Discussion of. An Estimated Two-Country DSGE Model for the Euro Area and the US Economy. by Gregory de Walque, Frank Smets and Raf Wouters

Transcription

1 Discussion of An Estimated Two-Country DSGE Model for the Euro Area and the US Economy by Gregory de Walque, Frank Smets and Raf Wouters Martin Ellison University of Warwick and CEPR

2 Summary of the contribution Construct a two-country DSGE model of the EA and US, and estimate it using Bayesian methods and 22 time series Add oil shocks (hoping to explain positive co-movements in the data correlation between US GDP and EA GDP is 0.4) Provide clear discussion of results using all standard techniques (impulse response functions, variance decompositions, historical decompositions)

3 First impression Wow!!! Gregory Frank Raf

4 Second impression Results are a little disappointing Estimated international spillovers very low (>95% of output fluctuations in EA and US are caused by domestic shocks) Oil shocks don t do very much Implied international co-movements almost zero Real exchange rate mostly driven by its own UIP shock Related to above? Estimate of elasticity of substitution between foreign and domestic goods is all over the place.

5 My comments Is the model correct? Is the modelling of oil correct? Is the estimation technically correct?

6 Is the model correct? Lots of modelling choices to make Producer currency pricing or local currency pricing? Homogenous pricing or pricing to market? Flexible prices or sticky prices? Complete assets markets or incomplete asset markets? Incomplete asset markets across countries or within countries Production structure: tradeables and non-tradeable goods? Coordinated or uncoordinated fiscal and monetary policies? etc, etc

7 Schematic view of real side of model ε a p, K ~ j, Lj; Oj ; CES Leontief M p j Intermediate goods y j Kimball aggregator Domestic goods D C, I, Final goods F X θ;o f D d ; M f Distribution services M d M d ; D f Intermediate goods II θ Leontief Leontief CES

8 Nominal side of model 8 Calvo parameters = new world record? Monetary policy rule: R t = Rt + ( ) rπ π t + terms in output gap Final good price inflation (affected by import prices) Determinacy issues (Bullard and Schalling 2005)

9 Is the modelling of oil correct? Oil is modelled as an AR() process Likely to hit problems for at least two reasons:. Hamilton (999) Clear evidence of nonlinearity oil price increase much more important than oil price decrease 2. Ellison-Scott (2006) High oil price in 970s due to OPEC, high oil price 2000 onwards due to scarcity very different effects Not enough substitutability too much Leontief

10 Oil consumption per unit output US EA Where are these trends going in the estimation?

11 Is the estimation technically correct? Yes, of course! But still have problems in estimating, the elasticity of substitution between foreign and domestic goods. Final goods are CES aggregate of foreign and domestic goods. µ µ θ = ) ( M D Using this and cost minimisation one derives factor demands and so on

12 But CES aggregator also has secondary role as aggregator of input prices to marginal cost MC θ = µ P D + ( µ ) PM Now take first-order log-linear approximation of this aggregator: MC θ = µ P D + ( µ ) P M In first-order approximation the elasticity of substitution drops out On pricing side it is as if we had simple Cobb-Douglas aggregator (or even Leontief!)

13 Does log-linearisation matter? Take second-order log-linear approximation: MC θ µ ( µ ) = µ P D + ( µ ) P M + P D + P M 2 2 Second-order effect? But Juan and Jesus have taught us that second-order approximation errors can have first-order effects on the likelihood Could this be killer app for their work?

14 Back to Tintin and our Belgian heroes Bruce and Martin