OUTPUT/INFLATION DYNAMICS Derivations and Analysis: Dynamic Aggregate Demand (DAD) curve and the Dynamic Short-run Aggregate Supply (DSAS) curve Jeremiah Allen ©1982; 2004 II. SAS The Model The derivation of the Dynamic Short-run Aggregate Supply curve, DSAS, begins with the derivation of the static Short-run Aggregate Supply curve, SAS. The derivation of the SAS begins with an observed empirical regularity, the Phillips curve. [Note: BJM begin here too. They add mark-up pricing to get to their Aggregate Supply curve in Chapter 9. This isn’t necessary; we need only the empirical fact that Prices don’t fall.] In 1958 A. W. Phillips observed a remarkably stable relation between the unemployment rate, u, and the rate of change of money-wages, W, in the UK for nearly 100 years: 1861 –1957. There were three interesting features of the Phillips relation: 1) It appeared to be stable over 100 years. 2) It was non-linear. 3) Each business cycle followed counter-clockwise loops. Richard Lipsey, a few years later, provided a simple Demand and Supply model, with downward wage rigidities, as the theoretical basis for the Phillips curve. In a lovely piece of theory, Lipsey also provided an explanation for the loops. [This explanation of the loops forms the basis for the labour market dynamics we’ll see later.] For the theoretical analysis at this stage, I ignore the loops and treat the Phillips curve as linear. There are two SAS curves: the Flexible-Price SAS(FP) and the New-Keynesian SAS(NK). As with the MAS curve, with FP aggregate prices, P, can fall; that is, inflation, , can be negative. So the SAS(FP) is a straight line. With NK aggregate prices can not fall; that is, inflation is constrained: 0. So the SAS(NK) is horizontal at = 0. NK is more realistic since Prices have not fallen measurably in an industrial market economy since 1938, and it is closer to the curve of the original Phillips curve. For the derivation of the SAS curve, first transform the Phillips curve with the observation that money-wage increases are the sum of productivity increases (prod) and inflation (): W = (prod) + . This shifts the horizontal axis up from the line [W = 0] to the line [ = 0], and changes the vertical axis from W to . Now write the equation of a linear approximation to the transformed Phillips curve: W = – r(u), where u is the unemployment rate and r is a parameter (II. i) Add the transformation: t + (prod) = – r(u); and subtract (prod) from the left-hand side: t = – r(u). (II. ii) Now denote the point where the transformed linear approximation crosses the new abscissa – the u axis here – as un. un is “full employment”. In Friedman’s language it is “the natural rate of unemployment”, or, in a more neutral term, it is the “NAIRU” (Non-Accelerating Inflation Rate of Unemployment.) This is the level of unemployment associated by Okun’s Law with potential output, Yn. Add this transformation: The linear approximation to the twice transformed Phillips curve is: t = –ř(u – un). which is shown on Figure II-1. (II. iii) DAD and DSAS page 2 Figure II-1 transformed Phillips curve linear approximation (II. iii) u Finally, substitute Okun’s Law for the right hand side. Okun’s Law is: u= un –v[(Y – Yn)/Yn] (u – un) = –v[(Y – Yn)/Yn] (II. iv) where v is “Okun’s constant”. Okun’s constant was once considered to be around 1/3, but indications are that it is now somewhat higher. Substitute Okun’s Law, equation II. iv, into equation II. iii. This gives a linear SAS: SAS: = f [(Y – Yn)/Yn], (II. v) where f = (ř v). The two SAS curves are shown on Figure II-2. Figure II-2 MAS SAS(FP&NK) SAS(NK) 0 Y Yn SAS(FP) DAD and DSAS III. page 3 DSAS A. The Model The seeming stability of the Phillips curve, and thus of the SAS curve, was challenged initially by Milton Friedman, in the late 1960s. The “challenge” was formalized by Edmund Phelps a few years later. The basis of the challenge is that the SAS represents unacceptably stupid behavior, if inflation persists. Their extension of the Phillips curve analysis makes it dynamic, and that analysis is where we begin the derivation of the Dynamic Short-run Aggregate Supply curve, the DSAS curve The underlying model of the Phillips curve, as developed by Lipsey, is simple Demand and Supply in labour markets. Over the period Phillips observed his curve, inflation never persisted. There were always cycles; expansions always ended quickly. Money-wages increased during expansions, when excess supply of labour (unemployment) was low or negative. But as the economy shifted from expansion to contraction, the rate of money-wage inflation quickly dropped as excess supply (unemployment) increased. Friedman argued – and virtually all macro-economists now agree – that if governments tried to hold the rate of unemployment at less than un for some length of time by keeping inflation constant at a positive level, then that level of inflation would become expected. This would cause the Phillips curve, and thus the SAS curve, to shift up. The shift is caused by people – both buyers and sellers in the labour market – coming to expect the level of inflation that the government is maintaining. During the period Phillips observed, this had never happened. But once governments began to think they could permanently trade-off a constant level of inflation for a level of unemployment below un, it could happen, and probably did. This is nicely described in the Bruce reading, but I’ll describe it briefly here too. The argument is simple: that the Phillips curve, and thus the SAS curve, will shift up or down as expectations rise or fall. The DSAS curve builds those shifts into its equation. The equation has the DSAS curves intersect the MAS curve at the expected rate of inflation,e. We model these shifts as taking place in discrete time, just like our shifts of the DAD. So we start by giving time subscript, t, to the variables in SAS, which becomes SASt :t = f [(Yt – Yn)/Yn] . (II. v) [At this stage, having derived the SAS curve from the Phillips curve, I will derive DSASt from the SASt, rather than beginning with the Phillips curve again.] Consider Figure III-1 below. Suppose the government were to attempt to hold the economy at Point A, with inflation of x% and output of YX > Yn ↔ uX < un The government would do this if it thought SAS0 was permanent and stable, and if it wanted to keep the unemployment rate at uX, which is determined by Okun’s Law from level of output, YX . But after a period of inflation equal to x%, people begin to expect inflation of x% (e = x%). The SASt curve shifts up – from SAS0 to SAS1 – to cross Yn at = x% . This insight led to the DSAS curve, which is derived by having the SAS curve augmented with expectations. The DSASt is the short-run aggregate supply curve we will use. The equation of the DSASt is: DSASt: t = f [(Yt – Yn)/Yn] + et . (III. i) Note that this is not yet quite dynamic; there are no explicit lags in the DSASt equation. The lag structure has to enter through expectations – that is, by making et a function of lagged DAD and DSAS page 4 variables. (There is a lag in DSASt; Yt is a function of Yt-1, which is a function of t-1 , but it’s implicit not explicit.) Figure III-1 MAS NK SAS1 1 SAS0 x% 0 0 YX FP Y Yn This now forces attention to how expectations are formed. Notice that we now have two equations, DSASt and DADt, but we now have three unknowns: Yt , t , and et . To able to solve these we need a third equation. The equation I will use here is the one developed by Friedman and Phelps, and is called “adaptive expectations”. The basic idea is that people develop expectations on the basis of experience, and that it takes some time for experience to change expectations. The equation of adaptive expectations is: AEt: et = et-1 + g(t-1 – et-1); 0<g≤1 (III. ii) Equation (III. ii ) gives a value of et which is determined exclusively by lagged values of itself and ; these are all pre-determined. This also converts DSASt to be explicitly dynamic; et is a function of t-1 , giving an explicit lag structure. Note that if g = 1, et = t-1 ; any period’s inflation immediately becomes the next period’s expected inflation. But if g < 1 and t-1 is held constant, expectations converge to t-1. [NOTE: Expectations formation is still controversial. Variations include the strangely-named Rational Expectations models. Note also, that while it hasn’t been explicitly formulated, or put into a model, it’s clear from observations – especially in the major recessions of the early 1980s and early 1990s – that inflation falls, and quickly, when unemployment becomes large. Thus, it would be both “adaptive” and “rational” to have lower inflationary expectations when unemployment is substantially higher than un. In a more fully developed model, then, (ut – un) would be an argument in the equation of expectations. We won’t do that; adaptive expectations works well and is much simpler to work with.] We have now three equations, DADt, DSASt and AEt, and three unknowns: Yt , t , and et . The system has a solution. Reduced-form equations, in either Yt or t , can be found from these three equations. [The equation for et is already a reduced-form equation.] DAD and DSAS page 5 __________________________________________________ EXERCISE: Find the reduced-form first-order difference equations for Yt and t . What will be the dynamic behavior of each? Your answer to that last question for Yt will turn out to be wrong. Can you guess why? ---------------------------------------------------------------------------------One solution, the one that will be the starting point for two-thirds of the analysis below, is shown on Figure III-2: Figure III-2 DAD0 MAS DSAS0 0 Yn Y To work out solutions for this system, I suggest the following algorithm: 1. You begin with all variables in Period 0: Y0 , 0, ande0. These values are always given. 2. A good short cut is to substitute the DSASt: t = f [(Y1 – Yn) / Yn ] + et-1 , for t in DADt and solve for a reduced form equation in Yt. You will have an equation in just Yt , Yt-1 , et , and numbers. [This equation is given on Problem 4; it is: Yt = [Yt-1 + at + mt + f + et]/[1 + (f /Yn)] The divisor is all numbers, so find this equation as numbers and variables. You will use this a lot, so keep it handy; it will make all solutions easier. Call it the working equation.] 3. Find e1 . This is entirely determined by lagged values, that is, by values of e0 and 0 . 4. Substitute the values of Y0 and e1 into the working equation; that equation is now all numbers and Y1 ; solve for Y1. 5. Substitute this value of Y1 into DSAS1; that equation is now all numbers and 1 ; solve for 1 . This is recursive, so just keep repeating, substituting t=2 for t=1; then t=3 for t=2; then etc. EXCEPT: New Keynesian: 4. and 5: IFt < 0, set t = 0, and substitute t = 0 into DADt and solve for Yt. OR DAD and DSAS page 6 IF there has been a supply shock, then 1 > 0 > 0 exogenously. Substitute the exogenous value of 1 into DAD1 and solve for Y1. This works ONLY for Period 1. For all periods after, follow steps 3 thorough 5. B. Analysis The analysis is all in the Y, space. Our model is now fully defined in that space. The curves we use are the DADt, the DSASt, the MAS(NK), and the AEt equation. The basic model is shown in Figure III-2 above, beginning with a long-run equilibrium of full employment, Y0 = Yn, and no inflation, 0 = 0. This is the starting point for the first two parts of the analysis below. Those two are actions which cause inflation. The NK model and the FP model are the same for Periods 1 and 2, so the descriptions below apply to both. Because I don’t want to give you the answers to Problem #4 yet, I will describe only Period 1 and Period 2. I will show only DAD0 , DAD1 and DAD2 , and DSAS0 , DSAS1 which is the same as DSAS0 , and DSAS2 . I will use only g = 1 in the descriptions below. I.A on Problem #4: Expansionary policies This is illustrated in Figure III-3. The economy begins with full employment, Y0 = Yn, and no inflation, 0 = 0. In Period 1, the economy is hit with expansionary policy, either fiscal or exogenous-monetary, and DAD1 shifts up, to the right, of DAD0. This gives a Short-run value, Y1 , that is greater than Yn : 1050. It must be Short-run since it is greater than Yn – that is, it is not on the MAS curve. It also gives a value 1 , that is greater than 0: 1 = .05. Because 1 > 0, if g = 1 the DSAS curve shifts up by the amount, 1 , for Period 2. Since e2 = 1, DSAS2 crosses MAS at 1 = .05. What happens next depends on the monetary response and the values of the parameters, f and . With NO accommodation, and with the special parameter values of Problem #4, where f = 1 and = Yn, the DAD curve shifts back. With these parameter values, it shifts halfway back. So the DAD and DSAS in Period 2 intersect at Y2 = Yn and 2 = .05. If there is lagged accommodation, in Period 2 the DAD curve stays at DAD1. The DSAS curve has shifted up, as above. The intersection is halfway up the DAD curve from the equilibrium of Period 1. This is shown as the asterisked values of “2”, “2*” on Figure III-3 below. This continues for all time periods, the DSAS curve in each period shifting so the equilibrium moves a distance halfway up the DAD from the old equilibrium to the Medium Run Equilibrium, MRE, which is where DAD1 (= DADt for all t) crosses MAS. Figure III-3 DAD2 DAD1 2* DAD0 2 MRE MAS DSAS2 DSAS0,1 2* 1,2 1 0 0 Yn Y1 Y0,2 Y DAD and DSAS page 7 Y2* I.B on Problem #4: Supply Shock This is illustrated on Figure III-4. The economy begins with full employment, Y0 = Yn, and no inflation, 0 = 0. In Period 1, the economy is hit with a supply shock. Prices rise exogenously such that inflation in Period 1 is 1 =.10. The SAS curve shifts up, to be flat at 1 . This gives a short-run value of Y1 that is less than Yn. What happens after this depends on the monetary response and the values of the parameters, f, g, and . For g =1 and NO accommodation, the DAD curve shifts down to the left, crossing the Output axis at 900. DSAS shifts up to cross MAS at .10. The two intersect at y = 900 and = 0. With accommodation, DAD2 is the same as DAD1 and DSAS2 is the same as with no accommodation. The case with accommodation is shown as the asterisked values of “2”, “2*” on Figure III-4 below. Figure III-4 MAS DAD2 DAD0,1 SAS1 2* DSAS2 DSAS0 1 1 0 0 2 Y1,2 Yn Y MAS(NK) 2* II. on Problem #4: Disinflation This is illustrated on Figure III-5 for Case 1: “cold turkey”. The economy begins with full employment, Y0 = Yn, and 10% inflation, 0 = .10. In Period 1, the government instructs the monetary authorities to set mt = 0 for all t. DSAS1 is the same as DSAS0 because it intersects MAS at e1 0 . The DAD curve shifts down, to the left as m1 is set to zero. Output is reduced in Period 1: Y1 = 950 < Yn. Inflation is reduced in Period 1 because of the leftward shift of DAD1 : 1 = .05. In Period 2, DSAS2 is below DSAS0,1 . On Figure III-5, the case where g = 1 is illustrated, so DSAS2 crosses MAS at 1 = .05. The values given in Problem #4 will have the system move to zero inflation, 2 = 0, and unemployment, Y2 = 950 < Yn. DAD and DSAS page 8 In future periods, t = 3,n : in the NK model, Yt = 950, and t = 0, for all t; in the FP model, Aggregate prices, P, will fall so that inflation will be negative ( < 0), for several time periods. With FP, eventually full employment and zero inflation is reached. Figure III-5 DAD2 DAD0 DAD1 MAS 0 DSAS0,1 DSAS2 0 = .10 1 1 2 0 Y1,2 Yn Y