ECON 3141/4141: International macro and finance. Term paper, spring 2005. Annotated version Ragnar Nymoen 1. See individual comments to your answers. 2. In chapter 3 of Introductory Dynamic Macroeconomics, you find the following expression, as well as an explanation of symbols and of variables:1 Ut = 0.005 + 0.005 · S1989t + 0.8Ut−1 + εU,t Show that according to this equation, the steady-state rate of unemployment is 0.025 before 1989, and 0.05 after 1989. ANSWER SUGGESTIONS: (note: originally there was a typo as the coefficient of S1989t was 0.005, both were corrected though). The equation is an ADL model of the rate of unemployment, Ut . S1989t is the only exogenous variable (corresponding to ”xt ” in the lectures). Since 0.8 < 1 the model has a stable solution for any given initial value of U, and the actual/chosen values of S1989t and εU,t . (In the case of S1989t the values are known form the specification of the dummy, while the εU,t are set to zero by us). From IDM we know the expression for the stable steady-state solution of Ut , which we can denote U ∗ . 0.005 = 0.025 if S1989t = 0 0.2 0.01 = 0.05 if S1989t = 1 = 0.2 U∗ = U∗ This means that, if we for example take 1985 as the initial period, and assume that U1985 = 0.025 then the solution will be U1986 = U1987 = U1988 = 0.025 and U1989 = 0.03, U1900 = 0.034, and so on: a gradual rise of actual unemployment to the new equilibrium of 5%. 3. Assume that, in a closed economy, the rate of unemployment in year t, denoted Ut is given by: (1) Ut = γ 0 + γ 1 (it − π t ) + γ 2 (it−1 − π t−1 ) + γ 3 Ut−1 , γ 1 + γ 2 > 0, 0 ≤ γ 3 < 1 where it denotes a nominal interest rate and πt denotes the rate of inflation. Hence it − π t denotes the real interest rate in year t. In order to simplify the notation, the random disturbance term is omitted from (1)–and also from (2) and (3) below. 1 Note that the typo in IDM has been corrected (i.e., Ut−1 ). 1 (a) Explain intuitively (on the basis of your earlier studies for example) the economic rationale for the distributed lag of the real interest rate in equation (1). ANSWER SUGGESTIONS: Some points that might be mentioned: U is a variable which is affected both by factors that influence labour demand and labour supply. It is most reasonable that the effect of real interest rate is via demand, which in turn depends on (the demand of) GDP output: A rise in the real interest rate affects both private consumption (both a substitution effect and an effect though disposable income) and private investments in real capital. Moreover, adjustment lags both in consumption and investment decisions make the distributed lag specification in (1) reasonable. Due to such decisions lags, and if we think of the time period as annual for example, then it might be realistic to assume that γ 2 > γ 1 , i.e., the second year effect of a rise in the interest rate is larger than the first year (impact) effect. Note 1: You can of course also establish the link between U and GDP demand by invoking Okun’s law, as we do in the lectures. Note 2: You are not expected to comment on the autoregressive part of (2), but a reasonable remark is that both (further) adjustment lags in labour demand and separate lagged adjustments of labour supply when labour market conditions change, motivate the presence of Ut−1 in the equation, with a positive coefficient. (b) Assume that both the nominal interest rate and the inflation rate are exogenous variables (this assumption also applies to c) and d) below). What is the expression for the long-run multiplier of the rate of unemployment with respect to the real interest rate? ANSWER: Since this is an ADL model, we have from the general properties of ADLs that the long-run multiplier in this case is given by ∂U ∗ γ + γ2 . = 1 ∂(i − π) 1 − γ3 (Note that the parameter restriction γ 3 < 1 is important, as γ 3 = 1 corrupts the multiplier.) (c) Draw a typical graph of the dynamic multipliers of Ut with respect to a permanent change in the real interest rate. Explain. ANSWER: If both γ 1 and γ 2 are positive: a monotonously increasing multiplier (of course maintaining 0 ≤ γ 3 < 1 as given in (1)). Many of you draw a graph that approaches 1 asymptotically. Thus the long run multiplier is restricted to 1. However, take care to note that there is nothing in the question (or altogether in the concept of the long run multiplier) that restricts its value to 1 (In many applications, such a restriction might however be theoretically appealing. For example: the consumption function or a PPP based model of the domestic price level. Why?). 2 (d) Calibrate the parameters of (1) so that the steady-state rate of unemployment is 0.03 for a constant real interest rate of 0.01. For a constant real interest rate of 0.06, what is the corresponding steady-state rate of unemployment? ANSWER: Again, since this is an ADL model, the expression for the steady-state rate of unemployment is U∗ = γo γ + γ2 + 1 0.01 1 − γ3 1 − γ3 Calibration means “choosing parameter values so that the model matches some criterion” (theoretical or empirical: in this case the criterion is a long run value of Ut equal to 0.03). Several constellations of the three parameters are consistent with U ∗ = 0.03. For example, and for simplicity set γ 0 = 0 and set γ 3 = 0.8. Then the calibrated value of γ 1 + γ 2 becomes 0.6. (Note that we don’t need do identify values of γ 1 and γ 2 separately, only of the sum). For the same parameter values, and if the constant real interest rate is 0.06, we get U ∗ = 0.18, so a high real interest rate has a huge impact on long run unemployment in this calibrated model. 4. Simplify equation (1) by setting γ 1 = 0 and γ 3 = 0, to obtain (2) Ut = γ 0 + γ 2 (it−1 − πt−1 ). Assume next that the nominal interest rate is exogenous, but that the rate of inflation is endogenized by the Phillips curve: (3) π t = β 0 + β 1 Ut + β 2 πt−1 , β 1 < 0, 0 ≤ β 2 ≤ 1. (a) Assume that β 1 = −0.5. Calibrate the other parameters of the system made up of equation (2) and (3) in such a way that the equilibrium rate of unemployment (identical to the natural rate in this case) is equal to 0.06. Also give an expression which defines the corresponding equilibrium rate of inflation. ANSWER SUGGESTION: Since this is a Phillips curve system, the steady state rate of unemployment is given by −β 0 + (1 − β 2 )π ∗ U∗ = β1 However, unless β 2 = 1, U ∗ depends on the steady-state rate of inflation (corresponding to a downward sloping rather than a vertical long-run Phillips curve). Hence, the easiest way to answer this question is to set β 2 = 1, which is admissible since equation (3) specifies 0 ≤ β 2 ≤ 1. In the case of β 2 = 1, for example β 0 = 0.03 and β 1 = −0.5 gives U ∗ = 0.06. 3 There is an important “tactical lesson” here: (nearly) all of you attempt to answer the question for the more general case of 0 ≤ β 2 < 1 which is of course very commendable, but also much more difficult! In the school exam, an “A” would typically (I expect) be based on the explicit assumption of β 2 = 1! However, returning to the case of 0 ≤ β 2 ≤ 1,we have the following long run system, defining the equilibrium values π ∗ and U ∗ in a situation where it = it−1 = i: U ∗ + γ 2 π∗ = γ 0 + γ 2 i −β 1 U ∗ + (1 − β 1 )π∗ = β 0 Solving this long run model for U ∗ gives: U∗ = (γ 0 + γ 2 i)(1 − β 2 ) − β 0 γ 2 (1 − β 2 ) + γ 2 β 1 Unless β 2 = 1, U ∗ is seen to depend on i, the nominal interest rate, so a relevant critique of the question (always a good thing to include in an exam answer (if you are right that is!!)) is that it is wrong to say that U ∗ is “identical to the natural rate in this case”. Presumably, a natural rate should not depend on nominal variables (see for example your essays on the natural rate!). Having said that, one calibration may be β 2 = 0.95, β 1 = −0.5, β 0 = 0.03, γ 2 = 0.6, γ 0 = 0.01 and i = 0.09: (0.01 + 0.6 × 0.09) × (1 − 0.95) − 0.03 × 0.6 ≈ 0.06. (1 − 0.95) + 0.6 × (−0.5) (b) Explain what we mean by a solution to the system defined by (2) and (3). ANSWER: See for example chapter 1.9 of IDM: Assume all parameters are known. Assume also that π 0 is known and that we have a known series of values i0 , i1 , i2 , ...iT . Then (2) and (3) delivers two unique times series U1 , U2 , ...UT and π1 , π 2 , ...π T which represent the solution. It is found recursively: First U1 and π 1 , then U2 and π 2 and so on. (c) Show that a sufficient condition for a stable solution of the system is: β 2 − γ 2 β 1 < 1. ANSWER: Substitute the expression for Ut in (3) and apply the condition for a stable solution of an ADL. See for example chapter 1.9 in IDM. Specifically, the so called final equation for π t is: π t = β 0 + β 1 γ 0 + β 1 γ 2 it−1 + (β 2 − γ 2 β 1 )πt−1 4 (d) Is a vertical long-run Phillips curve consistent with a stable solution of the system? Explain. ANSWER: In this model a vertical Phillips curve is synonymous to setting β 2 = 1. If γ 2 ≥ 0 (as in 3 d), the system is then dynamically either unstable or explosive: From known initial conditions, the solution does not converge to U ∗ , unless U = U ∗ to start with. Intuition: assume that, for some reason or another, that we get a higher initial value (π 00 ) than the value π 0 which is consistent with U1 = U ∗ : then π 1 increases with the same amount from the lagged term alone. In addition, U1 is reduced (below U ∗ ) which increases π 1 further, this leads to instability. Most believers in the Phillips curve would not like this implication, since a vertical long run Phillips curve is seen as theoretically sound it will be seen as problematic (an internal inconsistency almost) that the model is unstable in the case of β 2 = 1. When we reconsider equation (2) is is clear that the “problem” is that the rate of inflation only enters through the real interest rate (in fact the AD schedule in this case has the “wrong slope” compared to the text-book case). Hence one would have to propose other effects of inflation on Ut than the effect through the real interest rate, and include them in the model (so that the slope of the AS schedule is turned around). In the closed economy case this may not be easy, although so called effects of real balances on aggregated demand (real money in the consumption function for example) represent one possibility. In the case of open economy models, the real exchange rate offers a way out of this puzzle, which in fact is what we have in the AD-AS model in B&W. (e) Derive the dynamic multipliers of inflation with respect to the nominal interest rate as fully as you can–or at least: the impact and long-run multipliers. Set β 1 = −0.5, γ 2 = 0.1, β 2 = 0.75. ANSWER NOTE: The first multiplier (impact) is 0. The second multiplier is δ1 = β 1 γ 2 which becomes −0.05. The third multiplier: δ 2 = β 1 γ 2 × 1 + (β 2 − γ 2 β 1 )δ 1 = −0.09. The long-run multiplier: β1γ2 = −0.25. 1 − β2 + γ2β1 5