Stabilization policy with rational expectations. IAM ch 21. Ragnar Nymoen Department of Economics, UiO Revised 20 October 2009 ECON 3410/4410: Lecture 7 Backward-looking expectations (IAM 21.1) I From the notes to IAM Ch 20, we have the …nal equation for in‡ation = t 1 (1 + ) t 1 + (1 + ) zt + 1 (1 + ) st + (1 + ) (1) or: ( t )= 1 (1 + ) ( where = )+ t 1 2h (1 + zt = 2 b) 1 , = (1 + 1 1+ ) zt + , and (gt g ) + vt (1 + 2 b) ECON 3410/4410: Lecture 7 1 (1 + st ) (2) Backward-looking expectations (IAM 21.1) II We now investigate in‡ation expectations errors, assuming that t = 0 is the initial situation, and that expectation for period t = 1; 2; :: are made at the start of period 1, using information available at the end of period 0. Expectations are made for the (in…nite) horizon t = 1; 2; 3; :::::. Although we have so far referred to backward-looking expectations formation as static, we are in fact "using the model" to generate the agents’expectations for period t = 2; 3; :::. In more detail, the expectations formation is e t e t = = 0 , for t emod = t = 1 but t 1; for t = 2; 3 > 1 (3) (4) where emod denotes the model based forecast and t 1 is the t period t 1 solution obtained by the use of the …nal equation. ECON 3410/4410: Lecture 7 Backward-looking expectations (IAM 21.1) III Note that this do not entail that the agents has to know the model (i.e. to sit in this auditorium). Hence expectations are in fact not backward-looking in the strict sense of et = 0 for all t. To analyze the expectations errors, we assume that gt = g and vt = st = 0 both in the economy and in the expectations formation process. This is the same simpli…cation as on p 630 in IAM. The period 1 error (remember that 0 is the initial period): e1 = = = 1 e 1 =( 1 (1 + ) (1 + ) ( ( 0 ) 1 ) ( ( e 1 0 0) ECON 3410/4410: Lecture 7 ) ) (5) Backward-looking expectations (IAM 21.1) IV Period 2 error: e2 = e 2 2 = 1 1+ = ( 2 = ( =( ) 2 ( 2 ( 1 (1 + ) 0 )( 0 ) 1)( 0 ) e 2 ) ) ( 0 ) And generally: et = t 1 ( 1)( 0 ) for t = 2; 3; :: (6) which is equation (10) in IAM p 630. Assume that 0 = . Then (5) and (6) shows that the expectations generated by (3) and (4) are always accurate. ECON 3410/4410: Lecture 7 Backward-looking expectations (IAM 21.1) V If 0 6= , the forecasts are biased but become gradually closer to the in‡ation target. The in‡ation forecast during the phase of adjustment may be as depicted in Figure 21.1 in IAM, for the case of 0 = 3% and = 0. The critique against backward-looking expectations is that the forecast bias (systematic error) only goes away gradually, cf also the Lucas critique at the end of this set. ECON 3410/4410: Lecture 7 AD-AS model with rational expectations (IAM 21.2) I The rational expectations hypothesis (REH) equates agents’ subjective expectations about a variable, for example et+1 , with the mathematical expectation conditional on an information set It . Rational expectations are without systematic errors, which, in macroeconomics entails that the true model of the economy is part of the information set. With this extra assumption, RE amounts to model consistent expectations. The agents must know the model before they can form rational expectations. e Let etjt 1 and ytjt 1 denote the rational expectations for period t conditional on information available at the end of period t 1. ECON 3410/4410: Lecture 7 AD-AS model with rational expectations (IAM 21.2) II The short-run AD-AS model with RE is given by yt y = 1 (gt g) 2 rt = it e t+1jt 1 ; it = r+ e t+1jt 1 e +b ytjt t = e tjt 1 (rt + (yt (7) (8) +h 1 r ) + vt ; e t1jt 1 (9) y ; y ) + st : (10) vt and st are assumed to have zero expectations. They have variances 2v and 2s ; and they are uncorrelated. For simplicity we set e gtjt 1 =g ECON 3410/4410: Lecture 7 Solving the RE model I Step 1 Express yt and yt t = y+ = t in terms of expectations and exogenous variables 1 (gt h + 2 h e tjt 1 + + h 2 h g) e tjt 1 1 (gt e tjt 1 e + b(ytjt i y ) + vt 1 g) e + b(ytjt 1 (11) (12) i y ) + vt + st : ECON 3410/4410: Lecture 7 Solving the RE model II Step 2 Obtain the functional relationships for the mathematical e expectations etjt 1 and ytjt 1 from the expressions in Step 1: e ytjt 1 = y+ and e tjt 1 1 | e gtjt = + 0 = h g + } 1 {z =0 e tjt 1 + n h 2 h e tjt 1 2 h e tjt 1 h e + b(ytjt 1 y) i (13) 1 e gtjt e tjt 1 1 g + e + b(ytjt e + b(ytjt 1 1 y) ECON 3410/4410: Lecture 7 y) io (14) Solving the RE model III Together with (13) this is seen to imply: e ytjt 1 e tjt 1 = y; + (14) = We now have found the rational expectations for GDP and in‡ation. Since these are variables in the AD and AS equations, the last step of the solution is to insert these REs into the expressions in Step 1. ECON 3410/4410: Lecture 7 Solving the RE model IV Step 3 Insertion back into (11) and (12) gives the rational expectations solution: yt t = y+ = + 1 (gt 1 g ) + vt (gt g ) + v t + st Which is a static model. There are no persistence or business cycles in this solution ECON 3410/4410: Lecture 7 (15) (16) Policy ine¤ectiveness I (gt g ) is a …scal policy “surprise”, which does not appear in IAM since they set gt = g from the outset. We have used the e weaker assumption that gtjt 1 = g . A simple model for gt which is consistent with this is gt = g + "gt where "gt has conditional expectation zero. Using, gt g = "gt ; the RE solution can be expressed as: yt t = y+ = 1 "gt + vt ; + vt + st : (17) (18) Since none of the policy variables or parameters ( , h or b) enter into (17), this is called the policy ine¤ectiveness proposition. ECON 3410/4410: Lecture 7 Policy ine¤ectiveness II Systematic monetary policy does not a¤ect GDP. This is because the rule based monetary policy cannot generate in‡ation surprises which drives yt via the AS function (yt y) = 1 ( t e tjt 1 ) + 1 st which in this form is known as Lucas’supply function. To avoid the conclusion about policy ine¤ectiveness the model needs to be modi…ed: Assume that the central bank acts on the basis of actual in‡ation and output. Instead of the Taylor rule (9), we use eq (28) in IAM rt = r + h ( t ) + b (yt y) to represent the situation that central bank can react to t and yt after in‡ation expectations have been formed in the private sector. ECON 3410/4410: Lecture 7 Policy e¤ectiveness under RE With the modi…ed Taylor rule, IAM p 636 and 637, shows that the RE solution is (use Step 1 - Step 3), and set gt = g as in the book): t yt (1 + 2 b)st + vt , 1 + 2 (b + h) vt 2 hst = y+ 1 + 2 (b + h) = + (19) (20) GPD is now in‡uenced by systematic monetary policy. For example: Increased weight (b) on GDP in Taylor rule reduces the impact of supply and demand shocks. The reason is that the central bank can a¤ect the real interest rate by changing the nominal rate after in‡ation expectations have been set. ECON 3410/4410: Lecture 7 Optimal stabilization policy under RE I The analysis is in terms of the same social loss function as in ch 20. What matters is the variability of in‡ation and GDP. Taking variances on both sides of (19) and (20): 2 y = 2 = 2 v +( [1 + 2 2 v 2 h) 2 2 s ; + h)]2 + (1 + 2 b)2 (21) 2 (b [1 + 2 (b 2 s + h)]2 (22) We also follow the method of …rst analyzing the case of pure demand shocks, and then the case of isolated supply shocks. ECON 3410/4410: Lecture 7 Optimal stabilization policy under RE II Demand shocks only ( 2 s 2 y = 2 = = 0): 2 v 1+ [1 + 2 (b + 2 2 v 2 (b h) + h)]2 Both variances are reduced by choosing high positive values for b and h. Optimal to choose highest possible values of both b and h, subject to it 0. No con‡ict between GDP and in‡ation stabilization. This is the same qualitative conclusion as with static expectations. ECON 3410/4410: Lecture 7 Optimal stabilization policy under RE III Supply shocks only ( 2 v = 0): From (21) and (22) we have: 2 y = 2 = ( 2 2 2 h) s [1 + 2 (b + h)]2 (1 + 2 b)2 2s [1 + 2 (b + h)]2 It is not possible to see, by direct inspection, how choices of h and b in‡uence these variances. Therfore we need the derivatives with respects to b and h. ECON 3410/4410: Lecture 7 Optimal stabilization policy under RE IV To calculate, it is practical to …rst take logs and both sides ln ln 2 y 2 2 2 2 ln [1 + 2 (b + h)] 2 h) s ] 2 2 ln[(1 + 2 b) s ] 2 ln [1 + 2 (b + = ln[( = and then …nd the derivatives. @ 2y @b @ 2y @h @ 2 @b @ 2 @h = = = = 2 2 [1 + 2 (b + h)] 2 y < 0 if b > 0 2(1 + 2 b) 2 h [1 + 2 (b + h)] y 2 22 h 2 (1 + 2 b) [1 + 2 (b + h)] 2 2 2 < 0 if b > 0 [1 + 2 (b + h)] ECON 3410/4410: Lecture 7 h)] Optimal stabilization policy under RE V @ 2 2 We see that the “direct derivatives” @by and @@h are negative subject to b > 0 (countercyclical interest rate setting). Remember that h > 0 from the Taylor principle @ 2y @h 2 But and @@b are both positive (given that b > 0), meaning that when supply shocks are dominant, there is a con‡ict of priorities between GDP stabilization and in‡ation stabilization. So in general, there is a trade-o¤, re‡ected by the choice of parameter in the social loss function. Again, these conclusions are qualitatively the same as in the case with static in‡ation expectations. ECON 3410/4410: Lecture 7 A modi…ed Taylor-rule (gradualism) I In the RE solution, all endogenous variables adjust instantaneously to a shock. This includes the nominal interest rate. IAM contains an important paragraph (“The optimality of a modi…ed Taylor rule under forward-looking expectations”) explaining why a Taylor rule that comes closer to empirical Taylor-rules may actually be optimal. In empirical studies we often estimate: it = r + h( t ) + h(yt y ) + cit 1; c >0 (23) Which corroborates that many central banks normally prefer to adjust the interest rate in “small but frequent steps”. ie, gradualism in interest rate setting. ECON 3410/4410: Lecture 7 A modi…ed Taylor-rule (gradualism) II The theoretical point is that, with c rather large, an adjustment of it will signal a higher interest rate in the future. The result is that a policy interest rate change is likely to a¤ect mortgage rates and other long term rates more e¤ectively than would be the case if c = 0. The policy instrument gets a stronger e¤ect in the monetary transmission mechanism. ECON 3410/4410: Lecture 7 The Lucas critique I The Lucas critique states that models with backward-looking expectations give wrong predictions about policy e¤ects. The example with a reduction in illustration. in period t serves as an Remember that the AD and AS equations with static expectations are: (yt y) = ( = t 1 t t + (yt ) zt , y ) + st 0 showing that if is reduced in period t, GDP is reduced. This due to increased real interest rate. Already, this is di¤erent from the RE solution, compare (17) and (20), which show no reduction in GDP when is reduced. ECON 3410/4410: Lecture 7 The Lucas critique II The dynamic e¤ects on GDP in the static model are intuitively clear: There must be a gradual increase in GDP, back to y . This is achieved by a gradual reduction in in‡ation, and a reduction of the real interest rate. It may be confusing that the …nal equation that we (and IAM) have derived for (yt y ) does not contain as a parameter, so the algebraic solution of the model does not seem to match the economic interpretation. The explanation is that in the derivation, we have assumed that is a constant parameter, and it therefore drops out from the solution. ECON 3410/4410: Lecture 7 The Lucas critique III To allow for a time varying in‡ation target, we write the model as (yt y) = ( = t 1 t t) t zt + (yt y ) + st and re-do the derivation of the …nal equation, which becomes (yt y ) = (yt 1 y ) + (zt zt 1) st + for GDP, and t = t 1 + t + z t + st for in‡ation. = 1 1+ , and (1 )= 1+ as before. ECON 3410/4410: Lecture 7 ( t t 1) The Lucas critique IV We now see that a permanent change in the in‡ation target has the same e¤ect in the backward-looking model as a permanent negative demand shock. The sequence of dynamic multipliers are therefore negative. but diminishing in magnitude. In‡ation is gradually reduced down to the new target, the long run-multiplier for in‡ation is long run = 1 =1 as we know it should be. The critique is that this gives a misleading analysis of the e¤ects of a lower in‡ation target, if expectations are in fact rational. As we have seen, in the RE solution, expectations adjust fully in period t if the change was announced in period t 1: ECON 3410/4410: Lecture 7 The Lucas critique V GDP is una¤ected, meaning that there is less welfare loss under RE. Today, the validity of Lucas critique is taken for granted in large parts of the economist profession. But the relevance of the critique hinges on the REH being a su¢ ciently good approximation to real world expectations. ECON 3410/4410: Lecture 7