A Monetarist Model for Economic Stabilization, Schemes and Mind Maps of Business

Download A Monetarist Model for Economic Stabilization and more Schemes and Mind Maps Business in PDF only on Docsity! A Monetarist Model for Economie Stabilization* by LEONALL C. ANDERSEN and KEITH M. CARLSON rji I. HE MONETARIST VIEW that changes in the money stock area primary determinant of changes in total spending, and should thereby be given major emphasis in economic stabilization programs, has been of growing interest in recent years. From the mid-1930’s to the mid-1960’s. monetary policy re- ceived little emphasis in economic stabilization policy. Presumed failure of monetary policy during the early years of the Great Depression, along with the development and general acceptance of Keynesian economics, resulted in a main emphasis on fiscal ac- tions — Federal Government spending and taxing programs — in economic stabilization plans. Monetary policy, insofar as it received any attention, was gen- erally expressed in terms of market rates of interest. Growing recognition of the importance of money and other monetary aggregates in the determination of spending, output, and prices has been fostered by the apparent failure of stabilization policy to curb the inflation of the last half of the l960’s. Sharply rising market interest rates w’ere interpreted to indi- cate significant monetary restraint, while the Reve- nue and Expenditure Control Act of 1968 was con- sidered a major move toward fiscal restraint. Despite these policy developments, total spending continued to rise rapidly until late 1969, and the rate of inflation accelerated. Those holding to the mone- tarist view were not surprised by this lack of success aOffering helpful suggestions throughout the study were Denis Karnosky of this bank, William P. Yohe of Duke Uni- versity and visiting Scholar at this bank, 1969-70, and David Fand of Wayne State University. Susan Smith pro- vided programming assistance and Christopher Babb and H. Albert Margolis advised on statistical problems. The authors thank the following for their comments on earlier drafts, without implying their endorsement of either the methods of analysis or the conclusions F. Gerard Adams, Philip Cagan, E. Gerald Corrigan, Richard Davis, Ray Fair, Edgar Fiedler, Milton Friedman and members of the Money and Banking Workshop at the University of Chicago, Edward Gramlieh, Harry C. Johnson, John Kalchbrenner, Edward Kane, Michael Keran, Allan Meltzer, Franco Modigliani, George Morrison, David Ott, Joel Popkin, Thomas Saving, Roger Spencer, Henry Wallich, Clark Warburton, Manfred Willms, and Arnold Zellner. in curbing excessive growth in total spending, largely because the money stock grew at a historically rapid rate during the four years ending in late 1968. Eco- nomic developments from 1965 through 1969 were in general agreement with the expectations of the monetarist view. This article develops a model designed to analyze economic stabilization issues within a framework which focuses on the influence of monetary expan- sion on total spending. Most of the major econometric models have not assigned an important role to the money stock or to any other monetary aggregate.1 Furthennore, most econometric models contain a large number of behavioral hypotheses to be empirically estimated and integrated with each other, because they are designed to aid in understanding the deter- mination of many economic magnitudes. By compari- son, the model presented in this article is quite small. It is designed to provide information on the most likely course of movement of certain strategic eco- nomic variables in response to monetary and fiscal actions. Frank de Leeuw and Edward M. Cramlieh, “The Federal Reserve-MI’I’ Econometric Model,” Federal Reserve Bul- letin (January 1968), pp. 11-40, and “The Channels of Monetary Policy: A Further Report on the Federal Reserve MIT Econometric Model,” Federal Re~~~erveBulletin (June 1969), pp. 472-91; James S. Duesenberry, Gary Fromm, Lawrence R. Klein, and Edwin Kuh (ed), The Brookings Quarterly Econometric Mode? of the United States (Chicago: Rand McNally, 1965), and The Brookings Model: Some Further Results (Chicago: Rand McNally, 1969); Michael K. Evans and Lawrence R. Klein, The Wharton Econometric Forecasting Model, 2nd Enlarged Edition (Philadelphia: University of Pennsylvania, 1968); Maurice Liebenberg, Albert A. Hirsch, and Joel Popkin, “A Quarterly Econo- metric Model of the United States: A Progress Report,” Survey of Cur,’ent Business (May 1966), pp. 423-56; Daniel M. Suits, “The Economic Outlook for 1969,” in The Economic Outlook for 1969, Papers presented to the Six- teenth Annual Conference on the Economic Outlook at The University of Michigan (Ann Arbor: University of Michigan, 1969), pp. 1-26. For a discussion of the role of money in these models, see David I. Fand, “The Monetary Theory of Nine Recent Quarterly Econometric Models of the United States,” forthcoming in the Journal of Money, Credit, and Banking. Page 7 ~AL RESERVE SAI?or ST LOUIS - APRIL 1970 The model presented here is the authors’ own ver- sion of how monetary and fiscal actions influence the economy. Other economists (including those of a inonetarist persuasion) may prefer to develop certain aspects of the model in a different way. Two such modifications are presented in Appendix C. The model is considered open to revision, hut is presented at this time with a view to stimulating others to join in quantifying relationships that are generally asso- ciated with the monetarist view. This article is divided into five major sections. A general monetarist view of the response of spending, output, and prices to monetary and fiscal actions is summarized first. Next, the specific features of the model are discussed within a formal framework of analysis. Statistical estimates of the model’s param- eters are presented in the third section. The fourth section tests the performance of the model with sev- eral dynamic simulation experiments.. Finally, by simulating the response of the economy to alterna- tive rates of monetary expansion, an illustration is provided of how the model can be used for current stabilization analysis. General Monetarist View The general monetarist viesv is that the rate of monetary expansion is the main determinant of total spending, commonly measured by gross national product (GNP ) .~ Changes in total spending, in turn, influence movements in output, employment, and the general price level. A basic premise of this analysis is that the economy is basically stable and not neces- sarily subject to recurring periods of severe recession and inflation. Major business cycle movements that have occurred in tile past are attributed primarily to large swings in the rate of growth in the money stock. This view regarding aggregate economic relation- ships differs from prevailing views which consider aggressive policy actions necessary to promote sta- bility. Monetarists generally hold that fiscal actions, in the absence of accommodative monetary actions, exert little net influence on total spending and there- fore have little influence on output and the price level. Governsnent spending unaccompanied by ac- commodative monetary expansion, that is, financed by taxes or borrowing from the public, results in a crowding-out of private expenditures with little, if any, 2 Ceneral references on the monetarist view are Karl Bmnner, “The Role of Money and Monetary Policy,” this Review (July 1968), pp. 9-24; David I. Fand, “Some Issues in Monetary Economics,” this Review (January 1970), pp. 10-27, and “A Monetarist Model of the Monetary Process,” forthcoming in the Journal of Finance, net increase in total spending. A change in the money stock, on the other hand, exerts a strong independent influence on total spending. Monetarists conclude that actions of monetary authorities which result in changes in the money stock should be the main tool of economic stabilization. Since the economy is con- sidered to be basically stable, and since most major business cycle movements in the past have resulted from inappropriate movements in the money stock, control of the rate of monetary expansion is the means by which economic instability can be minimized. The theoretical heritage of the monetarist position is tile quantity theory of money.8 This theory dates back to the classical economists (particularly David Ricardo) in the early 1800’s. The quantity theory in its simplest form is characterized as a relationship between the stock of money and the price level. Classical economists concentrated on the long run aspects of the quantity theory in which changes in the money stock result in changes only in nominal magnitudes, like the price level, but have no influ- ence on real magnitudes like output and employment. The quantity theory of money in its modem form recognizes the important influence that changes in the money stock can have on real magnitudes in the short run, while influencing only the price level in the long run. The modern quantity theory postu- lates that in the short run a change in the rate of growth in money is followed with a moderate lag by changes in total spending and output, while changes in the price level follow with a somewhat longer lag.~ These changes in total spending, output, and prices are in the same direction as the change in the rate of monetary expansion. The modern quantity theory still accepts the long- run postulates of its older version. A change in the rate of monetary expansion influences only nominal niagnitudes in the long run, namely, total spending (GNP) and the price level. Real magnitudes, notably 3 The classic work on the quantity theory is Irving Fisher, The Purchasing Power of Money (New York: Macmillan, 1911). For an extensive review of the quantity theory literature, see Arthur W. Marget, The Theory of Prices: A Re-examination of the Central Problems of Monetary Theory (New York: Prentice-Hall, 1938), volume II, pp. 3-133. 4 Many of the ideas prevalent in current monetarist doctrine can be found in the writings of Clark Warburton in the 194O’s and early 1950’s. Many of his important articles have been reprinted in his Depression, inflation, and Monetary Policy, Selected Papers, 1945-1953 (Baltimore: The Johns Hopkins Press, 1966). See also Milton Friedman, (ed), Studies in the Quantity Theory of Money (Chicago: Uni- versity of Chicago Press, 1958), and Lloyd W. Mints, Monetary Policy in a Competitive Society (New York: MeCraw-Hill, 1951). Page 8 FEDERAL RESERVE BANK OF ST. LOUIS APRIL, 1970 the response of output and prices to monetary and fiscal actions, not to test a hypothesized structure. The focus is on the response in the short run — periods of two or three years — but the long-run properties of the model also are examined. Estimation of the Model The general form of the model indicates those variables that are included in each equation. Estima- tion requires selection of the algebraic form of the equations and the techniques to be used in estimnation. Each of the equations of the model is estimated by ordinary least squares. Lag structures, with one exception, are estimated by the Almnon lag technique. The reported relationships reflect considerable ex- perimentation with the number of lags and the de- gree of the polynomial.” The sample period starts with 1933 for the spending equation and with 1955 for all the others. The data are quarterly and, with the exception of interest rates, are seasonally adjusted. Criteria used in the selection of the equations were minimizing the standard error of estimate and eliminating serial correlation in the estimated resid- uals. In addition, the signs and statistical significance of the estimate’d coefficients received consideration, along with the pattern of the lag distribution. Since these criteria frequently could not be satisfied simul- taneously, an elemnent of subjectivity was present in- selecting the “best” equation. The change in total spending is specified as a function of current and past changes in the money stock (demand deposits and currency held by the nonbank public) and in high-employment Federal expenditures (expenditures on goods and services plus transfer payments adjusted to remove the influence of variations in economic activity on unemployment benefit payments)~The choice of the particular equa- tion (Table I) is based on previous work by Andersen and Jordan.’2 Implicit in this choice is the assump- tion that the change in the money stock is an exog- enous variable. ~kmore complete model would specify a mechanism whereby the money stock is determined by actions of the monetary authorities, the public, and the banking system. ‘‘For discussion of the use and interpretation of the Almon lag technique, see Keran, p. 10. ‘ tm Leonall C. Andersen and Jerry L. Jordan, “Monetary and Fiscal Actions: A Test of Their Relative Importance in Economic Stabilization,” this Review (November 1968), pp. 11-24. See also Keran, pp. 5-24. Table I TOTAL SPENDING EQUATION $qmp)e Period: 1/1953 P1/1969 Conaretrsts 4th Degree Polynomial (en or 0,ms—e 0) 4 At 267 2 mAMt + 2 esAs R 2 ’ .66 (3.46) ~ ° SE 384 OW— L75 in — 122 (273) 00— 56 ( 2.57) en, = ISO 17-34) at — .45 ( L437 mm 162 (425) ez 01 ( 08) in — 87(3657 ea .43 ( 318) en .06 ( .12) e — 54 ( 2477 2,,, 557 (8067 Ze 05 ( 17) rabol a C nd AT dolla ehng ntotal p dinglOMP non ntpe I ‘n qu,u’te S E_j on, eJ,an n money took aqua , En dlrchan -nh,h-ompoyo,etPe am p thtu ‘nq No - tattOo pp wuheach a es oncoeffse cab 1, pareti, E’, be 05 at nun hda unsent ,ablewh , epbunedhyv ian nth undpendtn a’abl Eu U, tanad ro I, ‘ma fl-Wi hobu atm in The pattern of thc coefficients indicates a large and rapid influence of monetary actions on total spending relative to that of fiscal actions.” Changes in high-employment expenditures, with the money stock held constant, first have a positive influence on total spending, but the influence becomes signifi- cantly negative after three quarters. Fiscal actions, unaccompanied by changes in money, have little net effect on CNP over five quarters.’4 For short periods, and for extended periods in which the rate of change of Federal expenditures is either accelerating or de- celeratirig, fiscal effects are significant. The estimated coefficients for changes in money and changes in Federal expenditures are in general agreement with the monetarist view of the response of total spending to these two variables. The specification of the total spending equation, as shown in Table I, has been criticized as being incom- ‘ 3 Andersen and Jordan tried several measures of fiscal actions in their basic equation. The best results were obtained by using only high-employment expenditures, rather than the high-employment surplus or both high-employment ex- penditures and receipts. They justify their choice by ap- pealing to the notion that financing expenditures by borrow- ing from the public and taxes have essentially the same impact on total GNP. For some results that contradict those of Andersen and Jordan, see E. C. Corrigan, “The Measure- ment and Relative Importance of Fiscal Policy,” forthcom- ing in Federal Reserve Bank of New York Monthly Review. It should be repeated that, a priori, specification of the total spending equation was sufficiently general as to be consistent with a numher of theories of CNP detennination, ‘-‘Andersen and Jordan, p. 18, indicate that these results are consistent with a “crowding-out” theory of eftects of gov- ernmnent spending. Page 11 FEDERAL RESERVE BANK or sr LOUIS -— - — APRIL 1970 plete in that it allegedly ignores the effects of interest rates on velocity.’5 However, since the spending equation is a reduced form, such effects are embodied in the coefficients of money.’° flUIItCS The total spending equation is the cornerstone of the model, providing its monetarist character. The focus of this paper, hosvever, is on determining the division of the change in total spending between price and output changes. Price changes are estimated as a function of (1) current and past demand pres- sure, and (2) anticipated price change. Demand pressure — As a measure of demand pres- sure on prices, the change in total spending is related to the potential change in output (GNP in constant prices).’7 These two variables, when combined, provide a measure of the economy’s demand for goods and services relative to its capacity to supply goods and services, The change in prices is specified as a positively related linear function of this measure of demand pressure (see Appendix A). Demand pressure, Dn, is defined as: Dr = Al, _(Xr, — Xn_,), where AY, is the change in total spending in quarter t; X”, is potential (full employment) GNP in 1958 prices in quarter t; and X,_, is real GNP in the previous quarter.’8 Given the GNP gap, defined as X~’, X,—,, the larger is the change in total spending (AY,), the greater is the spillover into higher prices. Given ~Y,, the larger is X”~— X,—,, the greater is the expansion of output and the less the spillover into higher knees. In addition to current values, past values of the demand pressure variable are included in the price equation. The purpose of including past values is to allow for lags in th0 detennination of prices in re- sponse to changing de,nand. Furthennore, the impact of changing demand through changing input prices and costs of production is given a chance to operate by including lagged values for the demand pressure variable, Anticipated Price Change — The other independent variable included in the price equation is’’~inieasure of anticipated price change (AP’n). The purpose of including this variable as a factor influencing current changes in the price level is to allow anticipations of future price movements to influence the decisions of market participants. Since such a variable is not bI 1 ANTICIPAT 1) PRICE DEFtNFTION p from Løng T us late oat Rote Eq eStee) ‘ Pl 1~ 4) 01 5] ps 02 py 08 p 03 p 06 p, 06 p 0 p 36 p—04 p 06 r ~ pa 03 p-—04 06 p,r 2 p .07 pa 7 Pi 96 yrabolaa C tee I s oil a a 15 a S haagemnawr 01 (10 00) 4 ~n namlo en e tat I t (0 40 mu tat (ON? mu p’eea) luqu observable, it has to be constructed. This is accom- plished by assuming that anticipations about future price changes are formed on the basis of past price experience. The measure of price anticipations used in this study is a by-product of estimating long-term market interest rates.’° Yohe and Karnosky showed that long-term market interest rates respond to price an- ticipations of borrowers and lenders, since commit- ‘~‘Forother ways of handling expectations, see Appendix C on alternative price equations. tmm See Paul S. Anderson, “Monetary Velocity in Empirical Analysis,” in Controlling Monetary Aggregate,-, Proceedings of the Monetary Conference held on Nantucket Island (June 1969), pp. 37-51, and the discussion of that paper by Leonall C. Andersen, pp. 52-55. See also Henry A. Latand, “A Note on Monetary Policy, Interest Rates and Income Velocity,” Southern Economic Journal (January 1970), pp. 328-30. ‘ 6 See A. A. Walters, “Monetary Multipliers in the U. K.: 1880-1962,” Oxford Economic Papers (November 1966): ‘ 7 This measure was apparently first used by Ray Fair of Princeton University. See his “The Determination of Ag- gregate Price Changes,” forthcoming in the Journal of Political Economy. For a similar specification of a price equation, see Milton Friedman, “A Theoretical Framework for Monetary Analysis,” also forthcoming in the Journal of Political Economy. See also a paper by William Considine of Stanford University, “Public Policy and the Current Inflation,” prepared as a part of a summer intern program at the U.S. Treasury Department (Septe,nber 5, 1969). ‘- 5 The series on potential output is based on that used by the Council of Economic Advisers. Currently, potential output is estin,ated to be rising at a 4.3 per cent annual rate. For alternative estimates of potential output, see Fair, “The Determination of Aggregate Price Changes.” Page 12 FEDERAL RESERVE BANK OF St. LOUIS APRIL, 1970 ments to borrow and lend funds nec1uirc’ an assessment of anticipated changes in the priee level for the period of the loan. The problem consists of isolating this price effect on market intc’rest rates from factors in- fluencing the real rate. In the process of constructing a measure of antici- pated price change, past changes in prices are ad- justeci by a summary measure of current c’eononlic conditions. Since price changes tend to lag changes in total spending, the degree of resource utilization as measured by the unemployment rate is used as a leading indicator of future price movements.’°For example, if unemployment is rising relative to the labor force, decision-making economic units would tend to discount current inflation in forming anticipa- tions about future price move,nents. Reflecting this consideration, the price change in each quarter is divided by an index of the unc-inplovment rate ap- plicable to that quarter. Thus the measure of price anticipations would be less for a given inflation rate accompanied by high or rising unemployment than when unemployment is low or falling. The specific definition of price anticipations is shown in Table II. The weights and the length of the lag period were obtained from the estimated long- tenn interest rate equation.2’ Estimated price equation — The estimated price equation is shown in Table III, where AP, is defined as the dollar change in total spending due to price changes in quarter t. The influence on prices of the demand pressure variable, D—,, is significant and positive for five quarters but very small thereafter.22 The pattern of influence is one of steady decay, \Vith 70 per cent of the total effect of demand pressure taking place in the first three quarters and 95 per cent in the first five quarters. Anticipated price change, represented by AP5,, is a significant determinant of current price change. ‘ 0 For purposes of exposition the unemployment rate was not included in the definition of anticipated price change in Exhibits I and II. 2tThe price expectations variable as shown in Table II is scaled in dollar units. This transforn,ation is ,nade because prices are estimated as the dollar change in total spending due to price changes. 22 M/hen the price equation is estimated with the components of D,~ separated, the coefficients for the AY, portion are not statistically significant at the five per cent level, im- plying that the gap portion, (XF, - X,,), explains most of the changes in AP,. However, there may be eollinearity problems which influence the estimated coefficients. Fur- the,’more, the D,_u form is used because, theoretically, it is a measure of excess demand (see Appendix A), ~Iat tIm PRICE EQUATION Sample Period I 1955 IV 1949 Conur met uS Oo9r a Pofynotaucd - (4— 0 1 du_0) A~ 270 , do,, 86AP R 87 (70) ‘ (855 SE 107 0~W 141 4 02(263 Cs 01 (186) 4 02 ~633} 6 11 81 & 0 (66) 4 09 fLI 4 0 (29) 8 a I ii e,nt C I I d ‘e n a r S an n I a Ce’, a t S ~I s I t on I P ‘ tub a I ut .5 tat 8 11, t 1’ ‘ mn, I,, n ‘.m2k n i n We baa! Though significant the measure of the impact of this variable should not be taken too literally because its construction indicates that it cannot really be viexved independently of the demand pressure van ible. ‘~ The influence of the e tno variables should perhaps b’ viewed in combination rather than as independent and separate influences.’4 Determination of output — Given AY as deter- mined by the total spending equation and ~P from the price equation the dollar change in total spend ing due to output changes defined as can be dcnved from the follouing identity: AY, — Al’ + AX, (P — P -,) (X X—,). The cross product term is a sumcd equal to zero.25 Thus, AX, = AY, — AP,. 23 From the standpoint of the model as a unit, price an- ticipations are important only in determining the division of total spending between prices and output, not the level (or change) of spending itself. To allow for the possible direct influence of price expectations on total spending, the spending equation was estimated with the price anticipa- tions variable. l’he coefficient of the price anticipations variable was not significant for this specification. 2 ’There is, however, some evidence that the price anticipa- tions variable may he interpreted as an independent and separate influence. When the price equation is estimated without Al’ 5 1 , the sun, of the coefficients on I),—, is only slightly more than shown in Table HI and the standard error is increased considerably. ~“The value of this cross-product term was calculated from 1953 to the present and provide-s ample justification for Page 13 FEDERAL RESERVE BANK OF St LOUIS APRIL, 1970 model. Of interest in evaluating the model as a unit is the implied pattern for the endogenous variables when only an initial set of lagged endogenous varia- bles and the time paths of the exogenous policy variables (money stock and high-employment Federal expenditures) are assumned known. To conduct such a test, several dynamic simulation experiments were performed. These simulations take the form of cx post dynamic simulations and an cx ante dynamic simulationY8 An cx post dynamic simulation is confined to the sample period from \vhich the estimated relation- ships are derived. Actual values for all current and lagged exogenous variables are used, but only initial actual values for the lagged endogenous variables are used. The model generates solution values for the endogenous variables in the first simulation pe- riod, xvhich are then used to generate solution values for the second period, and so on for each succeeding period.29 A comparison of these calculated time paths for the endogenous variables with their actual time paths enables one to formulate some judgment as to how well the niodel performs as an interdependent unit in tracking the movements of certain strategic economic variables. Ex post dynamic simulations were conducted for several subperiods within the sample period (1955- 69). The results for the entire sample period are summarized in Chart I on the next page. When simulations are conducted for subperiods within the 1955-69 period, the pattern of movement as shown for the whole period simulation tends to hold, but the levels are closer to the actual values at the be- ginning of each subperiod. ables quite well during the 1955-69 period. Since criteria for judging the performance of the model in such a simulation have not been developed, any con- clusions are necessarily subjective,30 The tendency for the model to avoid diverging sharply from the actual path for extended periods is an especially important feature. Such a feature provides some basis for trust- ing the tracking ability of the model over several quarters, even if on a quarter-by-quarter basis it may appear to be off the mark. To gain additional information about the predictive performance of the model, a comparison is made with an cx post simulation from another model. Results of an cx post simulation for 1963 and 1964 have been published for the Wharton model. The results for the model are compared with those of the Wharton model in Table VII. The period 1963-64 includes the 1964 tax cut, which, according to the Wharton model, is considered an important factor influencing economic developments in 1964. However, the St. Louis model, which does not emphasize such fiscal actions, did about as well, on average, for the years 1963 and 1964 (see Table VII). The main difference to be remembered in eval- uating these simulations is that the St. Louis model contains three primary exogenous variables, while the Wharton model contains forty-three. :uoSee Robert H. Rasche and Harold T. Shapiro, “The FEB. - M.I.T. Econometric Model: Its Special Features,” American Economic Review, Papers and Proceedings (May 1968), p. 142. Chart 1 indicates that the model tends to track the move- ment of the endogenous van- 25 For a discussion of the different ways of assessing the tracking ability of econometric models, see Carl F. Christ, “Econometric Mod- els of the Financial Sector,” forth- coining in the Journal of Money, Credit, and Banking. For a dis- cussion of simulation procedures and results with an income-expen- diture model, see Evans and Klein, pp. 50-69. tm9 See de Leeuw and Gramlich, “The Channels of Monetary Policy p. 485. Page 16 Toble VII ALTERNATIVE EX POST SIMULATIONS: ACTUAL MINUS PREDICTED1 Comparison of Wharton and St. Louis Models for 1963-64 Nominal GNP 2 Real GNP’ Price Level Unemplolmen? Rate’ Wharton St. Louis Wharton St ouk Wharton St .ois Wharo.i St tours 1963. I 4.6 0.4 3.9 0.4 01 02 09 0.3 II 0.2 0.3 0.4 0.7 0.1 0.2 0.7 0.1 III 1.3 1.5 2.5 0.6 0.3 Ci 0.9 0.1 IV 0.9 2.1 2.2 04 02 05 1.2 02 1964 I 0.9 1.7 2.7 1.4 03 Co 14 02 II 1.1 0.1 2.3 3.2 -0.3 06 14 02 1.5 1.7 4.0 --2.7 0.5 0.8 16 02 IV 0 --1.7 2.2 68 0.4 09 12 0.2 Average Error 0.11 076 1.55 1 88 0.28 049 1.16 0.16 Root Mean Squared Error 200 1.49 2.92 309 0.33 0.60 1.28 021 $:sr.,plr. ps., 4 E~,no-rs..u. car silOs \\ lssu:., It’ It-I :1.. WI: rn’,, Si. I . l:c..—:~’b.’ st. I ‘rn. :!C,.l..., of b-liar-u. Unns;rutr.d fr.,m the eel of impl,est -rice rId an5 Per cent. Srmrrru M. K. K’r 1 rst and L. B. KleNs. 7k. 11 he,!.,,. irp,.o.,,tr.r ct .u...W,4s, 2r.rI Er:iar,’rri Edition Phila,l,’l:,hia:t.ricersr..y ‘‘I Pr nn—vh~::in, ]9~c . mi Er—an’ R,.e—,e i:rmnk ot St I.,siie. FEDERAL RESERVE BANK OF ST. LOUIS APRIL. 1970 Page 17 FEDERAL RESERVE BANK OF ST. LOUIS The comparison is not meant to imply that the St. Louis model is superior. Rather, the suggestion is offered that a small model con- structed within a monetarist framework may yield as much informa- tion about the key aggregates as a large structural model. In summary, small monetarist models may be useful as a guide in the formulation of stabilization policy. i%nte ~ An cx ante dynamic simulation is like an cx post dynamic simula- tion, except that it extends beyond the sample period. To conduct such a simulation for this model, it was necessary to re-estimate the model for a subpeniod within the full sample period. All equations of the model were re-estimated with data through 1967. The period of the cx ante dynamic simulation is 1968 and 1969. The results are summarized in Chart II and in Tables VIII and IX. The success of the cx ante dynamic simulation can be assessed by comparing it with the tracking record of the cx post simulation for the same period. A comparison of the errors associated with the cx ante simulation with those of the cx post simulation (where the errors in both cases are computed with reference to actual values) suggests that any structural shifts that occurred in the 1968-69 period were not of such a magnitude that the cx ante tracking ability of the model was significantly different from that of cx post simulation. Any conclusions about the tracking ability of the model are neces- sarily tentative, because they are based on only one cx ante dynamic simulation experiment. Nevertheless, these results provide a tentative basis for confidence in the tracking ability of the model in es- Table VIII MODEL SIMULATIONS 1968 1969 I II Ill IV I II Ill IV GM? t veI (Balboa of dollars) Acluat 835 3 858 7 8764 892.5 908.7 924 8 942 8 952 2 ExAntet 8341 8567 8789 8999 9176 9323 9459 9572 Ex Post2 8346 854.7 *77.7 897.8 9149 929.4 9434 9551 Annual Rate at Change in V Atuol 97 11.7 85 76 75 73 8.0 40 ExActs 9.1 11,3 10.8 9.9 81 66 60 43 E,rPast 94 11,0 102 9.5 7.8 65 6.1 51 Annual Rote of Change in X Actual 5 9 7.4 40 3.2 2.6 2.0 2.2 0.4 ExAnte 5.7 7-6 70 59 40 24 17 07 ExPot 5.4 6.7 57 4.8 3.0 16 11 0.0 Annual Rate af Change in P Actual 37 40 40 43 49 52 54 4.7 Es Ante 3.3 3,4 3.6 3.8 4,0 4.1 4.2 42 EAPost 38 40 4.3 44 47 4.9 5,0 50 Unemployment Rate (per cent) Actual 3.7 3.6 3 6 3.4 3.4 3.5 3 4 3 6 ExAnte 39 3.8 35 33 32 32 33 3.5 Es Post 3.9 37 35 3.4 34 3.5 3.7 39 Corparate Aoa Rate (per cent) Actual 61 6.3 61 6,2 67 69 7.1 75 ExAtte 58 5.9 60 6.4 6.5 67 7,0 71 E Post 59 60 61 65 6.7 6.9 72 7.4 Commercial Paper Rate (per cent) Actual 5 6 6.1 &0 60 6J 74 8 5 84 8 Ante 51 6.1 Si 56 57 60 66 67 Es Post 58 58 5.8 6.5 66 70 78 80 IC to bbreviations V Nominal ON? N Real ON? P ON? Iiitee deftato Sunulatton based a equation est,n-sated h ough Tv/ba?. i5 ulatson based on squat o a estimated tbmugh mV/igsa rage is APRIL, 1970 Results of Ex Ante Dynamic Simulation i toni Percent Per Cent 12 2 ii 11 10 9 9 8 8 7 7 6 6 5 5 4 4 3 3 2 2 0 0 Real ONP 8 7 7 6 6 5 5 $ 4 3 3 2 2 0 2 -2 Implicit Price_Deflator 6 --______ --— - -~ 6 5 -~— ‘. 5 Ac Is a I ___- -__ 1967 1968 1969 - me,, c FEDERAL RESERVE BANK OF ST. LOUIS APRIL, 1970 is incomplete for long-run analysis; nevertheless, it yields results that are of interest and may not be too far removed from results that might evolve from a more complete specification.31 When simulations are conducted for long periods into the future (thirty years), the model demonstrates properties consistent with those expounded by the classical economists. Over the long run, monetary actions have no effect on real magnitudes; the rate of growth of output, the unemployment rate, and the real rate of interest all tend to move toward some equilibrium rate, regardless of which rate of money growth is maintained. The effects of alterna- tive rates of monetary expansion are on nominal magnitudes, namely, total spending, prices, and mar- ket interest rates. Based on the assumptions of the model, a six per cent rate of growth in money along with a six per cent growth rate in Federal expenditures, for example, would lead ultimately to about a six per cent rate of growth in total spending, a four per cent rate of growth in output, a two per cent rate of increase in prices, and market interest rates about two percent- age points in excess of the real rate. Alternatively, a two per cent growth rate in money would result ap- proximately in a two per cent growth in total spend- ing, a four per cent rate of growth in output, a two per cent rate of decline in prices, and market interest rates about two percentage points below the real rate, Over the long run, the model indicates that high employment and price stability are compatible. Summary The main purpose of this study has been to quantify the effects of monetary and fiscal actions 3tThe shortcomings of the model for the long-run analysis are quite evident. There are no assumptions specified as to labor force growth and productivity. Furthermore, there is no investment function and, therefore, the capital stock is not an endogenous variable. All long-run assumptions are embodied in assumptions about the growth rate ot potential output. With these assumptions, policy actions cannot affect the economy’s long-run growth rate. within a small-model framework and thereby offer an alternative to existing large-scale econometric models. Such a model has been formulated and the effects of monetary and fiscal actions on spending, output, prices, employment, and interest rates have been estimated. The model developed in this article is primarily “monetarist” in character, The estimated equations indicate that monetary actions, as measured by changes in the money stock, play a strategic role. Fiscal actions, as measured by high-employment Federal expenditures, have some short-run effects, but for periods of a year or more the net effect on spending, output, and prices is near zero. Simula- tions of alternative rates of monetary expansion pro- duce short-run and long-mn responses which are consistent with the general monetarist view of the economy. One of the chief advantages of this model is that it depends primarily on information about only two variables — the money stock and high-employment expenditures.32 Considerable insight can be gained about the pattern of expected movements of certain strategic economic variables by considering alterna- tive courses of monetary and fiscal actions. How- ever, since the model is limited to only monetary and fiscal influences, to the exclusion of other inde- pendent forces, it is not suitable for exact forecast- ing.33 Its primary purpose is to measure the general pattern of influence of monetary and fiscal actions on several strategic economic variables. Since the econ- omy is viewed as being basically stable, other factors influencing total spending, output, and prices are not considered to be of great importance in estimating the response to monetary and fiscal actions. i2This feature has led John Deaver to conjecture that the standard error of forecast in the Andersen-Jordan model may he far lower than that of the FEB-MIT model. See his “Monetary Model Building,” Business Economics, (Sep- tember 1969), p. 30. 33 See Andersen and Jordan, pp. 15, 23, 24, and Leonall C. Andersen, “Money in Economic Forecasting,” Business Economics, (September 1969), p. 17. This article is available as Reprint No. 55. The Appendices to this article begin on the next page. Page 21 FEDERAL RESERVE BANK OF ST. LOUIS APRIL. 1970 APPENDIX A EXPLANATION OF THE PRICE EQUATION The and lags) isprice equation (omitting timescripts = f (D, ~pA) where D, demand pressure, is defined as D = — (X 5 — X). AY is the change in total spending, (X” — X) is the G~~Pgap, that is, the difference between potential and actual output, and ~p~’ is anticipated price change. This specification of the price equation is based on standard theory of macroeconomic equilibrium. Macroeconomic equilibrium can be depicted graph- ically as in Figure I. The solid downward-sloping line, is the total spending line, which represents the combinations of prices and output consistent with a particular level of total spending, Y. This total spending line can he interpreted as total demand for output. P Figure I Macroeconomic Equilibrium (Determination of Output and Prices) The upward-sloping line, labeled X5, is the total supply line. This line corresponds to that combination of prices and output which maximizes profits of firms, given the prices of factors of production, the degree of competition among firms and the stock of human and physical capital (defined to embody the state of technology). The intersection of total supply and total demand detennines the levels of output and prices. The equilib- rium price level is that level which equates the amount of output supplied \vith the amount demanded. The focus of the model is on the change in prices and the change in output. In terms of Figure I, changes in prices and output are brought about by shifts in demand and/or supply. Since X~is drawn for a level of total spending, a shift of that line upward and to the right to ~ represents an increase in total spending. If the total supply line remains fixed, the effect of AY on prices depends on (1) the magnitude of AY, and on (2) the slope of the total supply line, X5 The purpose of the model is to estimate the response of spending, output, and prices to monetary and fiscal actions, not to test a hypothesized structure. Conse- quently, rather than attempt to determine the shape of the total supply line empirically, its variable slope is proxied by the difference between potential output and actual output. As drawn in Figure I, there is a one-to-one relationship between X’~ X5 and the slope of X5. As- suming that this relationship is approximately linear within the range of experience since 1955, and that the observed values fall on the supply line, the effect of a variable slope for X~can be approximated by X5 — X. In this way the term [AY — (XF — X)l brings to- gether both the magnitude of demand shift and the slope of the supply line. The other term in the price equation, anticipated price change, APA, is considered as a separate influence on prices. In terms of Figure I, the anticipated price term is a shift parameter for the total supply line (an increase in APA shifts X5 upward and to the left). Including it in this way allows for the influence of past prices on current pricing policies of firms and factors of production. ‘ N x Page 22 FEDERAL RESERVE BANK OF ST. LOUIS APRIL, 1970 APPENDIX B GRAPHICAL ILLUSTRATION OF THE MODEL The workings of the model can be demonstrated with graphical techniques. Figure II is a representation of the core 0f the model, showing the determination of changes in spending, output, and prices. Panel A of Figure II is a graphical representation of the total spending equation with AM on the horizontal axis and AY on the vertical axis. Changes in AE shift the total spending line. Panel B shows prices (AP) as a function of A?. A short-run price line (APs) is drawn consistent with empir- ical results showing that AP is not very sensitive to AY in the short run. Important determinants of the position of the short-mn price line are the size of the CNP gap and anticipated price changes. The long-run price line (AT’ (LE)) is drawn to show the relationship between AP and AY when the CNP gap is zero and anticipated prices are equal to actual prices. Its slope (45 degrees ày St 2 ÀY 1 ax (A) Figure Model in Graphical Form from its origin on the AY axis) is based on the monetarist view that in the long run, AM influences only AP. Panel C expresses the total spending identity in graph- ical terms. Total spending is divided between output and prices; to reflect this, the line in panel C is drawn as a 45 degree line with its position determined by the magni- tude of total spending (AY). There is a family of 45 degree lines, one for each possible AX’. Also included in panel C is a horizontal line representing the long-mn growth rate of output. It is shown as a horizontal line to indicate that long-run output growth is exogenously de- termined by resource growth and technology. In panel D, the AX1 line shows the relationship between money (AM) and output (AX) as derived from the other three panels. The equation for this line is not shown in Exhibit I in the text, but it can be derived from the other equations of the model. Figure II is drawn to represent an initial equilibrium for a given AM, which has associated with it the short- run price and output lines, AP1 and AX,. The effect of a change in AM, given AE, is shown as a movement along the spending line in panel A from 0 to ~ Given the initial price line, AP1, and the changed AY, the effect on prices and output is shown in panels B, C and D as a movement from 0 to 0 This case illustrates the impact of a change in AM in the short run. For longer periods, anticipated price changes and the CNP gap will also change; they become endogenous variables in a long-run model. To illustrate the ef- fects for the long run, the long-run price line, AP(LR), in panel B, is rele- vant. The interpretation of the long- run price line is that changes in AM are reflected only in AP, with AX de- termined by considerations of resource growth and technology. The horizontal line in panels C and D is the long-run relation between prices and output. In the short-mn, the solution of the model need not lie on the long-run price line in panel B (or the long-mn output line in panels C and D). How- ever, a -succession of short-runs (shown as a shift of the AP and AX lines to AP. and AX2) will tend to move equi- librium toward the long-run price and output lines, as anticipated prices ad- just to actual prices and the GNP gap goes to zero. Spending Equation (B) Price Equation Page 23