The full text on this page is automatically extracted from the file linked above and may contain errors and inconsistencies.
CO o GO O c 03 February 1980 Vol. 62, No. 2 QQ CD > CD CO CD Od "cU a5 TD CD 3 On the Costs and Benefits of Anti-Inflation Policies 15 Access to the Discount Window for All Commercial Banks: Is It Important for Monetary Policy? 25 The New M onetary Aggregates The R e v i e w is published monthly by the Research Department of the Federal Reserve Bank of St. Louis. Single-copy subscriptions are available to the public free of charge. Mail requests for subscriptions, back issues, or address changes to: Research Department, Federal Reserve Bank of St. Louis, P.O. Box 442, St. Louis, Missouri 63166. Articles herein may be reprinted provided the source is credited. Please provide the Bank’s Re search Department with a copy of reprinted material. On the Costs and Benefits of Anti-Inflation Policies LAURENCE H. MEYER and ROBERT H. RASCHE A l \ prominent policy issue of the 1970s and one that seems certain to dominate the early 1980s is the appropriate response to a prevailing high rate of in flation. The view that there is a long-run trade-off between inflation and unemployment, widely accepted at the end of the sixties, is now held by only a small minority of economists. It is still widely believed, however, that there is a short-run trade-off between inflation and unemployment, which implies that re strictive macroeconomic policies designed to reduce inflation would cause a temporary rise in the unem ployment rate. Therefore, both the time pattern of the response of inflation and unemployment to demand management policies and the relative cost of inflation and unemployment remain dominant issues in the design of macroeconomic policy. There is relatively little consensus on either the costs or benefits of reducing inflation. Both income-expenditure and monetarist macroeconometric models indicate that a deceleration in monetary growth would gradu ally eradicate inflation but at a sizable cost in terms of foregone output. In contrast, recent theoretical anal ysis based on “rational expectations” suggests that the cost of reducing inflation could be small. Moreover, the literature contains still less information on the cost of inflation, which makes it difficult to obtain a careful balancing of the costs and benefits of policies intended to reduce or eliminate inflation. This paper develops three views of the dynamics of inflation and unemployment: the expectations-augmented Phillips Curve model, a monetarist model of the relation of monetary change to both inflation and unemployment, and a rational expectations model. Based on each of these models, estimates of the cost of reducing inflation are presented. Finally, the size of the permanent per period gains associated with eradicating inflation that would justify incurring these temporary costs are estimated using both the Phillips Curve and monetarist models. THREE VIEWS OF THE RELATION HETWEEN INFLATION AND UNEMPLOYMENT The Phillips Curve approach, as employed in virtually all large-scale, income-expenditure macro econometric models, relates inflation to the unemploy ment rate and inflation expectations and almost uni formly specifies inflation expectations as dependent exclusively on past actual inflation rates. Some mone tarists, however, prefer to relate both inflation and unemployment directly to monetary change and reject the regularity between inflation and unemployment embodied in the Phillips Curve. A third view accepts the Phillips Curve but introduces an alternative hy pothesis about expectations formation. This rational expectations approach yields conclusions quite differ ent from the first two approaches. The Phillips Curve The Phillips Curve relates the rate of change in nominal wages to both the unemployment rate and the rate at which wages and/or prices are expected to rise. This is usually combined with a mark-up model of price determination in which prices are related to wages. A combination of the two hypotheses yields a relation between inflation (p ), unemployment (u), and expected inflation (pe): (l ) p = a + P u + p*, |3 < 0. There are two basic sources of inflation identified in equation 1: a demand factor and an expectations factor. The unemployment rate is a proxy for the 3 F E D E R A L R E S E R V E B A N K O F S T . L O U IS balance between supply and demand in the labor market. The lower the unemployment rate, the greater the demand relative to the supply of labor. When there is excess demand for labor, wages are bid up at a rate proportional to the degree of excess demand. Workers and firms bargain directly about nominal wages, but labor supply and demand depend on the real wage rate.1 Hence, the bargaining for nominal wage increases over any period will also depend on the rate of inflation expected over that period. According to the Phillips Curve perspective, de mand management policies affect inflation by affect ing aggregate demand and, hence, unemployment. While in principle economic policy could affect in flation expectations directly, the specification of the Phillips Curve in macroeconometric models generally assumes that inflation expectations are formed adap tively, that is, they depend exclusively on past infla tion rates. A simple form for such an equation is: ( 1 ') p t = a + p ut + pt-i, where pt is the rate of inflation over some period, ut is the average unemployment rate over the period, and pt_! is last period’s inflation rate and, hence, this period’s expected rate. The role of lagged inflation in the Phillips Curve may also reflect the direct influence of past wage and price changes on current inflation due to catch-up effects and long-term contracts. Contracts that fix wages over some period, typically from one to three years, permit current wage settlements to influence wages and prices over the duration of the contract, building an element of persistence or inertia into the inflation process.2 In the specification given by equation 1', there exists a critical unemployment rate consistent with either price stability or constant inflation.3 Setting pt = pt-i, the critical value of u is u* = -a/p. This is often referred to as the natural rate of unemployment or the nonaccelerating inflation rate of unemployment JIf escalator clauses were both universal and complete, bar gaining would be in terms of real wages. The existence of partial escalator clauses for some workers speeds the response of wage to price change, but doesn’t alter the fact that nom inal wage bargains must directly reflect inflation expectations over the duration of the contract. 2See John B. Taylor, “Staggered W age Setting in a Macro Model,” American E conom ic Review, Papers and Proceedings (M ay 1 9 7 9 ), pp. 108-13, for a model that incorporates both forward-looking and backward-looking elements in the wage setting decisions. 3The existence of an equilibrium or natural rate of unemploy ment independent of the rate of inflation depends on the co efficient of inflation expectations in equation 1 or lagged in flation in 1' being equal to unity. 4 FEBRUARY 1980 Figure 1 (NAIRU). Anti-inflation policy operates by raising u above u°; as long as u is greater than u°, inflation de celerates and ultimately is eradicated. Figure 1 depicts this cycling down process. First, policy reduces aggregate demand. This raises u above u° and induces a deceleration in inflation (link 1). The decline in the actual inflation rate, in turn, re duces inflation expectations via the p i term (link 2), which further reduces actual inflation in the next period (link 3). As long as u remains above u°, this cycling down continues. Ultimately, u returns to u* when inflation has been fully eradicated. Thus, eradi cating inflation requires a temporary rise in the un employment rate during the transition to price stability. Monetarist Reduced-Form Equations Stein has developed a “monetarist” framework for assessing the relation between inflation and unemploy ment.4 Stein’s basic inflation equation can be expressed as: ( 2 ) p t - pt-i = a ( m t - pt - i ) , where mt is the rate of monetary growth in period t. According to this specification, inflation accelerates when monetary growth exceeds the previous period’s rate of inflation. The distinctive feature of the monetarist equation is not that it identifies monetary growth as the key factor driving inflation while the Phillips Curve ig4Jerome L . Stein, “Inflation, Employment and Stagflation,” Journal o f Monetary Econom ics (April 1 9 7 8 ), pp. 193-228. Similar approaches have been presented by Carlson and Tatom: Keith M. Carlson, “Inflation, Unemployment, and Money: Comparing the Evidence from Two Simple Models,” this R eview (September 1 9 7 8 ), pp. 2-6; and John A. Tatom, “Does the Stage of the Business Cycle Affect the Inflation Rate?” this R eview (September 1 9 7 8 ), pp. 7-15. F E D E R A L R E S E R V E B A N K O F S T . L O U IS nores the influence of monetary change on inflation. The Phillips Curve itself is consistent with monetary change as a dominant influence on the inflation rate. However, it is a specification of the structure of the inflation process, that is, how monetary change affects inflation. According to the Phillips Curve interpreta tion, a decline in monetary growth moderates inflation by temporarily raising unemployment. The monetarist equation, in contrast, directly relates monetary growth to inflation and is essentially a reduced-fomi equation relating inflation to policy instruments. In contrast to the Phillips Curve, however, equation 2 allows onlij monetary change to affect inflation. It also appears to make the acceleration of inflation independent of de mand conditions in the economy. In fact, however, the unemployment rate itself is also affected by mone tary growth in the Stein model.5 Hence, in both the Phillips Curve and monetarist frameworks, a decline in monetary growth both increases unemployment and reduces inflation. Both approaches therefore permit us to calculate the temporary rise in unemployment asso ciated with anti-inflation policy.6 Equilibrium Models and Rational Expectations Recently there has been renewed interest in equi librium models in which monetary change results in an immediate proportional change in the price level.7 A useful point of departure here is a simple inflation reduced-form equation in which the inflation rate equals the rate of monetary growth: ( 3 ) p = m. This differs from traditional monetarist models in allowing the full effect of monetary change on prices to occur immediately. The Phillips Curve then deter mines the unemployment rate. It is convenient to rearrange equation 1 as: ( 4 ) u = u ° + 1/|3 (p - pe), which demonstrates that unemployment deviates from 5Stein’s unemployment rate equation is presented later when his model is used to derive the response of inflation and un employment to monetary change. 6Note, however, that inflation is expected to fall as long as u > u* according to the Phillips Curve; in the monetarist model, on the other hand, an acceleration in monetary growth induces an acceleration in inflation even if the unemployment rate initially is above its critical u“ level. Despite this dif ference, the two approaches yield similar implications for the time path of unemployment and inflation in response to a deceleration in monetary growth. 7Fo r example, see Thomas J. Sargent and Neil Wallace, “Ra tional Expectations and the Theory of Economic Policy,” Journal o f Monetary Econom ics (April 1 9 7 6 ), pp. 169-85. The model developed below is similar to the one developed by Sargent and Wallace. FEBRUARY 1980 its natural rate only in response to unanticipated in flation. Monetary change determines the actual inflation rate (via equation 3). Inflation expectations, accord ing to the rational expectations view, are formed on the basis of the relevant economic theory — in this case on the relevant model of the inflation process — and are conditional on all relevant available informa tion. Taking the expected value of inflation from equation 3, ( 5 ) p* = E ( p ) = E ( m ) , where E(m ) is the expected rate of monetary growth conditional on information available prior to the period over which the expectations apply, equation 4 can be rewritten as: ( 4 ') u = u* + 1 /p (m - E ( m ) ). This implies that unemployment is affected only by unanticipated monetary change. In the previous two models, a deceleration in mone tary growth reduces the growth in nominal demand, but inflation unwinds gradually; the decline in the growth of nominal demand, therefore, initially falls heavily on real demand and, hence, initially reduces output and employment. In this model, in contrast, monetary deceleration increases unemployment only if the monetary deceleration is unexpected. If it is an nounced in advance and reflected in expectations, a rapid deceleration of inflation results with no teno porary rise in unemployment and, hence, no cumu lative output loss. However, a question remains: How rapidly would the expected rate of monetary growth decelerate even if the Fed announced a policy of phased deceleration? Barro’s attempt to implement this model empirically holds that economic agents base their expecta tions of monetary growth in part on past monetary growth.8 In a sense, such a specification substitutes past actual rates of monetary growth for past actual inflation rates and, therefore, does not entirely re move the inflation inertia embedded in the traditional Phillips Curve. Meltzer also has recently emphasized the gradual process whereby expectations of future monetary growth respond to current observations of monetary policy actions.9 The overall framework, 8Robert J. Barro, “Unanticipated Money Growth and Unem ployment in the United States,” American Econom ic Review (M arch 1 9 7 7 ), pp. 101-15. 8Allan H. Meltzer;> “The Case for Gradualism in Policies to Reduce Inflation, ’ Stabilization Policies: Lessons from the 1970s and Im plications for the 1980s (St. Louis: Washington University, Center for the Study of American Business), forthcoming. 5 F E D E R A L R E S E R V E B A N K O F S T . L O U IS FEBRUARY 1980 C hart ] M oderation in Inflation: Perry Equation R ate o f W age Change Rate of W age Percent Change Pe rce n t 11.0 11.0 10.0 9.0 8.0 7.0 6.0 5.0 4.0 3.0 2.0 1.0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 1 Year however, suggests that the Fed can minimize the cumulative output loss by carrying out its anti-infla tion policy in a manner that makes it easy for the public to discern its intent; this means developing a reputation for meeting its monetary aggregate tar gets and reducing the volatility of monetary growth so it is easy to recognize changes in the target rate when they occur. THE RESPONSE OF INFLATION AND OUTPUT TO ANTI-INFLATION POLICY Results Based on Estimated Phillips Curves When developing an estimate of the cost of anti inflation policy using a Phillips Curve, anti-inflation “policy” is identified with an increase in the unem ployment rate above the critical rate. Underlying the change in the unemployment rate, but implicit in the analysis, are changes in monetary and fiscal policy instruments. As long as the unemployment rate is maintained above the critical rate, inflation will de celerate. Based on the assumed initial inflation rate and on the estimated parameters in the Phillips Curve, it is simple to calculate how long it will take to eradi cate inflation. The final step in estimating the cost of an anti-inflation policy is to convert the increased unemployment into a measure of the cumulative output loss. This 6 is done via Okun’s Law: Each 1 percentage point increase in the unemployment rate reduces real out put by 3.2 percent.10 Thus, at the 1978 value for potential output, for example, a 1 percentage-point rise in unemployment translates into a 45.6 billion dollar loss in output. The cumulative but undis counted loss can be found by assuming that potential output will rise at a 3.3 percent rate in line with pro jections by the CEA. The cumulative loss (L ) in this case is: (6)L = Z t = y ( u - u * ) (1 + p )', 0 where, Y = the Okun coefficient, u = the level of unemployment brought on by policy, u#= the critical unemployment rate, p = the rate of growth of potential output, and n = the number of years required to eradicate inflation. 10Arthur M. Okun, “Potential GNP: Its Measurement and Sig nificance,” from Proceedings of the Business and Economics Statistics Section of the American Statistical Association (1 9 6 2 ), pp. 98-104. More recent estimates of Okun’s Law suggest that the output loss might be only 2.5 percent for each 1 percent increase in unemployment. See, for example, the estimates in the St. Louis model in Leonall C. Andersen and Keith M. Carlson, “A Monetarist Model for Economic Stabilization,” this R eview (April 1 9 7 0 ), pp. 7-25, and Tatom’s discussion of Okun’s Law in “Does the Stage of the Business Cycle?” F E D E R A L . R E S E R V E B A N K O F S T . L O U IS FEBRUARY 1980 Chart 2 C u m u la tiv e O utput Loss: O utput Perry Equation O u tp u t Loss Loss lio n s o f 197 2 D o lla r s B illio n s o f 1972 D o lla rs 2,200 2,200 2,000 2,000 1 ,800 1 ,800 1 ,600 1,600 1 ,400 1,400 1,200 1,200 1,000 1,000 800 600 400 200 1 12 13 14 15 1 Year The discounted output loss is simply the product of the initial year’s loss and the number of years required to complete the program.11 The estimate of the cumulative output loss based on a Phillips Curve equation presented in this section is derived from the results of a study by Perry.12 Perry’s results are based on a wage change equation, using the inverse of his weighted unemployment rate and lagged wage change, estimated using annual ob servations over the 1954-77 period: (7 ) A in W = -1 .8 8 + 7.44 (1 /U w ) + 0.79 A l n W - + ( - 2 .2 ) ( 3 .5 ) (4 .6 ) 0.21 A in W -, + 1.07 DNIX SE = 0.70, ( 1 .1 ) (2 .9 ) where W is adjusted hourly earnings in the private nonfarm sector and DNIX is a dummy for the con trols equal to -1 in 1972 and 1973 and -f-1 in 1974 and 1975. His preferred equation yielded a NAIRU of 4.0 in terms of his weighted unemployment rate (which corresponds to about 5.5 percent in the of ficial unemployment rate in 1977). Hence, any un n This is the same as discounting future years’ losses by a 3.3 percent real interest rate. 12George L. Perry, “Slowing the W age Price Spiral: The Macroeconomic View,” Brookings Papers on Econom ic Activity ( 2 : 1 9 7 8 ), pp. 259-91. Similar estimates have been presented by Okun ancf Gramlich. Arthur M. Okun, “Efficient Disinfla tionary Policies,” American Econom ic Review, Papers and Proceedings (M ay 1 9 7 8 ), pp. 348-52; and Edward M. Gram lich, “Macro Policy Responses to Price Shocks,” Brookings Papers on Econom ic Activity ( 1 : 1 9 7 9 ), pp. 125-66. employment rate above 5.5 percent, if maintained long enough, would eradicate inflation. In the following simulations, inflation is assumed initially to be 7.5 percent, and the economy is as sumed initially to be at the critical unemployment rate (NAIRU). The time it would take to eliminate infla tion if unemployment were raised by either 1 or 3 percentage points is then calculated. The response of inflation to a rise in the unemploy ment rate and the accompanying cumulative output loss are depicted in charts 1 and 2: Perry 1 cor responds to a 1 percentage point rise in the unem ployment rate and Perry 2 to a 3 percentage-point rise. Beginning with Ain W (approximately the percentage change in the wage rate) equal to 10 percent, the unemployment rate is raised above NAIRU and held there until the rate of wage change declines to 2.5 percent, the rate presumed consistent with the trend growth in labor productivity and, hence, with price stability. When unemployment is raised 1 percentage point, the rate of change in the wage rate falls from 10 percent to 9.6 percent in the first year and declines about 0.3 percentage points per year thereafter, taking 23 years to reach the 2.5 percent rate consistent with zero price inflation. The undiscounted cumulative out put loss is $1.5 trillion, and the discounted cumulative output loss is $1 trillion. 7 F E D E R A L R E S E R V E B A N K O F S T . L O U IS FEBRUARY 1980 C hart 3 M o d e ra tio n in Inflatio n : St. Louis a n d Stein M o d e ls I n f l a t i o n Ra te I n f l a t i o n Rate Percent Percent 8.0 8.0 5 6 7 8 9 10 11 12 0 Year If unemployment is raised by 3 percentage points, inflation is eliminated after 11 years. The cumulative output loss, however, is greater in this case: $1.8 tril lion in the undiscounted case and $1.5 trillion in the discounted case.13 Results Based on a Monetarist Model According to Stein’s monetarist model, monetary change affects both inflation and unemployment. Stein’s two-equation model is: ( 8 ) Aut — 3.0 - 0.6 Ut-i + 0.4 p t-i - 0.4 m t~i ( 9 ) Apt = - 0.4 p t-i + 0.4 m u , If monetary growth remains constant, inflation con verges to the rate of monetary growth, and unemploy ment converges to a constant rate, equal to 5 percent in Stein’s model. Hence, the equilibrium rate of in flation equals the rate of monetary growth, and the critical unemployment rate is 5 percent. If monetary growth declines below the rate of inflation, inflation 13While most reported Phillips Curves yield high estimates of cumulative output loss in line with Perry’s, there are some that imply much lower estimates. Fo r example, see the Phillips Curve presented by Phillip Cagan in “The Re duction of Inflation by Slack Demand,” in William Fellner, Project Director, Contemporary Econom ic Problems in 1978 (Washington, D .C .: American Enterprise Institute for Public Policy Research, 1 9 7 8 ), pp. 13-45. The cumulative output loss based on Cagan’s equation is only about one quarter of that implied by Perry’s equation. decelerates and unemployment temporarily rises above its equilibrium rate. The simulation used to derive an estimate of the cumulative output loss from Stein’s model differs from that used in the Phillips Curve approach. Since inflation and unemployment are both linked directly to mone tary change in the monetarist model, the rate of mone tary growth can be used as the policy instrument. As sume that the rates of monetary growth and inflation are both 7.5 percent initially and that the economy is at the equilibrium unemployment rate. Anti-inflation policy is identified with a deceleration in the rate of monetary expansion. Now, consider two scenarios: a phased deceleration of monetary growth by 1 percent age point per year until the rate of monetary growth declines to a rate which, if maintained, would be con sistent with price stability (zero in this model) and an immediate deceleration to the rate consistent with long-run price stability. Imposing these alternative paths of monetary change on the model generates the associated implied paths of inflation and unemploy ment; the rise in unemployment above 5 percent is then translated into a measure of the cumulative out put loss. The effects of each policy on inflation and output loss are depicted in charts 3 and 4; the Stein 1 lines correspond to the gradual deceleration in monetary F E D E R A L R E S E R V E B A N K O F S T . L O U IS FEBRUARY C ha rt 4 C u m u l a t i v e O u t p u t Loss: O u t p u t Lo ss St. L o u i s a n d S t e in M o d e l s O u t p u t Loss B illio n s o f 1972 D o lla rs B illio n s o f 1972 D o lla rs growth, the Stein 2 lines represent the more aggres sive policy. Under the gradual policy, unemployment begins to rise in year 2, peaks in year 8 at 6.6 per cent, and returns to almost 5 percent by year 16. The inflation rate, on the other hand, begins to decelerate in year 2, initially at a 0.4 percentage point a year rate, ultimately reaching 1.0 point per year by year 7. The inflation rate is down to 2 percent by year 8 and thereafter declines gradually to about zero by year 16. The undiscounted cumulative output loss is $687.5 billion. Interestingly, the more aggressive pol icy incurs a smaller undiscounted output loss, $613 billion. Note that, qualitatively, the results are similar to those based on the Phillips Curve: Restrictive demand management policy temporarily will raise unemploy ment as it induces deceleration in the rate of inflation. The size of the cumulative output loss in the Stein model, however, is dramatically lower than that based on Perry’s equation. 1980 key equations in the St. Louis model are a reduced-form equation for the rate of growth in nominal income based on the Andersen-Jordan equation and an expectations-augmented Phillips Curve.15 The rate of monetary change is the principal de terminant of the rate of change in nominal income, although the rate of change in high-employment government expenditures also has a small, transitory effect. Thus, a decline in the rate of monetary growth is quickly translated into a decline in the rate of increase in nominal income. The distri bution of the latter decline between prices and output depends on the Phillips Curve; the slower the deceleration of inflation as nominal income falls, the greater the im pact of monetary change on output and the greater the resulting cumulative output loss of anti-inflation policy. To begin, a base run in which the rate of monetary growth is held steady at 7.5 percent from III/1968 through IV/1978 was generated. This builds in inflation in ertia and provides the base against which to evaluate the effects of gradual monetary deceleration. Begin ning in 1/1973, monetaiy growth was gradually decel erated by 1 percentage point in the first quarter of each year. The results from this policy run were then compared with those from the base run and the cumu lative output loss associated with this policy was derived by comparing the output solution assuming monetary growth remains at 7.5 percent per year with that assuming a phased monetary deceleration.16 The first set of simulations was generated using the St. Louis model estimated over the sample period I/1953-IV/1978. The results, labeled StLl, are re ported in charts 3 and 4. The inflation rate begins to decline slowly; indeed, it takes 2 years to reduce it by 1 percentage point. Thereafter, the deceleration speeds up; after 5V2 years, inflation has declined by 7.5 percentage points. The unemployment rate initially rises slowly, and the maximum increase is only 1.8 Results Using the St. Louis Model The St. Louis model is in some sense a compromise between the Phillips Curve and the monetarist reduced-form approaches developed above.14 The two 15Leonall C. Andersen and Jerry L. Jordan, “Monetary and Fis cal Actions: A Test of Their Relative Importance in E co nomic Stabilization,” this Review (November 1 9 6 8 ), pp. 11-24. 14The St. Louis model is described in Andersen and Carlson, “A Monetarist Model.” 16Because the model produces direct solutions for the response of output to monetary deceleration, the use of Okun’s Law is not required. 9 F E D E R A L R E S E R V E B A N K O F S T . L O U IS percentage points. The undiscounted cumulative out put loss is only about $200 billion.17 The estimate of the cumulative output loss in this case is dramatically lower than for either the Stein model or the Perry equation. However, the small size of the cumulative output loss reflects, in part, the suspiciously large coefficient on the demand slack variable in the model’s Phillips Curve — almost three times the size of the same coefficient estimated through 11/1971 or 1/1975, for example. This rise in the demand slack coefficient is neither readily ex plained nor mirrored in other estimates of Phillips Curves. Consequently, two additional runs with modi fied versions of the St. Louis model were made. First, a simulation of the model estimated through III/1971 in which the coefficient on the demand slack variable is substantially smaller was run. The results are depicted by the StL2 lines in charts 3 and 4. In flation decelerates much more gradually; after six years, it has declined by only 4 percentage points. The cumulative output loss, already at $350 billion, is escalating rapidly. In the second modified version of the St. Louis model, the Phillips Curve was replaced with a mone tarist reduced form for the inflation rate in which in flation depends on a 20-period distributed lag on the rate of change in the money supply.18 The lines la beled StL3 in charts 3 and 4 present the implications of gradual monetary deceleration on inflation and out put in this case. The results are remarkably similar to those generated by the first modified version of the St. Louis model (StL2 lines in charts 3 and 4). The inflation rate declines somewhat more rapidly and the output loss is a bit smaller, but both the time pattern and magnitude of inflation deceleration and output loss are very close. The StL2 and StL3 simulations were not run long enough to eradicate inflation and, therefore, are not directly comparable with the Phillips Curve and mone tarist reduced-form results. Nonetheless, the results 17In the Perry and Stein results, the initial level of potential output was that for 1978. The $200 billion cost estimate for the St. Louis model is based on an initial level of income in 1973. To make the St. Louis result comparable with the Perry and Stein results, it would be appropriate to multiply it by a factor equal to the ratio of potential output in 1978 to that in 1973. 18The inflation reduced form was provided by Tatom and is similar to the one he presented in “Does the Stage of the Business Cycle?” In addition to the distributed lag on mone tary change, it also includes a four-quarter distributed lag on the differential in the rates of change in producer prices for . energy and the price index for the nonfarm business sector, as well as two dummies, one for the effects of the freeze and Phase II and one for the subsequent catch-up effect. FEBRUARY 1980 at the end of six years were qualitatively similar to the Perry and Stein results: Anti-inflation policies impose a sizable cost in the form of lost output dur ing the transition to lower inflation rates.19 The Credibility Effect and Rational Expectations The Phillips Curve-based results reported above re lated inflation to a distributed lag of past inflation rates, which implies a gradual unwinding of inflation in response to anti-inflation policies. In this specifica tion, inflation expectations are formed exclusively on the basis of past actual inflation. This ignores the pos sibility that the public will adjust their inflation ex pectations to both recent policy actions and expecta tions about future policy. A well-defined, credible anti-inflation policy might induce a more rapid decel eration of inflation expectations than is suggested by the conventional equations. Fellner, for example, maintains that “. . . the standard model coefficients . . . would change significantly for the better — in the direction of a much more rapid rate of reduction of inflation for any given slack — if a demand manage ment policy . . . changed to a credible policy of con sistent demand disinflation.”20 But, by how much do standard econometric ap proaches overestimate inflation inertia and the associ ated cumulative output loss? Unfortunately, reliable quantitative estimates of the extent to which policy makers can speed the deceleration of inflation by clearly defining their anti-inflation policies and con vincing the public that they intend to follow through do not exist. Nevertheless, there is widespread agree ment that anti-inflation policies ought to be set out clearly and supported by both the Federal Beserve and the Treasury to maximize credibility. There are, however, two empirical applications of rational expectations macro models that provide some insight into the predictions of that approach for the response to a phased monetary deceleration. Paul A. 1!,Dewald recently presented simulations of the response of in flation, output, and unemployment to monetary deceleration based on a modified version of the St. Louis model. William G. Dewald, “Fast vs. Gradual Policies for Controlling Infla tion,” Federal Beserve Bank of Kansas City Econom ic R e view (January 1 9 8 0 ), pp. 16-25. He estimates the Phillips Curve in the rate of change as opposed to the first difference form used in the St. Louis model. This procedure does not yield a coefficient on the demand slack variable as high as in the St. Louis specification. Hence, Dewald also finds that monetary deceleration yields a large cumulative output loss. 20William Fellner, “The Credibility Effect and Bational E x pectations: Implications of the Gramlich Study,” Brookings Papers on Econom ic Activity ( 1 : 1 9 7 9 ), pp. 167-78. F E D E R A L R E S E R V E B A N K O F S T . L O U IS Anderson modifies the St. Louis model by respecifying its Phillips Curve to be consistent with rational expec tations.21 He begins with a Phillips Curve of the fol lowing form: ( 1 0 ) p = a + p x + ep* + e, where x is a measure of demand slack in the economy, and e is a random disturbance term with mean zero. Instead of specifying pe as a distributed lag on past actual inflation rates as in the equation in the St. Louis model, Anderson imposes rational expectations by set ting p equal to the expected value of inflation based on equation 10. Setting pe = E ( p), he solves for the expected value of inflation: (10-)E(P) = T^ +TL » Anderson uses this equation to determine the infla tion rate in the St. Louis model, based on the estimates of a , (3, and e from the St. Louis Phillips Curve. In particular he sets e = .86. This procedure, in effect, dramatically raises the response of inflation to changes in demand slack. He runs simulations of the response to an acceleration in the rate of monetary growth. In the original St. Louis version, inflation increases grad ually and unemployment declines; in the rational ex pectations version, inflation increases more rapidly and the effect on unemployment virtually disappears. If e is viewed as the coefficient on expected infla tion, however, it seems inappropriate to employ its value of .86 as estimated in the St. Louis model in the rational expectations version of the St. Louis model because it was estimated originally under the assump tion that expectations are formed adaptively. Taking e = 1, as seems essential to the rational expectations framework, equation 10' is no longer a meaningful equation for p. Instead, setting pe = E(p ) and solv ing for E (p ), we obtain: ( 10" ) 0 = c c + P x , which indicates that there is a unique value of the demand slack variable (x ° = -a/ p), corresponding, of course, to the natural rate of unemployment. Only random disturbances (with zero mean) can cause x to differ from x*. In this case, the impact of monetary deceleration on the rate of growth of nominal income is transformed immediately and fully into a decline in inflation without any cumulative output loss. An alternative empirical application of a rational expectations macro model is presented by Barro.22 He -’■Paul A. Anderson, “Rational Expectations Forecasts from Nonrational Models,” Journal o f Monetary Econom ics (Jan uary 1 9 7 9 ), pp. 67-80. 22Barro, “Unanticipated Money Growth.” FEBRUARY 1980 relates deviations in unemployment from its natural rate to unanticipated monetary change ( as in equation 4' above) and, in addition, provides a model describ ing how economic agents form expectations about the rate of monetary growth. On the surface at least, Barro’s model seems well suited to provide an estimate of unemployment’s response to a policy of phased deceleration in monetary growth. The crucial issue here is how rapidly economic agents learn that the policy rule has in fact changed. They may learn this from an announcement by the Fed. Given some doubt about the Fed’s commitment to follow through on any announced deceleration, however, economic agents may insist on learning the new policy rule by observ ing the new pattern of monetary growth rates. This involves reestimating the policy rule and incorporat ing new observations each period. Eventually, eco nomic agents will learn that the Fed intends to decelerate monetary growth and then stabilize it at a noninflationary rate. But this learning process may take some time; meanwhile, monetary change will be less than expected and unemployment will exceed the natural rate. Hence, the Barro model also allows for the existence of a cumulative output loss during the transition to price level stability.23 The survey above provides the following cost esti mates: the extremely large estimate of the cumulative output loss based on Perry’s Phillips Curve, the smaller but still sizable loss based on Stein’s monetarist model, the evidence from simulations with the St. Louis model which, on balance, also suggest a large output loss, and the rational expectations results as modeled by Anderson, which suggest virtually no out put loss if monetary deceleration is perfectly antici pated. Unfortunately, in addition to the uncertainty surrounding the actual cumulative output loss likely to be associated with anti-inflation policy, there is also uncertainty about the benefits to be derived from eradicating inflation. BALANCING THE GAINS AGAINST THE COSTS The cumulative output loss is a measure of the costs of anti-inflation policies. To evaluate the desirability of such policies, an assessment of the gains from re ducing inflation is required. Unfortunately, the costs of 23This is an application of the learning mechanism emphasized by John B. Taylor, “Monetary Policy During a Transition to Rational Expectations,” Journal o f Political Economy (O c tober 1 9 7 5 ), pp. 1009-21; and by Benjamin M. Friedman, “Optimal Expectations and the Extreme Information Assump tions of ‘Rational Expectations’ Macromodels,” Journal of Monetary Econom ics (January 1 9 7 9 ), pp. 23-41. 11 F E D E R A L R E S E R V E B A N K O F S T . L O U IS inflation (and hence the benefits of reducing inflation) are not as clear cut or easily quantifiable as the costs of unemployment. Currently, no studies provide esti mates of the benefits that would accrue from reduc ing or eliminating inflation, which could in turn be compared directly to the cumulative output loss re quired to eradicate inflation. What can be computed, however, is the minimum size of the permanent gain in output per year due to the eradication of inflation that would justify incurring the cumulative output loss associated with the transition to price stability. The Costs of Inflation There are at least three dimensions to the costs of inflation.24 First, there are the costs associated with anticipated inflation that would be incurred even in a fully indexed economy where institutions have com pletely adapted to an inflationary environment. Second, there are the costs of anticipated inflation that arise from a set of institutions that have only partially adapted to the inflationary environment. Third, there are the costs associated with unanticipated inflation and uncertainty about the rate of inflation. A fully indexed economy is one in which all nominal payments and receipts (including wages, coupon pay ments, taxes, transfers, etc.) are tied (indexed) to the inflation rate. All debt instruments except for currency are indexed also. Currency is not indexed because in dexing it is assumed to be impractical. The major costs of anticipated inflation in a fully indexed economy can be labeled “menu” and “shoe leather” costs. Menu costs refer to the resource costs imposed by the necessity of frequent adjustments to published price lists in an inflationary economy. Shoe leather costs describe the costs incurred by more fre quent trips to the bank (or to the market) as a result of the incentive to economize on currency holdings.25 Recently there has been a growing emphasis on the costs imposed by inflation that reflect the existence of institutions that are not fully adapted to an infla tionary environment. The major source of these costs is the tax system, and the major effect is on saving and investment incentives and, therefore, on capital accumulation and the growth of output. The taxation 24The discussion of the costs of inflation in this section draws upon the recent survey by Stanley Fischer and Franco Modigliani, “Towards an Understanding of the Real Effects and Costs of Inflation,” W eltw irtschaftliches Archiv ( 4 : 1 9 7 8 ), pp. 810-33. 25For a discussion of the welfare cost of anticipated inflation, see John A. Tatom, “The Welfare Cost of Inflation,” this R eview (November 1 9 7 6 ), pp. 9-22. http://fraser.stlouisfed.org/ 12 Federal Reserve Bank of St. Louis FEBRUARY 1980 of personal interest income, for example, may induce a decline in the after-tax real rate of return to savers as inflation increases.26 Furthermore, the tying of the depreciation deduction for tax purposes to the histor ical rather than the replacement cost of capital goods tends to raise the cost of using capital goods in an inflationary environment. While inflation has many other effects on saving and investment, there is grow ing concern that its net effect is to discourage both saving and investment. Unanticipated inflation imposes costs by inducing redistributions of income and wealth. These “transfer effects” arise because contracts have been written in nominal terms embodying expectations about future inflation which turn out to be incorrect. The social cost of such redistributions is difficult to assess because there is a gainer for every loser. However, many con sider the “transfer” costs associated with unanticipated inflation the most serious cost associated with inflation.27 Uncertainty about the inflation rate may impose ad ditional costs by increasing the uncertainty associated with the outcome of economic decisions. Above, a cost was ascribed to the actual redistributions that follow from unanticipated inflation. There may also be utility losses associated with the increased likelihood of such arbitrary transfers when there is considerable uncer tainty about expected inflation. A number of studies have suggested that inflation uncertainty tends to be systematically related to the level of inflation. If this is the case, reducing the level of inflation will also re duce inflation uncertainty.28 Fischer and Modigliani do not provide estimates for the various effects of inflation since “any measures 26For a discussion of the effect of taxation of interest income on the response of nominal interest rates to a change in expected inflation, see Martin Feldstein, “Inflation, Income Taxes, and the Rate of Interest: A Theoretical Analysis,” American Econom ic Review (Decem ber 1 9 7 6 ), pp. 809-20. 27Franco Modigliani and Lucas Papademos, “Optimal De mand Policies Against Stagflation,” W eltw irtschaftliches Archiv ( 4 : 1 9 7 8 ), pp. 736-82. 28See, for example, Arthur M. Okun, “The Mirage of Steady Inflation,” Brookings Papers on E conom ic Activity (2 :1 9 7 1 ) , pp. 485-98; and Dennis E . Logue and Thomas D. Willett, “A Note on the Relation Between the Rate and Variability of Inflation,” Econom ica (M ay 1 9 7 6 ), pp. 151-58. These studies generally associate an increase in inflation uncertainty directly with an increase in the welfare cost of inflation. This follows only if increased uncertainty about inflation increases uncertainty about real income, real wealth, etc. For a dis cussion of the relation of inflation uncertainty to the welfare cost of inflation, see Lionel Kalish III, Laurence H. Meyer, and David H. Resler, “Inflation Uncertainty and the Welfare Cost of Inflation,” mimeographed (Federal Reserve Bank of St. Louis, 19 8 0 ). F E D E R A L R E S E R V E B A N K O F S T . L O U IS FEBRUARY 1980 F ig u r e 2 Output Table 1 The Minimum Value of the Per Year Gain (g) That Justifies Eradicating Inflation (billions of 1972 dollars) would be totally speculative at this stage.”29 In the absence of a measure of these costs, however, it is possible to compute the minimum total costs associ ated with continued inflation necessary to justify in curring the previously calculated costs of eradicating inflation.30 Evaluating the Minimum Gain Per Year Required to Justify Policies to Eradicate Inflation The solid line (X ) in figure 2 is the rate of growth of potential output if inflation remains indefinitely at 7.5 percent. If anti-inflation policies are pursued, out put is assumed to follow the dashed line (X '). The transitional costs of eradicating inflation occur be tween t0 and tn as unemployment rises above the rate associated with potential output. The cost of inflation may involve decreases in po tential output due to disincentives to saving and/or investment and/or welfare losses due to anticipated or unanticipated inflation. The benefit of eradicating inflation is shown in figure 2 as an increase in output above the level that would have prevailed had infla tion continued to average 7.5 percent; hence, the 29Fischer and Modigliani, “Towards an Understanding of the Real Effects and Costs of Inflation,” p. 813. 30This approach was suggested to us by Jerry Jordan and Allan Meltzer. Equation/ model Value of g Perry 1 $73.0 Perry 2 70.9 Stein 1 31.0 Stein 2 25.4 dashed X' line rises above the solid X line after tn.31 This analysis emphasizes the necessity of comparing the transitional cost incurred over the period during which inflation is eradicated with the permanent bene fit attributable to the eradication of inflation. G is the present value of the permanent per period output gain, evaluated from period n to oo: To simplify further, gt is assumed to be constant for all t > n. The value of g, which equates the dis counted cost of unemployment and the gain from eradicating inflation is then calculated. This is the minimum value of the permanent per period gain from eradicating inflation that would justify incurring the transitional costs. The values of g, based on transi tional costs estimated from the Perry equation and Stein model and calculated under the assumption of a 3.3 percent discount rate, are presented in table 1. The minimum value of g varies from about $25 billion per year in the Stein model to $73 billion based on Perry’s Phillips Curve under a moderate policy. This analysis provides an alternative perspective on the case for gradualism. A gradual policy will im pose a lower cost of eradicating inflation if the Phil lips Curve is nonlinear. For Perry’s nonlinear Phillips Curve, for example, the discounted cost was $1.0 bil lion for the gradual policy and $1.5 billion for the 31The gains of reducing inflation should begin being regis tered during the transition period. To simplify the calcula tion, the benefits are assumed to begin only at t„, when infla tion is eradicated. This assumption biases the calculation of the present value of benefits downward. 13 F E D E R A L R E S E R V E B A N K O F S T . L O U IS more aggressive policy. A more gradual policy, how ever, also postpones the benefits from eliminating in flation. The size of the permanent per period gain required to justify the anti-inflation policy may there fore be smaller under the more aggressive policy. In deed, this is the case for the Perry results. Although the cumulative output loss is smaller under the grad ual policy (Perry 1), the size of the per period gain required to justify eradicating inflation is smaller under the more aggressive policy (Perry 2). The more aggressive policy also vields a smaller minimum per period gain using the Stein model, although this was expected since the cost turned out to be lower in the radical case using Stein’s model. The calculations reported above presumed that the gains from reducing inflation could be meaningfully represented as a fixed real sum per period. Suppose, instead, that the gains are more meaningfully specified as a real sum which grows at the same rate as poten tial output. For example, the cost of a fully anticipated increase in inflation is generally measured by the re duction in the area under the demand curve for real money balances as wealthowners reduce their demand for money in response to the associated rise in nomi nal interest rates. The decline in demand for real money due to a rise in the interest rate is generally considered proportional to the overall scale of money holdings which, in turn, is determined by the level of transactions (e.g., real income). The cost of a given rate of inflation and, hence, the benefits of eliminating inflation may, therefore, grow at the rate of increase of potential output. In this case: http://fraser.stlouisfed.org/ 14 Federal Reserve Bank of St. Louis FEBRUARY 1980 where g is the value of the gain in period n (the first period in which a gain is registered). For p > r, G oo. This corresponds to the result recently derived by Feldstein: If the cost of inflation grows at a rate equal to or greater than the discount rate, any positive initial gain (any g > 0) is sufficient to justify incurring any finite transitional cost!32 CONCLUSION The existence of large transitional costs of eradicat ing inflation is not a sufficient reason to reject anti inflation policies. The potential existence of large benefits associated with reducing inflation requires a careful assessment of the net effects of anti-inflation policies. Unfortunately, the range of the estimates of the cumulative output loss, the uncertainty about the adjustment in those results required to allow for the credibility effect, and the lack of quantitative esti mates of the cost of inflation make it extremely diffi cult to obtain a meaningful comparison of the costs and benefits of an anti-inflation policy. Narrowing the range of estimates of output loss and developing a measure of the benefits associated with anti-inflation policies should be high on the priorities for macroeconomic research in the 1980s. :i-Martin S. Feldstein, “The Welfare Cost of Permanent Infla tion and Optimal Short-Run Economic Policy,” Journal of Political Econom y (August 1 9 7 9 ), pp. 749-67. Access to the Discount Window for All Commercial Banks: Is It Important for Monetary Policy? R. ALTON GILBERT ROPONENTS of legislation to induce or require more banks to be members of the Federal Reserve System often argue that the existence of nonmember commercial banks creates problems for the conduct of monetary policy. Two of the most frequently men tioned of these problems are: First, the ratio of net demand deposits to bank reserves may become more variable as nonmember banks have a larger share of demand deposit liabilities.1 Second, as banks with draw from membership, the average ratio of net de mand deposits to reserves rises. With a lower average reserve base, fluctuations in reserves due to such factors as changes in float and currency holdings of the public make net demand deposits more variable.2 These potential problems have been subjected to ex tensive theoretical analysis. Recently, a third possible problem with declining membership has been suggested — the fact that non member banks do not have access to the discount window on a day-to-day basis. Lack of direct access to the discount window for all commercial banks, it has been alleged, may preclude the adoption of appro priate monetary policy because of the Federal Re serve’s concern about the differential impacts of these policies on member and nonmember banks. During periods of tight monetary policy, for example, declin1Dennis R. Starleaf, “Nonmember Banks and Monetary Con trol,” Journal o f Finance (September 1 9 7 5 ), pp. 955-75; and Kenneth J. Kopecky, “Nonmember Banks and Empirical Measures of the Variability of Reserves and Money: A Theo retical Analysis,” Journal o f Finance (M arch 1 9 7 8 ), pp. 31118. 2See, for example, J. A. Cacy, “Reserve Requirements and Monetary Control,” Federal Reserve Bank of Kansas City Monthly R eview (M ay 1 9 7 6 ), pp. 3-13. ing Federal Reserve membership might increase the liquidity risk for the entire banking system, since fewer banks would be able to use the discount win dow to provide a temporary offset to unexpected reserve outflows.3 The validity of this argument hinges on whether credit from the discount window significantly would help nonmember banks adjust to deposit withdrawals. Under existing legislation, the Federal Reserve has the authority to lend to nonmember banks in unusual emergencies in which these banks would fail without additional reserves.4 Increasing the number of banks that are members, therefore, would not increase the ability of the Federal Reserve to respond to such emergency situations. 3G. William Miller, “Statement,” Monetary Control and the M em bership Problem , U.S. Congress, House, Committee on Banking, Finance, and Urban Affairs, 95th Congress, 2nd Ses sion, July 27, 1978, pp. 60-62; and statements by Sen. William Proxmire and Paul Volcker, Chairman of the Federal Reserve Board, before the U.S. Senate Committee on Banking, Hous ing, and Urban Affairs, February 4, 1980. 4Credit from Federal Reserve Banks is classified as reserve ad justment credit, seasonal credit, and emergency credit. Adjust ment credit is available for member banks to meet unexpected temporary credit demands caused by sudden deposit with drawals or unanticipated loan demand. Seasonal credit is avail able to relatively small member banks that have seasonal patterns in their deposits and loans. Emergency credit may be made available to member or nonmember banks with severe financial difficulties. For additional information on the condi tions under which the Federal Reserve makes credit available to banks, see R. Alton Gilbert, “Benefits of Borrowing from the Federal Reserve when the Discount Rate is Below Market Interest Rates,” this R eview (M arch 1 9 7 9 ), pp. 25-32. This paper analyzes use of adjustment credit by member banks. For an analysis of seasonal borrowing, see Stanley L . Graham, “Is the Fed’s Seasonal Borrowing Privilege Justified?” Federal Reserve Bank of Minneapolis Quarterly Review (F all 1 9 7 9 ), pp. 9-14. 15 F E D E R A L R E S E R V E B A N K O F S T . L O U IS The type of monetary policy that the Federal Re serve may be precluded from adopting out of concern for liquidity pressures on nonmember banks is pre sumably not that designed to create liquidity emer gencies for the banking system. Concern that declin ing membership would increase liquidity risk to the banking system suggests, rather, that nonmember banks have greater difficulty than member banks in adjusting to unanticipated deposit withdrawals or changes in demands for credit in more normal cir cumstances than those emergencies in which the Federal Reserve would make credit available to non member banks. Since monetary policymakers are jus tifiably concerned about possible influences of Federal Reserve membership on the conduct of monetary policy, it is important to clarify whether lack of direct access to the discount window for nonmember banks does, indeed, pose a serious problem for monetary policy. This article takes an indirect approach to determin ing whether nonmember banks have greater reserve management difficulties due to lack of access to the discount window. If borrowing from the discount window is a useful way to adjust to unanticipated reserve outflows, member banks would borrow from the discount window on at least a few occasions each year, taking advantage of their regular, reliable, dayto-day access to the discount window to meet unex pected withdrawals or to cushion themselves against temporary liquidity pressures. If, however, most mem ber banks manage their reserve positions without borrowing at the discount window, it is doubtful that nonmember banks have significantly greater difficulty than members in responding to similar reserve out flows. In this case, lack of access to the discount win dow for nonmember banks is simply irrelevant to the membership issue. DO MEMBER BANKS BORROW FREQUENTLY FROM THE DISCOUNT WINDOW? Most member banks do not borrow from the dis count window; those few that do so generally bor row infrequently. From 1974 to 1977, the proportion of the 430 member banks in the Eighth District that borrowed in any one year was as high as 25 percent only in 1974, when the discount rate was substantially below alternative short-term interest rates.5 Of the 115 member banks that borrowed during 1974, only 21 5Fo r further discussion of member bank borrowing, see Gilbert, “Benefits of Borrowing from the Federal Reserve.” 16 FEBRUARY 1980 did so on more than 10 occasions, and only 55 bor rowed five or more times (table l ) . 6 During 1975, only nine banks borrowed five times or more, and in 1976 only 10 banks borrowed that frequently. The infrequent borrowing of member banks from the Federal Reserve indicates that, in most circum stances, they adjust to reserve losses without resort to the discount window. This is sufficient evidence for rejecting the view that nonmember banks necessarily have greater problems than member banks in coping with reserve outflows because they lack direct access to the discount window. If borrowing from the dis count window were an important means for banks to adjust their reserve positions to reserve outflows, most member banks would borrow from the discount win dow.7 Yet, they do not do so. HOW DO MEMBER BANKS ADJUST TO RESERVE DRAINS? Most member banks manage their reserve positions by means other than borrowing at the discount win dow. This has implications for the significance of the discount window for the banking system’s operation. Effects of Deposit Fluctuations on the Reserve Positions of Member Banks Cash management by member banks is investigated by considering the factors that determine their de sired cash holdings, the types of events that cause their cash holdings to be different from desired levels, ^Although Reserve Banks classify borrowing by member banks as reserve adjustment or seasonal credit, it is often difficult to distinguish the purpose of borrowing in actual practice. Member banks that have the privilege of receiving seasonal credit often change the amount of seasonal credit they borrow daily or weekly. Use of the discount window for reserve adjustment is gauged in this article by the number of occa sions on which a bank borrows from the discount window dining a year, measured as the number of times when borrowing is positive followed by periods when borrowing is zero. Member banks that use the discount window for reserve adjustment on a routine basis tend to borrow on several occasions each year. "This conclusion is reinforced by noting that Fed reserve re quirements are binding for most member banks, in the sense that they hold larger cash balances than they would in the absence of Fed reserve requirements, whereas state reserve requirements are not binding for most nonmember banks. In most states, required cash reserves of nonmember banks are substantially smaller than the cash reserves nonmember banks actually hold. See R. Alton Gilbert, “Effectiveness of State Re serve Requirements,” this R eview (September 1 9 7 8 ), pp. 1628. If most member banks do not borrow at the discount win dow to offset reserve drains in order to meet binding reserve requirements, nonmember banks would be even less likely to borrow at the discount window, if made available to them, unless they were made subject to member bank reserve requirements. F E D E R A L R E S E R V E B A N K O F S T . L O U IS FEBRUARY 1980 Table 1 Frequency of Borrowing from the Federal Reserve by Member Banks Bank size (Total deposits in millions of dollars) Number of Eighth District member banks that borrowed on the following number of occasions each year 2-4 5-10 11-15 16-20 Over 20 1 Tot®1 ofT a n k s borrowing 1974 $ 0 - $ 10 7 8 7 8 0 0 30 10- 25 14 21 8 0 1 0 44 25- 18 50 3 2 11 2 0 0 5 0 - 100 2 1 4 1 0 1 9 Over 100 1 1 4 2 3 3 14 1975 $ 0 - $ 10 2 2 1 0 1 0 6 10- 25 2 8 1 1 0 0 12 25- 50 6 4 1 1 0 0 12 50- 100 3 2 0 0 1 0 6 Over 100 4 2 0 2 0 0 8 $ 0 - $ 10 2 2 1 0 0 6 1976 1 10- 25 4 0 1 1 0 0 6 25- 50 2 3 5 0 0 0 10 5 0 - 100 1 1 0 0 1 0 3 Over 100 4 4 0 0 0 0 8 $ 0 - $ 10 1 5 1977 1 0 0 1 8 10- 25 3 4 4 0 1 0 12 25- 17 50 5 6 3 3 0 0 5 0 - 100 2 4 2 0 0 0 8 Over 100 2 5 7 3 1 0 18 and their response to deviations of actual from de sired cash balances. Cash balances of member banks can be classified into three components: vault cash, reserve balances at Federal Reserve Banks, and demand balances due from correspondents. The amount of vault cash banks desire to hold is based upon their expectations of depositors’ demand for currency. Reserve balances held by most member banks at the Federal Reserve are determined by their required reserve balances, which are based upon their deposit liabilities and vault cash held two weeks earlier. Required reserve balances change each week, and member banks must meet their required reserves on a weekly average basis. Finally, demand balances of member banks due from correspondents are determined by both the volume of transactions through those accounts and the balances their corres pondents require as compensation for services pro vided. Correspondent banks generally do not charge respondents explicit fees for services; they require, instead, that respondents hold certain average de mand balances with them.8 The primary cause of deviation between a bank’s actual and desired cash balances is unanticipated fluc tuation in demand deposit liabilities. Time depos its mature on specific dates, and therefore bank man agement can anticipate when it must be ready to 8Robert E . Knight, “Correspondent Banking Part III: Account Analysis,” Federal Reserve Bank of Kansas City Monthly Review (December 1 9 7 1 ), pp. 3-17. 17 F E D E R A L R E S E R V E B A N K O F S T . L O U IS make payments to holders of time deposits. Demand deposit liabilities, however, fluctuate more from day to day and are, therefore, more difficult to estimate. The effects of unanticipated loan demand on banks’ reserve positions are reflected in changes in their demand deposit liabilities. When banks make loans, they increase the demand deposit balances of borrowers. If borrowers did not withdraw those de posits to make payments, there would be no dis turbance in the current week to the reserve positions of banks making loans. Increases in loans cause banks to lose reserves only when borrowers withdraw deposits. Bank customers withdraw deposits from their de mand accounts either by demanding currency or by writing checks. When depositors demand currency, a bank’s assets (vault cash) and liabilities (demand deposits) decrease simultaneously. Member banks lose reserves when depositors demand currency, but due to lagged reserve accounting, this decline in vault cash does not affect reserves available to meet reserve requirements in the current week. Likewise, the decline in demand deposit liabilities that results when depositors demand currency does not affect required reserves in the current settlement week, but two weeks later instead.9 Therefore, member banks may choose to make no initial response to a loss of reserves in the form of vault cash, unless total vault cash falls below some critical level at which the risks of temporary currency shortages become sufficiently great. When customers withdraw demand deposits by writing checks, the effect on a member bank’s cash assets depends upon how the bank clears checks and how it pays for checks drawn on its customers’ ac counts. Banks receive checks as their customers make deposits and, therefore, need a mechanism for collect ing checks drawn on other banks. Some member banks receive payment for checks by having their reserve balance at their Reserve Bank credited, while paying for checks drawn on their depositors’ accounts by having their reserve balances debited. Frequent debits and credits to member banks’ reserve accounts cause these balances to change on most business days. 9Required reserves of a member bank in the current settlement week are based upon its deposit liabilities two weeks earlier. The reserves that count toward meeting a member bank’s required reserves in the current settlement week are average daily vault cash held two weeks earlier and average daily reserve balances at the Fed in the current week. Fo r a com parison of how contemporaneous and lagged reserve account ing affects the reserve management of member banks, see R. Alton Gilbert, “The Effects of Lagged Reserve Requirements on the Reserve Adjustment Pressure on Banks,” Financial Analysts Journal ( September-October 1 9 7 3 ), pp. 34-43. Digitized for 18 FRASER FEBRUARY 1980 If a member bank clears checks through its reserve account, net withdrawals of deposit liabilities cause its reserve balance to decline. Required reserves for the current week, however, are not affected by de posit withdrawals in the current week due to lagged reserve requirements. Thus, if its initial reserve bal ance just equaled its required balance for the cur rent week, deposit withdrawals during the week would cause a member bank’s reserve balance to be deficient. Borrowing through the discount window is one means of increasing reserves on short notice. Other member banks clear checks through their bal ances at correspondents. They deposit checks with their correspondents and pay for checks drawn on depositors’ accounts by having their demand bal ances due from correspondents debited, including checks presented for payment by the Federal Reserve. If these banks do not use other Reserve Bank services that cause their reserve balances to change, these balances would change only when banks adjusted them to equal their required reserve balances.10 Thus, the reserve balances of member banks that use serv ices of correspondents tend to change less frequently than those of member banks that make greater use of Reserve Bank services. For a member bank that clears checks through a correspondent bank, net withdrawals of deposit liabil ities would not disrupt its balance between actual and required reserves for the current week. The decline in demand deposit liabilities instead would affect the bank’s required reserves in coming weeks. Since checks are cleared through the bank’s correspondent account, its reserve balance is unaffected by net deposit withdrawals. The response of these member banks to deposit withdrawals depends upon the constraints placed on them by their correspondents. If respAident banks were required to hold certain minimum amounts of demand balances at correspondents on a weekly av erage basis, they occasionally would have to obtain additional cash balances when experiencing net with drawals of deposit liabilities. Studies of correspondent banking, however, indicate that respondent banks may average their balances over longer periods of time, such as a quarter or even a year, in meeting the bal ance requirements of correspondents.11 Respondent 10Use of the following Reserve Bank services may involve debits and credits to reserve balances: wire transfers, coin and currency shipments, and collection of coupons on secur ities held in safekeeping and matured securities. n Knight, “Correspondent Banking.” FEBRUARY F E D E R A L R E S E R V E B A N K O F S T . L O U IS banks may have enough flexibility in managing their cash positions to simply let their demand balances due from correspondents fluctuate as their deposit liabil ities fluctuate with no additional response to deposit outflows. One method of determining whether respondent banks may temporarily reduce their demand balances due from correspondents by amounts equal to short term deposit outflows is to compare the dollar magni tude of changes in demand deposit liabilities to changes in their demand balances due from corre spondents. When large decreases in demand balances due from correspondents were compared to large de creases in gross demand deposit liabilities for 95 mem ber banks that obtain most of their services through correspondents, the ratio of the former to the latter averaged 1.04.12 Thus, large weekly decreases in de mand balances due from correspondents are of ap proximately the same dollar magnitude as large weekly decreases in gross demand deposits. These re sults indicate that respondents have enough short term flexibility in managing their cash positions that they can cope with relatively large decreases in their gross demand deposit liabilities by letting their de mand balances due from correspondents decline temporarily. Are Demand Deposit Withdrawals Large Enough to Induce Banks to Borrow? Only those member banks that clear checks through their reserve accounts have reductions in their re 12The 9 5 member banks, located in the Eighth District portions of Illinois and Missouri, had total deposits of less than $50 million in 1976. None of these banks cleared checks through the Reserve Bank regularly. These banks had so little ac tivity in their reserve balances at the Fed that the dollar amounts in their reserve balances remained unchanged for 150 days or more during 1976. Relatively large declines in the demand deposit liabilities at individual banks are meas ured by calculating the changes in gross demand deposits from each reserve settlement week to the next and deter mining the fifth largest decline. That amount is compared to the fifth largest weekly decline in demand balances due from correspondents, a measure of relatively large weekly declines in balances due from correspondents. An alternative comparison would be of the greatest weekly decline in de mand deposit liabilities to the greatest weekly decline in de mand balances due from correspondents. Such comparisons were not reported because, for many banks, the greatest weekly declines were substantially different from declines in other weeks and, therefore, not representative of the rela tively large weekly declines in these series. Examination of weekly declines in both series indicated that the fifth greatest weekly declines tend to be fairly representative of large de clines. Fo r instance, the third greatest weekly decline in demand balances due from correspondents divided by the third greatest decline in gross demand deposits averages 1.06. 1980 serve balances when experiencing net demand de posit outflows. These are the banks, therefore, that tend to borrow frequently from the discount window. Even for those banks that use their reserve balances for clearing checks, however, fluctuations in demand deposit liabilities may not be large enough to induce them to borrow to avoid reserve deficiencies. If fluc tuations in their reserve balances are smaller than their normal excess reserves, no response to declines in demand deposit liabilities is necessary. This issue is investigated by comparing average ex cess reserves to large decreases in demand deposit liabilities for a group of 102 member banks that make extensive use of Reserve Bank services.13 The fifth largest weekly decrease in gross demand deposits was larger than their average excess reserves in 1976 for all but one of these banks. Those relatively large decreases in deposits were, on the average, about 60 times larger than average excess reserves.14 There fore, excess reserves could not fluctuate by as much as demand deposits during at least several weeks each year. When demand deposit liabilities decline by relatively large amounts, member banks that clear checks through their reserve accounts must obtain additional reserves to avoid reserve deficiencies. Do Frequent Borrowers Clear Checks Through Their Reserve Accounts? The next step is to determine whether member banks that borrow from the discount window most frequently are, in fact, primarily among those banks that clear checks through their reserve accounts. A reliable indicator of whether a member bank uses its reserve balance for settlement in check collection is the frequency of changes in its reserve balance from day to day. Check collection is the only Reserve Bank l:iThe reserve balances of these member banks were unchanged on no more than five business days during 1976. 1*The ratio of the fifth largest weekly decline in gross demand deposits to average excess reserves was higher for larger banks, as indicated in the following display: Size group (Total deposits, in millions of dollars) $ Number of banks Sum of the fifth largest weekly decrease in gross demand deposits divided by the sum of average excess reserves 10 7 4.46 25 31 10.02 25- 50 33 84.88 50- 100 18 83.16 13 92.50 0 -$ 10- 100 and over 19 F E D E R A L R E S E R V E B A N K O F S T . L O U IS FEBRUARY 1980 Table 2 Comparison of the Frequency of Changes in Reserve Balances of Frequent Borrowers and Other Member Banks1 Size cateaorv (Annual average total deposits, millions of dollars) $0-$10 Distribution of banks by the measure of frequency of change in reserve balances at the Fed (Number °* cla^s ^ur‘n9 a year when a member bank’s reserve balances at the Fed were the same as on the previous day) Number 121- 131- 151- 176- 201251- 301 i ie of banks 120 130 ISO 200 250 300 & Over 175 Type of bank Up to 115 Infrequent borrowers 3 2 4 3 1974 5 3 1 2 1975 2 1976 1 3 5 18 8 1 1 2 7 53 Frequent borrowers in: 1977 $10-$25 Infrequent borrowers 2 2 1 2 1 1 17 15 7 9 8 10 1 1 5 14 10 2 82 Frequent borrowers in: $25 - $50 9 1974 7 1975 2 2 1976 2 2 1977 5 5 Infrequent borrowers 15 5 5 4 5 3 1 38 Frequent borrowers in: $5-$100 1974 12 1975 2 1976 4 1977 6 Infrequent borrowers 11 1 13 2 1 5 6 2 1 14 Frequent borrowers in: Over $100 1974 6 6 1975 1 1 1976 1 1 1977 2 2 0 0 1974 12 12 1975 2 2 1976 0 0 1977 11 11 Infrequent borrowers Frequent borrowers in: 1Banks designated as frequent borrowers are all Eighth District member banks that borrowed at the discount window on five or more occasions during a year. Those designated as infrequent borrowers are member banks in the Eighth District portions of Illinois and Missouri that did not borrow at the discount window in 1976 and were not frequent borrowers during 1974, 1975, or 1977. 20 F E D E R A L R E S E R V E B A N K O F S T . L O U IS FEBRUARY 1980 service that is likely to cause a member bank’s re serve balance to change each business day.15 row from the discount window frequently and others never do? The distribution of nonborrowers and frequent bor rowers by activity in their reserve balances is pre sented in table 2. Almost all of the Eighth District member banks that borrowed on five or more occa sions per year in 1975-77 had changes in their reserve balances on each business day. Incentive to borrow from the discount window was relatively great during 1974 due to the large spread between market interest rates and the discount rate. Several banks that bor rowed frequently in 1974 had relatively idle reserve balances. More of the frequent borrowers during 1974, however, had relatively active reserve balances than nonborrowers of comparable size.18 Thus, evidence in table 2 supports the hypothesis that banks which bor row frequently are primarily among those that have relatively active reserve balances. One possible answer is that, among member banks with relatively active reserve balances, the frequent borrowers have more highly variable deposit liabil ities, and thus are induced to borrow from the dis count window more frequently to replace reserves lost due to deposit withdrawals. Table 3 indicates that frequent borrowers in the years 1975-77 tend to have more highly variable demand deposit liabilities than other banks of comparable size with similar activity in their reserve balances.17 Why Are Only Some of the Member Banks with Active Reserve Balances Frequent Borrowers? Fluctuations in deposit liabilities at member banks that use correspondent services tend not to induce frequent, short-term borrowings from the discount window since those banks can adjust to decreases in deposit liabilities by letting their balances at corres pondents decline temporarily. Thus, banks that use services of correspondents may not be induced to borrow from any source in response to relatively large withdrawals of deposit liabilities. But among member banks that clear checks through their re serve accounts, and consequently have frequent changes in their reserve balances, why do some bor15The relation between use of Fed services and frequency of changes in reserve balances can be illustrated for member banks in the Eighth District portions of Illinois and Missouri, based on a survey of utilization of Fed services during 1976-77. Frequency of change in reserve balances is meas ured over 126 calendar days from mid-September 1976 to mid-January 1977. The minimum number of days that a member bank’s reserve balance would be unchanged during that period is 39. For the 68 banks that deposited more than five checks to their reserve balances during January 1977, reserve balances were unchanged an average of 39.9 days. A contrasting group is composed of 69 banks that cleared no checks through the Fed and used the reserve balances of their correspondents in remitting for the Fed’s cash letters and handling transactions with the Fed for coin and currency. Their reserve balances remained unchanged for an average of 75.6 days. 16Among member banks with total deposits up to $50 million, about 65 percent of the banks that borrowed frequently in 1974 had reserve balances that remained unchanged for 115 days or less, whereas only 20 percent of nonborrowers had such active reserve balances. Another factor that might account for the difference in frequency of borrowing is the difference in ratios of loan to deposits. Banks with higher ratios of loan to deposits tend to have smaller amounts of liquid as sets, which they can sell quickly to replace reserve drains. Frequent borrowers had significantly higher loan-to-deposit ratios in 1976; the difference was positive, but not statistically significant at the 5 per cent level in 1977. Thus, there is some evidence that, among the relatively small member banks with active reserve balances, those that borrow frequently from the discount window have higher ratios of loans to deposits.18 Differences in federal funds positions might also explain the varied use of the discount window by banks with active reserve balances. Reserve balances can be increased either by borrowing from the dis count window or by reducing federal funds sold to correspondents. Loans in the federal funds market 17The characteristics of member banks that borrowed from the discount window on five or more occasions in 1974, but were not frequent borrowers in the following three years, were also compared to those of member banks with rela tively active reserve balances that did not borrow fre quently in any of the years 1974-1977. The banks that borrowed frequently only during 1974 did not have greater variability in their demand deposit liabilities than the other banks. They did have higher ratios of loans to deposits than nonborrowers. However, the banks that borrowed fre quently only during 1974 had such large proportions of their assets invested in federal funds sold on average, and positive net federal funds sold so often, that they were not induced to borrow frequently from the discount window during the following years. Analysis in table 3 is limited to banks with total deposits up to $50 million because there were only two frequent borrowers in 1975-77 with total deposits be tween $50 million and $100 million (average for 1 9 7 6 ); including those two banks in the analysis, and making com parisons to nonborrowers with total deposits up to $100 million might bias some of the comparisons in table 3. 18A study of Tenth District member banks found that banks which borrowed from the discount window had higher ratios of loans to assets than nonborrowers of comparable size. See J. A. Cacy, “Determinants of Member Bank Borrow ing,” Federal Reserve Bank of Kansas City Monthly Review (February 1 9 7 1 ), pp. 11-20. 21 F E D E R A L R E S E R V E B A N K O F S T . L O U IS FEBRUARY 1980 Table 3 Comparison of Frequent Borrowers and Nonborrowers with Relatively Active Reserve Balances (1975-77)1 Measure Frequent borrowers Nonborrowers2 Difference T-Statistic Variability of demand deposits3 6.60% 4.82% 1.78% 2.0260 Percentage of average loans to average total deposits 1976 1977 61.81% 63.71 54.18% 58.52 7.63% 5.19 2.2991 1.5895 Number of weeks in which net federal funds sold was zero or negative 1976 1977 20.44 20.44 7.98 10.64 Average percentage of net federal funds sold to total deposits4 1976 1977 Number of banks 2.20% 2.32 18 5.87% 5.41 12.46 9.80 -3.67% -3.09 3.5466 2.7838 -2.2940 -2.1797 47 1Analysis is limited to banks with average total deposits in 1976 of less than $50 million dollars. 2The reserve balances of these banks were unchanged on no more than 120 days during 1976. 3Variability of demand deposits is measured in the following w ay: A bank’s demand deposit liabilities as of each business day are compared with average demand deposit liabilities in the five previous business days, and the percentage deviation of de posits of each day from the five-day moving average is calculated. The standard deviation of those percentage differences (positive and negative signs retained) is calculated for each year as a measure of the day-to-day variability of a bank’s demand deposit liabilities. For interpretation of this measure, suppose a bank has a standard deviation of 5 percent. For that bank the percentage deviations of demand deposit balances from the five-day moving average would lie within a range 5 percentage points below and above the mean on about two-thirds of the days during the year. Measuring deposit varia bility with this standard deviation weights the measure toward the more extreme percentage deviations from the five-day moving averages, which is appropriate for a measure that might be correlated with frequency of borrowing by banks to avoid reserve deficiencies. 4Net federal funds sold are counted as zero for weeks in which banks purchased more federal funds than they sold. generally have maturities of one day, and, therefore, banks that sell federal funds can increase their reserve balances quickly by reducing the amount of federal funds they sell. Federal Reserve regulations do not permit member banks to receive reserve adjustment credit through the discount window while selling federal funds. Infrequent borrowers with active reserve balances may sell federal funds so often that they are seldom permitted to borrow from the discount window. Two aspects of federal funds positions are analyzed in table 3. One is the number of weeks during each year in which a bank’s net federal funds sold was zero or negative. A bank with positive net federal funds sold during most weeks tends to have few occasions on which it would be induced to borrow. The number of weeks during which net federal funds http://fraser.stlouisfed.org/ 22 Federal Reserve Bank of St. Louis sold were zero or negative is significantly greater for frequent borrowers in both 1976 and 1977. The other measure of federal funds position is aver age federal funds sold as a percentage of average total deposits, a measure of the cushion of liquidity a bank maintains in the form of federal funds sold. This percentage is significantly lower for frequent borrowers in both 1976 and 1977. Analysis of the characteristics of banks that bor row frequently from the discount window indicates that use of the discount window for reserve adjust ment is a valuable service for frequent borrowers. Banks that borrow frequently are among those that make extensive use of Reserve Bank services. They tend to have greater short-term variability in their demand deposit liabilities and have higher F E D E R A L R E S E R V E B A N K O F S T . L O U IS ratios of loans to deposits. In addition, net federal funds sold, which are smaller on average and zero more often than for other banks, indicate that they substitute use of the discount window for holding liquid assets. Most member banks, however, prefer to adjust their reserve positions to deposit fluctuations by methods other than borrowing at the discount window. Among relatively small member banks (those with total de posits less than $100 million), most clear checks through correspondents. These banks can cope with deposit fluctuations by allowing their demand bal ances due from correspondents to fluctuate. Of the relatively small member banks that clear checks through their reserve balances, most prefer to adjust their reserve positions to deposit outflows by draw ing system that resulted from, or were compounded Only a small minority of member banks borrow from the discount window as a method of reserve adjustment. STABILITY OF THE CORRESPONDENT BANKING SYSTEM One of the reasons for creating the Federal Reserve System was concern over liquidity crises of the bank ing system that resulted from, or were compounded by, simultaneous demands for cash by respondent banks from their correspondents. This concern might still be relevant, given the continuing reliance of most banks on correspondents for liquid balances. Percentage changes from week to week in demand balances of individual banks due from correspondents are often quite large. Does it follow that demand balances of correspondent banks due to respondents are also highly variable from week to week? Are correspondent banks vulnerable to liquidity problems as a result of large fluctuations in their demand bal ances due to respondent banks? The variability of total demand balances due to respondents depends upon the size of the bank’s total balances due to respondents (see table 4). For the three largest correspondent banks, total demand balances due to respondents are about as variable as total demand deposit liabilities due to private non bank depositors. For smaller correspondent banks those balances are more variable than their other demand deposit liabilities. This contrast is greatest for the 21 smallest correspondent banks: The meas ures of variability in their demand deposit liabilities due to private nonbank depositors were about the same as for the larger correspondent banks, but bal FEBRUARY 1980 Table 4 Variability of Demand Balances Due to Commercial Banks and Private Nonbank Depositors1 Annual average of demand balances due to respondents (millions of dollars) Measure of variability of demand balances due to2 Nonbank depositors Commercial other than the banks U.S. government (10 largest correspondent banks) $ 162 5.5% 5.4% 141 7.3 5.7 141 7.8 6.7 91 6.1 3.7 60 14.1 3.0 59 10.6 3.8 53 8.0 4.0 35 9.8 7.1 27 11.7 7.5 8 12.0 3.5 (Averages for 21 smaller correspondent banks) $ 1.06 30.3% 5.1% 1These banks were selected from among all Eighth District member banks that borrowed from the discount window during 1976 and other Eighth District member banks in Illinois and Missouri. A bank is assumed to be offering correspondent services to other banks if its demand bal ances due to respondents are positive each day and change each business day. V ariability of these balances is measured by calculating average balances for each reserve settlement week in 1976, percentage changes in balances from each week to the next, and then taking the standard deviation of the per centage changes. ances due to respondents were substantially more variable. Their demand balances due to respondents were about as variable from week to week as de mand balances due from correspondents of individual banks that clear checks through correspondents. For 95 member banks that obtain most of their services through correspondents, the average measure of weekly variability in their demand balances due from correspondents is 29.5, which is approximately equal to the average measure of variability in demand balances due to respondents of 30.3 for the 21 smallest correspondent banks.19 19VariabiHty of demand balances of correspondents due to re spondents is measured as the standard deviation of percent age changes in those balances from week to week. Variability of demand balances due from correspondents is measured the same way. The 95 member banks which use correspond ent services are described in footnote 12. 23 F E D E R A L R E S E R V E B A N K O F S T . L O U IS These comparisons indicate that the relatively large correspondent banks have an advantage over smaller correspondents in coping with the variability in de mand balances due to individual respondent banks. For correspondents that serve the largest number of respondent banks, fluctuations in balances due to indi vidual respondent banks tend to cancel each other out, causing their total demand balances due to other banks to be no more variable from week to week than the sum of other demand deposit liabilities. Cor respondents that serve fewer respondents do not benefit as much from such cancellation. Thus, the largest correspondent banks are able to offer respond ents the service of reserve adjustment mechanisms with no more vulnerability to occasional liquidity problems resulting from fluctuations in their balances due to respondents than that from fluctuations in bal ances due to nonbank depositors. SUMMARY AND CONCLUSIONS It recently has been suggested that one adverse effect of declining Federal Reserve membership is that the Federal Reserve may become more reluctant to pursue restrictive monetary policy because of its uneven impact on the banking system. Nonmember banks, without access to the discount window, might not have the means to cope with liquidity pressures resulting from restrictive monetary policy. Therefore, extending access to the discount window to more banks, by inducing or requiring more banks to be members, would facilitate the implementation of monetary policy. The validity of this argument is investigated by examining the frequency with which member banks borrow at the discount window. If access to the dis count window helps banks adjust to occasional liquid ity pressures, most member banks would borrow 24 FEBRUARY 1980 frequently from the discount window as a means of coping with such pressures. Only a small minority of member banks, however, actually borrows frequently from the Federal Reserve. Therefore, while there are valid reasons for suggesting that reductions in mem bership may interfere with monetary control, analysis of reserve management by banks does not indicate that direct access to the discount window for non member banks is important for the conduct of monetary policy. Many member banks adjust to declines in their cash balances by letting their demand balances due from correspondents decline temporarily. This reaction is automatic for banks that clear checks through correspondent balances. For banks that clear checks through accounts at their Reserve Bank, net withdrawals of deposit liabilities cause reductions in their reserve balances, and may occasionally cause reserve deficiencies unless offset. Banks that borrow frequently at the discount window are among these banks that clear checks through their reserve ac counts. Most member banks that clear checks through reserve accounts, however, do not borrow frequently from the discount window, but instead rely upon federal funds sold to correspondents as a cushion of liquidity, reducing federal funds sold when expe riencing reserve outflows. The relatively large correspondent banks benefit from offsetting fluctuations in the deposits due to in dividual respondent banks. Thus, large correspondent banks can offer respondents means of adjusting their cash positions to deposit fluctuations without incur ring any more variability in their own liabilities than results from fluctuations in demand balances due to nonbank customers. This indicates that correspondent banks offer member and nonmember banks adequate means of adjusting their cash position to fluctuations in deposit liabilities. The New Monetary Aggregates R. W. HAFER A Jnumber of major financial innovations over the last decade have changed the composition of assets used by the public to make payments. Examples in clude the introduction of negotiable orders of with drawal (NOW accounts), the implementation of auto matic transfer systems ( ATS accounts) whereby funds from savings accounts can be automatically trans ferred to checkable deposits, the growing use of money market mutual funds as substitutes for conven tional savings accounts, and the dramatic growth in repurchase agreements (RPs).1 Because of these and other developments, it increasingly was argued that the existing monetary aggregates did not measure the true financial position of the public and, therefore, were inadequate tools of monetary policy. In response to these developments, the Federal Re serve Board recently announced redefinitions of the monetary aggregates.2 This article describes the new aggregates, compares them to the old measures, and discusses some technical issues involved in their measurement. 1For a discussion of NOW and ATS accounts and their effect on the old monetary aggregates, see Steven M. Roberts, “De veloping Money Substitutes: Current Trends and Their Implications for Redefining the Monetary Aggregates,” Im prov ing the Monetary Aggregates: Staff Papers, Board of Gover nors of the Federal Reserve System, Washington, D.C., 1978, pp. 147-70. Hereafter, this publication will be referred to as Staff Papers. See also John A. Tatom and Richard W . Lang, “Automatic Transfers and the Money Supply Process,” this R eview (February 1 9 7 9 ), pp. 2-10. An introduction to re purchase agreements is found in Norman N. Bowsher, “Re purchase Agreements,” this R eview (September 1 9 7 9 ), pp. 17-22. 2“Announcement,” Board of Governors of the Federal Reserve System, Washington, D.C., February 7, 1980. Definitions Tables 1 and 2 compare the old and new monetary aggregates. As shown in table 1, the new basic trans actions measure — MIA — is essentially the same as the old M l measure, except for the deletion of de mand deposits due to foreign commercial banks and Table 1 Comparison of Old and New Transactions-Type Monetary Aggregates Component Old M1 New M1A New M1B Currency in circulation X X X At commercial banks: Demand deposits inclusive of deposits due to foreign commercial banks and official institutions Demand deposits exclusive of deposits due to foreign commercial banks and official institutions X X X NOW accounts X ATS accounts X At thrift institutions: Demand deposits NOW accounts X ATS accounts X Credit union share draft balances X X 25 F E D E R A L R E S E R V E B A N K O F S T . L O U IS official institutions. This change, based on a recom mendation of the Advisory Committee on Monetary Statistics, was made because such balances consist primarily of compensating balances held by foreign commercial banlcs at U.S. commercial banks for serv ices performed.3 Since the total of such balances is a small percentage of the old Ml, this deletion does not produce large discrepancies between the old Ml and new MIA measures.4 Until the early 1970s, a clear distinction between interest-bearing deposits and non-interest-earning de posits held for transactions purposes existed. Since then, however, a series of financial innovations and regulatory changes have blurred this distinction.5 The significant changes have taken the form of NOW ac counts, which were established in several New Eng land states in the early 1970s, ATS accounts, and the rapid growth of credit union share drafts as an alter native payments mechanism. The new M1B aggregate combines those financial items that have the dual characteristic of being held both for check-writing purposes and as savings accounts. Nearly all of these items are interest-bearing checkable deposits.8 The new MIA and M1B aggregates focus on those mone tary components that are employed primarily as a means of payment. As shown in table 2, the old M2 aggregate was de fined as the sum of currency, demand deposits, sav ings deposits, time deposits, and time certificates of deposit (CDs) other than those issued in denomina tions of $100,000 or more by large weekly reporting banks. The new M2 measure is much broader in scope. It is calculated by adding savings deposits and small time deposits (those issued in denominations of less than $100,000) at all depositary institutions, overnight RPs issued by commercial banks, overnight Eurodollars (issued by Caribbean branches of mem 3See Improving the Monetary Aggregates: Report o f the Ad visory Com m ittee on Monetary Statistics, Board of Governors of the Federal Reserve System, Washington, D.C., 1976, pp. 15-20. Hereafter, this publication will be referred to as Report. 4Fo r example, during 1978 the amount of demand deposits due to foreign-related banking offices averaged less than 2 percent of total demand deposits. 5For an excellent description of this, see Roberts, “Developing Money Substitutes.” 6This distinction is necessary because not all of the new check able deposits are interest-bearing at present. For example, some depositary institutions currently offer non-interest-bear ing NOW accounts (N IN O W s), and demand deposits at mu tual savings banks do not currently pay interest. Also, some non-interest-eaming demand deposits that are held at thrift institutions and cannot be separated from interest-bearing checkable deposits are included in M1B. At present, the amount of such deposits is small. 26 FEBRUARY 1980 ber banks) held by U.S. nonbank residents, money market mutual fund shares, and a consolidation com ponent to the new M1B measure. This definition in cludes such a broad array of monetary components and institutions (e.g., commercial banks, U.S. agencies and branches of foreign banks, Edge Act corporations, foreign investment companies, mutual savings banks, savings and loan associations, and credit unions), that it is more directly comparable to old M3 than to the previous M2 definition (see table 2). There is, however, one major difference between the new M2 measure and the old M3 definition: The new M2 includes overnight RPs and Eurodollars, and money market mutual funds shares which were not included in the old M3 measure. The introduction of these items stems from their increasing substitutability for other non-transactionstype financial holdings already included in the broader monetary measures. For instance, money market mu tual funds shares are viewed as substitutes for other non-transactions-tvpe financial assets, despite the fact that owners of these shares are offered check-writing privileges. The fairly large minimum denomination re quirement (usually $500 or more) for checks written on these accounts and the fact that these balances typically exhibit relatively slow turnover rates suggest that these accounts are used primarily as savings rather than transactions accounts.7 The transactions and investment characteristics of overnight RPs have been subjected to considerable investigation in recent years. For example, studies by Garcia and Pak, Wenninger and Sivesind, and Tinsley, Garrett, and Friar have viewed these RPs as close sub stitutes for existing demand deposits. Consequently, they have explained a large part of the decline in the public’s demand for transactions balances which occurred in the mid-1970s by including RPs in the definition of a transactions-type money (i.e., old M l). In contrast, others have regarded RPs as short-term, highly liquid investment items that are significantly different from demand deposits.8 "Thomas D. Simpson, “The Redefined Monetary Aggregates,” F ederal Reserve Bulletin (February 1 9 8 0 ), p. 100. 8See Gillian Garcia and Simon Pak, “Some Clues in the Case of the Missing Money,” American Econom ic Review ; Papers and Proceedings (M ay 1 9 7 9 ), pp. 330-34; John Wenninger and Charles Sivesind, “Defining Money for a Changing F i nancial System,” Federal Reserve Bank of New York Quar terly R eview ( Spring 1979) pp. 1-8; Peter A. Tinsley, Bonnie Garrett, and Monica Friar, “The Measurement of Money Demand,” Staff Study # 1 3 3 (Board of Governors of the Fed eral Reserve System, 1 9 7 8 ); and Thomas D. Simpson, “The Market for Federal Funds and Repurchase Agreements,” Staff Study # 1 6 6 (Board of Governors of the Federal Reserve System, 19 7 9 ). FEBRUARY F E D E R A L R E S E R V E B A N K O F S T . L O U IS 1980 Table 2 Comparison of Old and New Non-Transactions-Type Monetary Aggregates Component Currency Old M2 New M2 Old M3 New M3 Old M4 Old M5 L X X X X X X X X X At commercial banks: Demand deposits inclusive of deposits due to foreign commercial banks and official institutions X X Demand deposits exclusive of deposits due to foreign commercial banks and official institutions X X X NOW accounts1 X X X ATS accounts1 X X X X X X Overnight RPs Savings deposits X X X X X X X Small time deposits (< $100,000) X X X X X X X Large time deposits Other than large negotiable CDs Including large negotiable CDs X X X X X X X X X X Term RPs X X At thrift institutions: Demand deposits X X X X NOW accounts1 X X ATS accounts1 X X Credit union share draft balances X X X X X Savings deposits (Mutual savings banks and savings and loan associations) X X X X X Small time deposits (< $100,000) X X X X X X X X X Large time deposits (> $100,000) Term RPs (Commercial banks and savings and loan associations) X X X Other: Overnight Eurodollar deposits of U.S. nonbank residents2 X X X Money market mutual funds shares X X X Term Eurodollars held by U.S. nonbank residents X Bankers acceptances X Commercial paper X U.S. savings bonds X Liquid Treasury securities M2 consolidation component3 X X X X 1These accounts were included previously in the savings deposit component of the definitions. 2Overnight Eurodollars issued by Caribbean branches of member banks. 3See text, p. 30, for a discussion of this component. 27 FEBRUARY F E D E R A L R E S E R V E B A N K O F S T . L O U IS There presently is no consensus as to whether over night RPs and Eurodollars, and money market mutual funds shares primarily constitute transactions- or in vestment-type assets. These items are included in the new M2 definition. Data on each of these series will be published separately, however, so a direct com parison of these components with the new transactions aggregates MIA and M1B will be possible. The new M3 series is defined as new M2 plus large time deposits (those issued in denominations of $100,000 or more) at all depositary institutions and term RPs issued by commercial banks and savings and loan associations. The new M3 aggregate is simi lar to the old M5 definition primarily because of the large-denomination time deposits component. The combination of the large-denomination time deposits and term RPs in this aggregate is based on the belief that these items are relatively close substitutes in many financial portfolios.9 The broadest of the new monetary aggregate defi nitions is the “L” series. This aggregate, which meas ures total liquid assets, adds to the new M3 series such financial items as other Eurodollar holdings by non bank U.S. residents, bankers acceptances, commercial paper, U.S. savings bonds, and liquid Treasury obliga tions.10 This measure closely approximates the credit expansion generated through the commercial banking sector and other financial channels. COMPARISON OF GROWTH RATES A comparison of the growth rates of the old and new monetary aggregates provides a useful way to assess the differences resulting from the redefinitions. As shown in table 3, there is relatively little quantitative difference between the annual growth rates of the Ml aggregate and the new MIA and M1B measures over the 1970-1979 period. For example, the average differ ence in annual growth rates between Ml and MIA over this period is only 0.18 percentage points, the largest divergence occurring in 1973 when Ml grew 0.5 percent faster than MIA. 1980 Table 3 Rates of Growth for Transactions-Type Monetary Aggregates Annual Rate of Grow th 1 Year Old M1 New M1A New M1B 1970 4.8% 4.8% 4.8% 1971 6.6 6.6 6.6 1972 8.4 8.5 8.5 1973 6.2 5.7 5.8 1974 5.1 4.7 4.7 1975 4.6 4.7 4.9 1976 5.8 5.5 6.0 1977 7.9 7.7 8.1 1978 7.2 7.4 8.2 1979 5.5 5.5 8.0 Q uarterly Rate of Grow th 2 Quarter Old M1 New M1A New M1B 1/1975 2.0% 2.6% 2.9% 11/1975 5.8 5.9 5.9 111/1975 7.2 7.0 7.3 IV/1975 3.0 2.9 3.2 1/1976 4.6 5.4 5.7 11/1976 6.4 5.8 6.3 111/1976 4.1 3.4 3.9 IV/1976 7.4 7.0 7.6 1/1977 7.4 8.8 9.3 11/1977 7.4 6.7 6.9 111/1977 8.6 6.0 6.5 IV/1977 7.4 8.4 8.7 1/1978 6.6 7.6 7.9 11/1978 9.2 8.7 9.1 111/1978 7.9 7.1 7.3 IV/1978 4.3 5.6 7.4 -1.3 0.2 4.8 1/1979 11/1979 8.1 7.8 10.7 111/1979 9.7 8.8 10.1 IV/1979 5.0 4.7 5.3 8Simpson, “The Redefined Monetary Aggregates,” p. 102. 10The Eurodollar holdings included in this measure incorpo rate those that are not captured in overnight Eurodollars issued by Caribbean branches of member banks. Liquid Treasury obligations consist of those issues with 18 months or less remaining to maturity. See Simpson, “The Redefined Monetary Aggregates,” p. 98. It should also be noted that the new M2, M3, and L ag gregates exclude the amounts held by depositary institutions, money market mutual funds, the federal government, the Federal Reserve, and foreign commercial banks and official institutions. See Simpson, “The Redefined Monetary Aggre gates,” pp. 98, 108. Digitized for 28 FRASER 1Fourth-quarter-to-fourth-quarter growth rates. -Annualized growth rates based on seasonally adjusted data. SOURCE: “The Redefined Monetary Aggregates,” table A l. Although the average difference between the M l and M1B measures is somewhat larger (0.51 percent), M1B has demonstrated a faster rate of growth over recent years relative to Ml. This faster growth—1.0 percent in 1978 and 2.5 percent in 1979 — results from F E D E R A L R E S E R V E B A N K O F S T . L O U IS FEBRUARY Table 4 Rates of Growth for Old M2, Old M3, and New M2 Monetary Aggregates1 Annual Rate of Growth Year 1970 Old M2 Old M3 7.2% 7.2% New M2 5.8% 1971 11.3 13.5 13.5 1972 11.2 13.3 12.9 1973 8.8 9.0 7.3 1974 7.7 7.1 6.0 1975 8.4 11.1 12.3 1976 10.9 12.7 13.7 1977 9.8 11.7 11.5 1978 8.7 9.5 8.4 1979 8.3 8.1 8.8 Quarterly Rate of Growth Quarter Old M2 Old M3 New M2 1/1975 6.4% 11/1975 9.5 12.4 14.9 111/1975 10.0 12.8 14.6 IV/1975 6.8 9.4 9.9 1/1976 10.5 12.0 13.0 11/1976 10.0 11.9 12.7 8.2% 7.8% 111/1976 8.9 11.0 11.3 IV/1976 12.6 13.8 15.2 1/1977 10.9 12.4 13.7 11/1977 9.0 10.5 11.2 111/1977 10.1 11.8 9.6 IV/1977 7.9 10.1 9.7 7.5 1/1978 7.0 8.1 11/1978 8.4 8.4 7.5 111/1978 9.8 10.3 8.2 IV/1978 8.5 9.8 9.5 1/1979 2.8 5.3 6.3 11/1979 8.8 7.9 10.2 111/1979 11.9 10.5 10.3 IV/1979 8.9 7.8 7.2 JSee footnotes accompanying table 3. SOURCE: “The Redefined Monetary Aggregates,” table A2. the increased use of NOW and ATS accounts as de mand deposit and other balances are shifted into these interest-earning checkable deposits. Growth rates of the old and new M l measures ex hibit greater divergence on a quarter-to-quarter basis. 1980 Over the last five years, the average difference be tween the M l and MIA quarterly growth rates was 0.80 percent; the average difference between Ml and M1B growth was 1.16 percent. The impact of NOW and ATS accounts again is demonstrated — during 1979, M1B grew 2.5 percent faster than either M l or MIA. As these figures suggest, the extension of NOW accounts nationwide may temporarily produce wider divergencies between the MIA (and old M l) and M1B growth rates. Annual and quarterly growth rates for the old M2, old M3, and new M2 measures are presented in table 4. As these figures show, growth rates of new M2 tend to be closer to those of the old M3 definition than to old M2. For instance, the average annual growth rate of old M2 was 9.2 percent over the last decade while the averages for old M3 and new M2 were 10.3 percent and 10.0 percent, respectively. An examination of the quarterly data reveals a similar relationship: From 1/1975 to IV/1979, old M2 grew at an 8.9 percent average rate while the average growth rates for old M3 and new M2 were 10.2 and 10.5 percent, respectively. In addition, the proportion of new M2 that consists of money market certificates and money market mutual funds has increased sharply since 1978.11 Table 5 presents the annual and quarterly growth rates for the old M4 and M5 aggregates together with the new M3 and L definitions. As noted earlier, the new M3 aggregate is relatively closer in construction to the old M5 measure than to old M4. The difference between the average annual rate of growth of old M4 and new M3 is 1.2 percentage points; that between old M5 and new M3 is only 0.5 of a percentage point. On a quarter-to-quarter basis, movements in new M3 and old M5 are even more similar. For example, over the period I/1975-IV/1979, the average quarterly rate of growth of old M4 was 8.1 percent while that of old M5 and new M3 was 9.6 percent and 10.4 per cent, respectively. The growth rates of L — total liquid assets — have been closer to new M3 than to the other monetary ag gregates. While the average quarterly growth rates of L and new M3 have been roughly similar over the past five years (11.1 percent and 10.4 percent, respec tively), there has been a growing divergence between these measures in more recent years. This is explained by the rapid growth of liquid assets issued by non n Simpson, “The Redefined Monetary Aggregates,” p. 105. 29 F E D E R A L R E S E R V E B A N K O F S T . L O U IS depositary institutions which form the distinction be tween new M3 and L.12 MEASURING THE NEW AGGREGATES: TECHNICAL CONSIDERATIONS Several technical problems arise in the measure ment of the new monetary aggregates: Certain de posits held by depositary institutions must be consoli dated to avoid double counting, the series used in calculating the new aggregates must be seasonally ad justed, and the data needed to construct the new aggregates must be gathered. In calculating the old Ml aggregate, the problem of double counting deposits was resolved by netting out cash items in the process of collection, interbank deposits, and Federal Reserve float from total com mercial bank demand deposits.13 A similar proce dure is followed in measuring the new MIA. At the M1B and M2 levels, however, it is assumed that thrift institutions hold demand deposits at com mercial banks to service their checkable deposits and ordinary savings deposits. Thus, in calculating M1B, the estimated proportion of demand deposits owned by thrift institutions used to service their checkable deposits will be removed;14 for new M2, total demand deposits owned by thrift institutions are currently netted out. At the new M2 and M3 levels, further consolidation measures are employed. For instance, in the calcula tion of new M2, savings and time deposits owned by all depositary institutions are netted out, and money market mutual funds’ holdings of RPs are deducted from the public’s holdings of overnight RPs. In addi tion, CDs held by these funds are also netted out of large time deposits in calculating new M3. Both of these latter items are netted out in the derivation of the total liquid assets aggregate (L ). 12Fo r example, the percentage increase in dollar amounts be tween January 1978 and December 1979 for these items, using seasonally adjusted data, are: Bankers accep tan ces............................ + 1 2 0 % Commercial paper ............................... + 48 Short-term Treasury securities ...... + 39 U.S. savings bonds .............................. + 4 and, based on seasonally unadjusted data, + 1 3 6 % for term Eurodollars. 13Fo r a discussion of this problem, see Report, pp. 12-14; Darwin Beck, “Sources of Data and Methods of Construc tion of the Monetary Aggregates,” Staff Papers, pp. 117-33; and Simpson, “The Redefined Monetary Aggregates,” pp. 107-10. 14At the present time, the amount of such holdings is negli gible and, therefore, is not omitted from M1B. 30 FEBRUARY 1980 Table 5 Rates of Growth for Old M4, Old M5, New M3, and L Monetary Aggregates1 Annual Rate of Growth Year Old M4 1970 10.2% 1971 12.8 14.3 14.8 10.4 1972 12.3 13.9 14.0 12.9 1973 12.0 11.0 11.7 12.3 1974 10.7 9.0 8.7 9.6 1975 6.6 9.7 9.4 9.8 1976 7.1 10.2 11.4 11.0 12.6 Old M5 New M3 9.2% 8.9% t 6.5% 1977 10.1 11.7 12.6 1978 10.6 10.6 11.3 12.3 1979 7.5 7.6 9.5 N.A. Q uarterly Rate of Growth Quarter Old M4 Old M5 New M3 L 1/1975 7.6% 8.9% 7.2% 7.1% 11/1975 5.5 9.5 9.4 9.5 111/1975 6.2 10.1 10.7 10.5 IV/1975 6.2 8.8 9.1 10.7 1/1976 6.0 9.0 9.9 10.1 11/1976 6.0 9.4 11.3 11.1 111/1976 6.3 9.2 10.3 10.0 IV/1976 9.5 11.8 12.1 10.8 10.1 11.8 12.4 11.5 1/1977 11/1977 8.3 10.0 11.4 11.8 111/1977 10.0 11.7 11.7 12.2 IV/1977 10.4 11.5 12.5 12.8 1/1978 10.2 10.0 10.5 11.2 11/1978 10.6 9.8 11.1 12.4 111/1978 9.9 10.4 10.3 11.3 IV/1978 10.1 10.7 11.5 12.2 1/1979 5.4 6.8 7.9 10.4 11/1979 3.7 4.9 8.8 13.1 111/1979 9.2 8.9 10.3 11.7 IV/1979 11.0 9.1 9.8 N.A. !See footnotes accompanying table 3. SOURCE: “The Redefined Monetary Aggregates,” table A3. Derivation of seasonally adjusted aggregates fol lows past procedures wherein the individual compo nents of the series are seasonally adjusted first, then aggregated to the desired level.15 At this time, how15See Simpson, “The Redefined Monetary Aggregates,” pp. 11011. Fo r a general treatment of the seasonal adjustment F E D E R A L R E S E R V E B A N K O F S T . L O U IS FEBRUARY 1980 Table 6 New and Proposed Data Sources Institution Member banks Nonmember banks Mutual savings banks Savings and loan associations Credit unions1 Component Collected Coverage / Freq uency Term RPs 125 large member banks/weekly Overnight RPs 125 large member banks/weekly NOW and ATS accounts all member banks/weekly Overnight Eurodollars at Caribbean branches approximately all/weekly Demand deposits sample/weekly NOW and ATS accounts sample/weekly Savings and small-denomination time deposits sample/weekly Large-denomination time deposits sample/weekly NOW accounts and demand deposits sample/weekly (Wednesday) Savings and small-denomination time deposits sample/weekly (Wednesday) Large-denomination time deposits sample/weekly (Wednesday) NOW accounts sample/thrice-monthly Savings and small-denomination time deposits sample/thrice-monthly Large-denomination time deposits sample/thrice-monthly Share drafts sample/weekly (Wednesday) Savings and small-denomination time deposits sample/weekly (Wednesday) 1The weekly sample (scheduled to begin in March 1980) will consist of 70 of the nation’s largest credit unions, plus a sampie of smaller credit unions to be collected once a month. SOURCE: “The Redefined Monetary Aggregates.” ever, several of the components used to calculate some of the new aggregates are not seasonally ad justed because of data insufficiencies or technical diffi culties. The individual series that have not been seasonally adjusted include NOW accounts, ATS ac counts, credit union share drafts, demand deposits at thrift institutions, overnight RPs and Eurodollars, money market mutual fund shares, term RPs at com mercial banks and savings and loan associations, and term Eurodollars held by U.S. nonbank residents. A much wider diversity of financial institutions now participates in the data reporting and collection process (see table 6). Financial institutions that have not been active participants in the previous deriva tion of the monetary aggregates will play an impor tant role. For example, the Federal Home Loan Bank Board now collects data on NOW accounts held at savings and loan associations; beginning in the spring of 1980, a sample of large credit unions will provide data on credit union share drafts and related items; problem, see Report, pp. 37-40 and, for a technical discus sion, David A. Pierce, Neva Van Peski, and Edward R. Fry, “Seasonal Adjustment of the Monetary Aggregates,” Staff Papers, pp. 71-90. the Investment Company Institute provides a weekly survey of money market mutual fund shares; and a daily survey of 125 large member banks forms the basis for the RP series. As this incomplete listing sug gests, the comprehensiveness of the new monetary ag gregates is greater than the previous measures. CONCLUSION The Federal Reserve Board recently has redefined the monetary aggregates to provide better measures of financial assets held by the public. The new basic transaction measure, called MIA, is equal to the former M l minus demand deposits held at commer cial banks due to foreign commercial banks and offi cial institutions. Large discrepancies between the growth rates of Ml and MIA are not anticipated. In addition, a broader transactions measure — M1B — has been introduced. This aggregate com bines those deposits that are held, for the most part, both for check-writing purposes and as savings ac counts. M1B, therefore, equals MIA plus NOW ac counts, savings accounts subject to automatic trans fer (ATS accounts), credit union share drafts, and 31 demand deposits at mutual savings banks (non-inter est-bearing). The growth of NOW and ATS accounts has contributed to faster growth of M1B relative to Ml or MIA. Consequently, if NOW accounts are legalized nationwide, more rapid growth in M1B rela tive to MIA is expected. The Board also has redefined M2 and M3 and has introduced a new aggregate, L, which is intended to measure total liquid assets held by the public. These redefinitions represent a consolidation of the 32 former M2, M3, M4, and M5 measures, For instance, the new M2 is similar in definition to the old M3; the new M3 is similar to the old M5. The new L aggre gate, unlike any previous measure, includes such items as term Eurodollars held by U.S. nonbank resi dents, bankers acceptances, commercial paper, U.S. savings bonds, and other liquid Treasury obligations. This broad measure of liquid financial assets is be lieved to provide a useful measure of credit in the economy which arises either through the banking sec tor or through other financial channels.