dsge models and central bank policy making: a critical review
TRANSCRIPT
DSGE Models and Central Bank Policy
Making: A Critical Review
Shiu-Sheng Chen
Department of Economics
National Taiwan University
12.16.2010
Shiu-Sheng Chen (NTU Econ) DSGE and Policy 12.16.2010 1 / 37
Motivations
On July 20, 2010, the U.S. House of
Representatives Committee on Science and
Technology held a hearing on “Building a Science of
Economics for the Real World.”
Shiu-Sheng Chen (NTU Econ) DSGE and Policy 12.16.2010 2 / 37
Motivations
The hearing explored◮ the failure of mainstream macroeconomic models to
foresee the recent financial and economic collapse◮ the promise and the limits of modern macroeconomic
theory in policy making.
The macroeconomic model called the Dynamic
Stochastic General Equilibrium model (DSGE) is
favored today, from academy to the world’s central
banks.
However, prominent economists disagree on the
extent to which applying this theoretical model to
policy decisions is practical.Shiu-Sheng Chen (NTU Econ) DSGE and Policy 12.16.2010 3 / 37
Motivations
The hearing explored◮ the failure of mainstream macroeconomic models to
foresee the recent financial and economic collapse◮ the promise and the limits of modern macroeconomic
theory in policy making.
The macroeconomic model called the Dynamic
Stochastic General Equilibrium model (DSGE) is
favored today, from academy to the world’s central
banks.
However, prominent economists disagree on the
extent to which applying this theoretical model to
policy decisions is practical.Shiu-Sheng Chen (NTU Econ) DSGE and Policy 12.16.2010 3 / 37
Motivations
The hearing explored◮ the failure of mainstream macroeconomic models to
foresee the recent financial and economic collapse◮ the promise and the limits of modern macroeconomic
theory in policy making.
The macroeconomic model called the Dynamic
Stochastic General Equilibrium model (DSGE) is
favored today, from academy to the world’s central
banks.
However, prominent economists disagree on the
extent to which applying this theoretical model to
policy decisions is practical.Shiu-Sheng Chen (NTU Econ) DSGE and Policy 12.16.2010 3 / 37
Agenda for Today’s Talk
Introduction to DSGE models.◮ To be fair, DGSE and similar macroeconomic models
were first conceived as theorists’ tools.◮ But why, then, are they being relied on as the platform
upon which so much practical policy advice is
formulated?
The critiques on the DSGE models.
The perspectives on the DSGE models.
Shiu-Sheng Chen (NTU Econ) DSGE and Policy 12.16.2010 4 / 37
Agenda for Today’s Talk
Introduction to DSGE models.◮ To be fair, DGSE and similar macroeconomic models
were first conceived as theorists’ tools.◮ But why, then, are they being relied on as the platform
upon which so much practical policy advice is
formulated?
The critiques on the DSGE models.
The perspectives on the DSGE models.
Shiu-Sheng Chen (NTU Econ) DSGE and Policy 12.16.2010 4 / 37
Agenda for Today’s Talk
Introduction to DSGE models.◮ To be fair, DGSE and similar macroeconomic models
were first conceived as theorists’ tools.◮ But why, then, are they being relied on as the platform
upon which so much practical policy advice is
formulated?
The critiques on the DSGE models.
The perspectives on the DSGE models.
Shiu-Sheng Chen (NTU Econ) DSGE and Policy 12.16.2010 4 / 37
Introduction to DSGE Models
What is DSGE Model?
Dynamic Stochastic General Equilibrium Model
Dynamic: studying how the economy evolves over
time.
Stochastic: the economy is affected by random
shocks.
General Equilibrium: It depicts the macro-economy
as the sum of individual choices and decisions made
by firms, households, the government, and the
central bank, according to their own preferences and
views about the future.
Shiu-Sheng Chen (NTU Econ) DSGE and Policy 12.16.2010 6 / 37
Introduction to DSGE Models
What is DSGE Model?
Dynamic Stochastic General Equilibrium Model
Dynamic: studying how the economy evolves over
time.
Stochastic: the economy is affected by random
shocks.
General Equilibrium: It depicts the macro-economy
as the sum of individual choices and decisions made
by firms, households, the government, and the
central bank, according to their own preferences and
views about the future.
Shiu-Sheng Chen (NTU Econ) DSGE and Policy 12.16.2010 6 / 37
Introduction to DSGE Models
Advantages of the DSGE Models
Provide a coherent framework of analysis. This
coherence is brought by restricting acceptable
behavior of agents to dynamic utility maximization
and rational expectations.
The dynamic mechanics are more transparent and
clear and (impulse response functions with
structural shocks).Overcome the Lucas Critique (Policy experimentsare more reliable)
◮ Lucas (1976) argues that the Keynesian
Macroeconometric models are useless for policy analysis
because they fail to take expectations seriously.
Shiu-Sheng Chen (NTU Econ) DSGE and Policy 12.16.2010 7 / 37
Introduction to DSGE Models
Advantages of the DSGE Models
Provide a coherent framework of analysis. This
coherence is brought by restricting acceptable
behavior of agents to dynamic utility maximization
and rational expectations.
The dynamic mechanics are more transparent and
clear and (impulse response functions with
structural shocks).Overcome the Lucas Critique (Policy experimentsare more reliable)
◮ Lucas (1976) argues that the Keynesian
Macroeconometric models are useless for policy analysis
because they fail to take expectations seriously.
Shiu-Sheng Chen (NTU Econ) DSGE and Policy 12.16.2010 7 / 37
Introduction to DSGE Models
Advantages of the DSGE Models
Provide a coherent framework of analysis. This
coherence is brought by restricting acceptable
behavior of agents to dynamic utility maximization
and rational expectations.
The dynamic mechanics are more transparent and
clear and (impulse response functions with
structural shocks).Overcome the Lucas Critique (Policy experimentsare more reliable)
◮ Lucas (1976) argues that the Keynesian
Macroeconometric models are useless for policy analysis
because they fail to take expectations seriously.
Shiu-Sheng Chen (NTU Econ) DSGE and Policy 12.16.2010 7 / 37
Introduction to DSGE Models
Historical Background
Indeed, the DSGE approach is not new at all.
What was DSGE Model?
The Real Business Cycle (RBC) models in 1980s.◮ Kydland and Prescott (1982)◮ Long and Plosser (1983)
However, why did the RBC models fail to grab
Central Bankers’ attention?
Well, the related question is: why are the Central
Bankers so fascinated with DSGE models nowadays?
Shiu-Sheng Chen (NTU Econ) DSGE and Policy 12.16.2010 8 / 37
Introduction to DSGE Models
Historical Background
Indeed, the DSGE approach is not new at all.
What was DSGE Model?
The Real Business Cycle (RBC) models in 1980s.◮ Kydland and Prescott (1982)◮ Long and Plosser (1983)
However, why did the RBC models fail to grab
Central Bankers’ attention?
Well, the related question is: why are the Central
Bankers so fascinated with DSGE models nowadays?
Shiu-Sheng Chen (NTU Econ) DSGE and Policy 12.16.2010 8 / 37
Introduction to DSGE Models
Historical Background
Indeed, the DSGE approach is not new at all.
What was DSGE Model?
The Real Business Cycle (RBC) models in 1980s.◮ Kydland and Prescott (1982)◮ Long and Plosser (1983)
However, why did the RBC models fail to grab
Central Bankers’ attention?
Well, the related question is: why are the Central
Bankers so fascinated with DSGE models nowadays?
Shiu-Sheng Chen (NTU Econ) DSGE and Policy 12.16.2010 8 / 37
Introduction to DSGE Models
Historical Background
Indeed, the DSGE approach is not new at all.
What was DSGE Model?
The Real Business Cycle (RBC) models in 1980s.◮ Kydland and Prescott (1982)◮ Long and Plosser (1983)
However, why did the RBC models fail to grab
Central Bankers’ attention?
Well, the related question is: why are the Central
Bankers so fascinated with DSGE models nowadays?
Shiu-Sheng Chen (NTU Econ) DSGE and Policy 12.16.2010 8 / 37
Introduction to DSGE Models
Historical Background
In the past, RBC models have the following features:
◮ Flexible prices◮ Focus on calibration.
⋆ Moments matching and policy experiments.
◮ However, no emphasis on estimation and forecasting.◮ The RBC models (old DSGE models) have difficulty to
match persistence in data with i.i.d. structural shocks
(weak internal propagation mechanism).◮ Most importantly, the RBC models emphasize real
shocks only.◮ Although some researchers try to incorporate money into
the RBC models, it is in general found that money is
neutral under flexible-price framework.
Shiu-Sheng Chen (NTU Econ) DSGE and Policy 12.16.2010 9 / 37
Introduction to DSGE Models
Historical Background
In the past, RBC models have the following features:
◮ Flexible prices◮ Focus on calibration.
⋆ Moments matching and policy experiments.
◮ However, no emphasis on estimation and forecasting.◮ The RBC models (old DSGE models) have difficulty to
match persistence in data with i.i.d. structural shocks
(weak internal propagation mechanism).◮ Most importantly, the RBC models emphasize real
shocks only.◮ Although some researchers try to incorporate money into
the RBC models, it is in general found that money is
neutral under flexible-price framework.
Shiu-Sheng Chen (NTU Econ) DSGE and Policy 12.16.2010 9 / 37
Introduction to DSGE Models
Historical Background
In the past, RBC models have the following features:
◮ Flexible prices◮ Focus on calibration.
⋆ Moments matching and policy experiments.
◮ However, no emphasis on estimation and forecasting.◮ The RBC models (old DSGE models) have difficulty to
match persistence in data with i.i.d. structural shocks
(weak internal propagation mechanism).◮ Most importantly, the RBC models emphasize real
shocks only.◮ Although some researchers try to incorporate money into
the RBC models, it is in general found that money is
neutral under flexible-price framework.
Shiu-Sheng Chen (NTU Econ) DSGE and Policy 12.16.2010 9 / 37
Introduction to DSGE Models
Historical Background
In the past, RBC models have the following features:
◮ Flexible prices◮ Focus on calibration.
⋆ Moments matching and policy experiments.
◮ However, no emphasis on estimation and forecasting.◮ The RBC models (old DSGE models) have difficulty to
match persistence in data with i.i.d. structural shocks
(weak internal propagation mechanism).◮ Most importantly, the RBC models emphasize real
shocks only.◮ Although some researchers try to incorporate money into
the RBC models, it is in general found that money is
neutral under flexible-price framework.
Shiu-Sheng Chen (NTU Econ) DSGE and Policy 12.16.2010 9 / 37
Introduction to DSGE Models
Historical Background
In the past, RBC models have the following features:
◮ Flexible prices◮ Focus on calibration.
⋆ Moments matching and policy experiments.
◮ However, no emphasis on estimation and forecasting.◮ The RBC models (old DSGE models) have difficulty to
match persistence in data with i.i.d. structural shocks
(weak internal propagation mechanism).◮ Most importantly, the RBC models emphasize real
shocks only.◮ Although some researchers try to incorporate money into
the RBC models, it is in general found that money is
neutral under flexible-price framework.
Shiu-Sheng Chen (NTU Econ) DSGE and Policy 12.16.2010 9 / 37
Introduction to DSGE Models
Historical Background
In the past, RBC models have the following features:
◮ Flexible prices◮ Focus on calibration.
⋆ Moments matching and policy experiments.
◮ However, no emphasis on estimation and forecasting.◮ The RBC models (old DSGE models) have difficulty to
match persistence in data with i.i.d. structural shocks
(weak internal propagation mechanism).◮ Most importantly, the RBC models emphasize real
shocks only.◮ Although some researchers try to incorporate money into
the RBC models, it is in general found that money is
neutral under flexible-price framework.
Shiu-Sheng Chen (NTU Econ) DSGE and Policy 12.16.2010 9 / 37
Introduction to DSGE Models
Historical Background
In the past, RBC models have the following features:
◮ Flexible prices◮ Focus on calibration.
⋆ Moments matching and policy experiments.
◮ However, no emphasis on estimation and forecasting.◮ The RBC models (old DSGE models) have difficulty to
match persistence in data with i.i.d. structural shocks
(weak internal propagation mechanism).◮ Most importantly, the RBC models emphasize real
shocks only.◮ Although some researchers try to incorporate money into
the RBC models, it is in general found that money is
neutral under flexible-price framework.
Shiu-Sheng Chen (NTU Econ) DSGE and Policy 12.16.2010 9 / 37
Introduction to DSGE Models
Historical Background
So how can we expect that Central Bankers would
enjoy a model that monetary policy plays no role?
Nowadays, the development of DSGE models
become a broad-ranging search to discover a set of
frictions to reproduce realistic dynamics.
In particular, the most popular frictions are the
Keynesian elements such as sticky prices/wages.
Frictions such as stick prices open up the possibility
of the non-neutrality of monetary policy.
No wonder the central bankers like the models now.
Shiu-Sheng Chen (NTU Econ) DSGE and Policy 12.16.2010 10 / 37
Introduction to DSGE Models
Historical Background
So how can we expect that Central Bankers would
enjoy a model that monetary policy plays no role?
Nowadays, the development of DSGE models
become a broad-ranging search to discover a set of
frictions to reproduce realistic dynamics.
In particular, the most popular frictions are the
Keynesian elements such as sticky prices/wages.
Frictions such as stick prices open up the possibility
of the non-neutrality of monetary policy.
No wonder the central bankers like the models now.
Shiu-Sheng Chen (NTU Econ) DSGE and Policy 12.16.2010 10 / 37
Introduction to DSGE Models
Historical Background
So how can we expect that Central Bankers would
enjoy a model that monetary policy plays no role?
Nowadays, the development of DSGE models
become a broad-ranging search to discover a set of
frictions to reproduce realistic dynamics.
In particular, the most popular frictions are the
Keynesian elements such as sticky prices/wages.
Frictions such as stick prices open up the possibility
of the non-neutrality of monetary policy.
No wonder the central bankers like the models now.
Shiu-Sheng Chen (NTU Econ) DSGE and Policy 12.16.2010 10 / 37
Introduction to DSGE Models
Historical Background
So how can we expect that Central Bankers would
enjoy a model that monetary policy plays no role?
Nowadays, the development of DSGE models
become a broad-ranging search to discover a set of
frictions to reproduce realistic dynamics.
In particular, the most popular frictions are the
Keynesian elements such as sticky prices/wages.
Frictions such as stick prices open up the possibility
of the non-neutrality of monetary policy.
No wonder the central bankers like the models now.
Shiu-Sheng Chen (NTU Econ) DSGE and Policy 12.16.2010 10 / 37
Introduction to DSGE Models
Historical Background
So how can we expect that Central Bankers would
enjoy a model that monetary policy plays no role?
Nowadays, the development of DSGE models
become a broad-ranging search to discover a set of
frictions to reproduce realistic dynamics.
In particular, the most popular frictions are the
Keynesian elements such as sticky prices/wages.
Frictions such as stick prices open up the possibility
of the non-neutrality of monetary policy.
No wonder the central bankers like the models now.
Shiu-Sheng Chen (NTU Econ) DSGE and Policy 12.16.2010 10 / 37
Introduction to DSGE Models
Historical Background
It is thus called a New Keynesian DSGE model, and
I will call it a “New DSGE” Model, in short.
Moreover, new DSGE models start to focus on
structural estimations (GMM, MLE, Bayesian).
So it is also called an Estimated DSGE Model,
which is able to provide forecasts on a regular basis.
Shiu-Sheng Chen (NTU Econ) DSGE and Policy 12.16.2010 11 / 37
Introduction to DSGE Models
Historical Background
It is thus called a New Keynesian DSGE model, and
I will call it a “New DSGE” Model, in short.
Moreover, new DSGE models start to focus on
structural estimations (GMM, MLE, Bayesian).
So it is also called an Estimated DSGE Model,
which is able to provide forecasts on a regular basis.
Shiu-Sheng Chen (NTU Econ) DSGE and Policy 12.16.2010 11 / 37
Introduction to DSGE Models
Historical Background
It is thus called a New Keynesian DSGE model, and
I will call it a “New DSGE” Model, in short.
Moreover, new DSGE models start to focus on
structural estimations (GMM, MLE, Bayesian).
So it is also called an Estimated DSGE Model,
which is able to provide forecasts on a regular basis.
Shiu-Sheng Chen (NTU Econ) DSGE and Policy 12.16.2010 11 / 37
Introduction to DSGE Models
Macroeconometric Models, SVAR and DSGE Models
Macroeconometric SVAR Models DSGE Models
Models
Dynamic Yes Yes Yes
Stochastic Regression Residuals Structural Shocks Structural Shocks
General Loosely based on Based on Explicitly based on
Equilibrium theory in ad hoc economic structure optimization in
manner and theory coherent manner
Invalid Exogeneity Yes No No
Assumption
Policy Experiment Yes Yes Yes
(Reliable?) (No) (Yes) (Yes)
Forecast Yes Yes Yes
Shiu-Sheng Chen (NTU Econ) DSGE and Policy 12.16.2010 12 / 37
Introduction to DSGE Models
DSGE and Central Banks
According to the overall victory for DSGE models in
the above Table, no wonder why DSGE models are
becoming increasingly popular in central banking
circles over the past decade.
European Central Bank US Federal Reserve Bank
Sveriges Riksbank Bank of Canada
Bank of England Norges Bank
Bank of Finland Reserve Bank of New Zealand
Bank of Spain Central Bank of Brazil
Central Bank of Chile Central Reserve Bank of Peru
Bank of Thailand IMF
Shiu-Sheng Chen (NTU Econ) DSGE and Policy 12.16.2010 13 / 37
Introduction to DSGE Models
DSGE and Central Banks
However, before the central bankers proceed to
move to the new framework, some warnings about
the flaws and limitations of DSGE models should be
heeded.
Shiu-Sheng Chen (NTU Econ) DSGE and Policy 12.16.2010 14 / 37
Introduction to DSGE Models
The State-of-the-Art DSGE Models
Smets and Wouters (2007) model◮ a large set of persistent exogenous shocks:
⋆ total factor productivity
⋆ investment-specific technology
⋆ risk premium
⋆ government spending
⋆ price mark-up
⋆ wage mark-up
⋆ monetary policy
◮ a large set of frictions:⋆ sticky wages and prices
⋆ sticky adjustment of capacity utilization
⋆ investment adjustment cost
⋆ habit formation in consumption
◮ a Taylor-rule-type monetary policy
Shiu-Sheng Chen (NTU Econ) DSGE and Policy 12.16.2010 15 / 37
Introduction to DSGE Models
The State-of-the-Art DSGE Models
The model is able to track and forecast 7macroeconomic time series data very well.
◮ GDP◮ consumption◮ investment◮ prices◮ wages◮ employment◮ short-run interest rates
Shiu-Sheng Chen (NTU Econ) DSGE and Policy 12.16.2010 16 / 37
Introduction to DSGE Models
The State-of-the-Art DSGE Models
I focus on this popular model because it is widely
considered to be the state-of-the-art DSGE model.
A version of the Smets-Wouters model is now being
used to inform policymaking at the European
Central Bank.
Hence, new DSGE models that follow approximately
this recipe are being formulated and coming into use
at central banks around the world.
Shiu-Sheng Chen (NTU Econ) DSGE and Policy 12.16.2010 17 / 37
Introduction to DSGE Models
The State-of-the-Art DSGE Models
I focus on this popular model because it is widely
considered to be the state-of-the-art DSGE model.
A version of the Smets-Wouters model is now being
used to inform policymaking at the European
Central Bank.
Hence, new DSGE models that follow approximately
this recipe are being formulated and coming into use
at central banks around the world.
Shiu-Sheng Chen (NTU Econ) DSGE and Policy 12.16.2010 17 / 37
Introduction to DSGE Models
The State-of-the-Art DSGE Models
I focus on this popular model because it is widely
considered to be the state-of-the-art DSGE model.
A version of the Smets-Wouters model is now being
used to inform policymaking at the European
Central Bank.
Hence, new DSGE models that follow approximately
this recipe are being formulated and coming into use
at central banks around the world.
Shiu-Sheng Chen (NTU Econ) DSGE and Policy 12.16.2010 17 / 37
Critiques
Critiques on New DSGE Models
The critiques I am going to offer are not meant to
humiliate the new DSGE models; however, I think
that the policymakers should ask hard questions
about the value of the model before adopting the
new approach.
The critiques:
◮ Lucas Critique strikes back◮ The shocks are dubiously structural◮ Empirical performance of the model
Shiu-Sheng Chen (NTU Econ) DSGE and Policy 12.16.2010 19 / 37
Critiques
Critiques on New DSGE Models
The critiques I am going to offer are not meant to
humiliate the new DSGE models; however, I think
that the policymakers should ask hard questions
about the value of the model before adopting the
new approach.
The critiques:
◮ Lucas Critique strikes back◮ The shocks are dubiously structural◮ Empirical performance of the model
Shiu-Sheng Chen (NTU Econ) DSGE and Policy 12.16.2010 19 / 37
Critiques
Lucas Critique Strikes Back
Indeed, the new DSGE models do not have solid
micro-foundation.
That is, some of the parameters are not “deep”
enough to be immune to the Lucas Critique.
The deep parameters are those that do not vary
with policy regime changes.
Shiu-Sheng Chen (NTU Econ) DSGE and Policy 12.16.2010 20 / 37
Critiques
The Lucas Critique
Lots of models in economics can be re-written as
yt = xt + bEtyt+1, |b| < 1.
By imposing no bubble condition, the solution is
yt =∞∑
j=0
bjEtxt+j
Suppose that xt follows a stationary AR(1) process
xt = φxt−1 + εt, |φ| < 1, εt ∼i .i .d . (0, σ2)
Shiu-Sheng Chen (NTU Econ) DSGE and Policy 12.16.2010 21 / 37
Critiques
The Lucas Critique
Hence,
Etxt+j = φjxt ,
and the model solution is
yt =1
1− bφxt .
It is called a reduced-form solution.
Shiu-Sheng Chen (NTU Econ) DSGE and Policy 12.16.2010 22 / 37
Critiques
The Lucas Critique
The structural model is
yt =
∞∑
j=0
bjEtxt+j
The reduced-form representation
yt =1
1− bφxt
depends on the process for xt with a particular form.
Shiu-Sheng Chen (NTU Econ) DSGE and Policy 12.16.2010 23 / 37
Critiques
The Lucas Critique
Hence, if the xt process change, the reduced-form
process will change. For instance, if the precess
changes to xt = ωxt−1 + εt , the reduced-form
representation changes as well:
yt =1
1− bωxt
Robert Lucas pointed out that the reduced-form
econometric model based on historical data
yt = βxt
is useless for policy analysis.Shiu-Sheng Chen (NTU Econ) DSGE and Policy 12.16.2010 24 / 37
Critiques
The Lucas Critique
Hence, if the xt process change, the reduced-form
process will change. For instance, if the precess
changes to xt = ωxt−1 + εt , the reduced-form
representation changes as well:
yt =1
1− bωxt
Robert Lucas pointed out that the reduced-form
econometric model based on historical data
yt = βxt
is useless for policy analysis.Shiu-Sheng Chen (NTU Econ) DSGE and Policy 12.16.2010 24 / 37
Critiques
Lucas Critique Strikes Back
Well, what’s wrong with the new DSGE models?
The new DSGE models generally exogenously
impose that firms can only change prices at certain
exogenously chosen points in time–those points may
be stochastic (Calvo pricing) or deterministic
(Taylor contracts).
These frictions are ad hoc, i.e., no micro-foundation
for price stickness is provided.
The deep parameters that govern the firms’ optimal
choice to fix the price have not been investigated.
Shiu-Sheng Chen (NTU Econ) DSGE and Policy 12.16.2010 25 / 37
Critiques
Lucas Critique Strikes Back
Well, what’s wrong with the new DSGE models?
The new DSGE models generally exogenously
impose that firms can only change prices at certain
exogenously chosen points in time–those points may
be stochastic (Calvo pricing) or deterministic
(Taylor contracts).
These frictions are ad hoc, i.e., no micro-foundation
for price stickness is provided.
The deep parameters that govern the firms’ optimal
choice to fix the price have not been investigated.
Shiu-Sheng Chen (NTU Econ) DSGE and Policy 12.16.2010 25 / 37
Critiques
Lucas Critique Strikes Back
Well, what’s wrong with the new DSGE models?
The new DSGE models generally exogenously
impose that firms can only change prices at certain
exogenously chosen points in time–those points may
be stochastic (Calvo pricing) or deterministic
(Taylor contracts).
These frictions are ad hoc, i.e., no micro-foundation
for price stickness is provided.
The deep parameters that govern the firms’ optimal
choice to fix the price have not been investigated.
Shiu-Sheng Chen (NTU Econ) DSGE and Policy 12.16.2010 25 / 37
Critiques
Lucas Critique Strikes Back
Well, what’s wrong with the new DSGE models?
The new DSGE models generally exogenously
impose that firms can only change prices at certain
exogenously chosen points in time–those points may
be stochastic (Calvo pricing) or deterministic
(Taylor contracts).
These frictions are ad hoc, i.e., no micro-foundation
for price stickness is provided.
The deep parameters that govern the firms’ optimal
choice to fix the price have not been investigated.
Shiu-Sheng Chen (NTU Econ) DSGE and Policy 12.16.2010 25 / 37
Critiques
The Shocks are Dubiously Structural
A structural shock must have two properties◮ The shocks must be uncorrelated with policy
interventions.◮ The shocks should have an appropriate economic
interpretation.
Most of the shocks in the new DSGE models:◮ wage markup shocks◮ price markup shocks◮ government spending shocks◮ risk premium shocks
fail to meet the above requirements, and are
dubiously structural.
Shiu-Sheng Chen (NTU Econ) DSGE and Policy 12.16.2010 26 / 37
Critiques
The Shocks are Dubiously Structural
A structural shock must have two properties◮ The shocks must be uncorrelated with policy
interventions.◮ The shocks should have an appropriate economic
interpretation.
Most of the shocks in the new DSGE models:◮ wage markup shocks◮ price markup shocks◮ government spending shocks◮ risk premium shocks
fail to meet the above requirements, and are
dubiously structural.
Shiu-Sheng Chen (NTU Econ) DSGE and Policy 12.16.2010 26 / 37
Critiques
The Shocks are Dubiously Structural
In particular, wage markup shocks are frequently
criticized on the basis that they have different and
somewhat conflicted interpretations/policy
implications.
Wage markup shocks◮ Bargaining power of the labor union
(Allocation is inefficient, the optimal policy is to limit
the monopoly power)◮ Worker’s leisure preference
(Allocation is efficient, laissez-faire is optimal)
Shiu-Sheng Chen (NTU Econ) DSGE and Policy 12.16.2010 27 / 37
Critiques
The Shocks are Dubiously Structural
Moreover, price markup shocks are not orthogonal
to other structural shocks.
Price markup shocks
◮ Change in tax rate◮ Degree of competition
Clearly, these interpretations are affected by
economic situations, which are certainly moved by
monetary shocks.
Shiu-Sheng Chen (NTU Econ) DSGE and Policy 12.16.2010 28 / 37
Critiques
Empirical Evaluation of the Model
The good performance (such as matching the
AR(1) dynamics of several macroeconomic
variables) may be due to the assumption of serially
correlated shocks, which is in question.
Morley (2010) describes that it is just like the “the
rabbit and the hat trick”: a rabbit was stuffed into
the hat and then a rabbit jumped out of the hat.
“What you put in is what you get out”!
Shiu-Sheng Chen (NTU Econ) DSGE and Policy 12.16.2010 29 / 37
Critiques
Empirical Evaluation of the Model
The good performance (such as matching the
AR(1) dynamics of several macroeconomic
variables) may be due to the assumption of serially
correlated shocks, which is in question.
Morley (2010) describes that it is just like the “the
rabbit and the hat trick”: a rabbit was stuffed into
the hat and then a rabbit jumped out of the hat.
“What you put in is what you get out”!
Shiu-Sheng Chen (NTU Econ) DSGE and Policy 12.16.2010 29 / 37
Critiques
Empirical Evaluation of the Model
The new DSGE model (i.e., Smets-Wouters
medium-scale model) are fit to much smaller set of
variables than that of the older macroeconometric
models.
The forecasting performance should be compared
with older macroeconometric models, not the
VAR/BVAR models.
Variables are linearly detrended. Fail to account for
possible unit root and cointegration.
Shiu-Sheng Chen (NTU Econ) DSGE and Policy 12.16.2010 30 / 37
Critiques
Empirical Evaluation of the Model
The new DSGE model (i.e., Smets-Wouters
medium-scale model) are fit to much smaller set of
variables than that of the older macroeconometric
models.
The forecasting performance should be compared
with older macroeconometric models, not the
VAR/BVAR models.
Variables are linearly detrended. Fail to account for
possible unit root and cointegration.
Shiu-Sheng Chen (NTU Econ) DSGE and Policy 12.16.2010 30 / 37
Critiques
Empirical Evaluation of the Model
The new DSGE model (i.e., Smets-Wouters
medium-scale model) are fit to much smaller set of
variables than that of the older macroeconometric
models.
The forecasting performance should be compared
with older macroeconometric models, not the
VAR/BVAR models.
Variables are linearly detrended. Fail to account for
possible unit root and cointegration.
Shiu-Sheng Chen (NTU Econ) DSGE and Policy 12.16.2010 30 / 37
Critiques
Empirical Evaluation of the Model
No formal tests such as the Diebold and Mariano
(1995) test are conducted for out-of-sample
forecasts.
Only a small set of variables is chosen to be
observable variables (7 of 14). Guerron-Quintana
(2010) find that the empirical results are not robust
to the selection of observable variables.
Using real time data from the US, Edge and
Gurkaynak (2010) find that the Smets-Wouters
medium-scale DSGE model forecasts inflation and
GDP growth poorly.Shiu-Sheng Chen (NTU Econ) DSGE and Policy 12.16.2010 31 / 37
Critiques
Empirical Evaluation of the Model
No formal tests such as the Diebold and Mariano
(1995) test are conducted for out-of-sample
forecasts.
Only a small set of variables is chosen to be
observable variables (7 of 14). Guerron-Quintana
(2010) find that the empirical results are not robust
to the selection of observable variables.
Using real time data from the US, Edge and
Gurkaynak (2010) find that the Smets-Wouters
medium-scale DSGE model forecasts inflation and
GDP growth poorly.Shiu-Sheng Chen (NTU Econ) DSGE and Policy 12.16.2010 31 / 37
Critiques
Empirical Evaluation of the Model
No formal tests such as the Diebold and Mariano
(1995) test are conducted for out-of-sample
forecasts.
Only a small set of variables is chosen to be
observable variables (7 of 14). Guerron-Quintana
(2010) find that the empirical results are not robust
to the selection of observable variables.
Using real time data from the US, Edge and
Gurkaynak (2010) find that the Smets-Wouters
medium-scale DSGE model forecasts inflation and
GDP growth poorly.Shiu-Sheng Chen (NTU Econ) DSGE and Policy 12.16.2010 31 / 37
Part III
Perspectives on New DSGE Models
Shiu-Sheng Chen (NTU Econ) DSGE and Policy 12.16.2010 32 / 37
Perspectives
On The Perspectives of DSGE Modeling and Policy Making
I will discuss perspectives from which the central
bankers can have some ideas about what they
should bear in mind when deciding to construct
their own DSGE models.
◮ The new DSGE models are not yet useful for policy
analysis.◮ Inclusion of a more sophisticated financial intermediation
sector.◮ Better forecasting.◮ More sensible shocks.
Shiu-Sheng Chen (NTU Econ) DSGE and Policy 12.16.2010 33 / 37
Perspectives
On The Perspectives of DSGE Modeling and Policy Making
First, as criticized above, the new DSGE models are not
yet useful for policy analysis.
So do not be frightened and confused by these
tedious and fancy mathematical equations.The current new DSGE model is full of dubiousstructures and features, which may misleadpolicymaking.
◮ Chari, Kehoe, and McGrattan (2009)◮ Faust (2008, 2009)
Keep in mind that the success in capturing business
cycle dynamics may be simply “made up” by an
implausible framework.Shiu-Sheng Chen (NTU Econ) DSGE and Policy 12.16.2010 34 / 37
Perspectives
On The Perspectives of DSGE Modeling and Policy Making
First, as criticized above, the new DSGE models are not
yet useful for policy analysis.
So do not be frightened and confused by these
tedious and fancy mathematical equations.The current new DSGE model is full of dubiousstructures and features, which may misleadpolicymaking.
◮ Chari, Kehoe, and McGrattan (2009)◮ Faust (2008, 2009)
Keep in mind that the success in capturing business
cycle dynamics may be simply “made up” by an
implausible framework.Shiu-Sheng Chen (NTU Econ) DSGE and Policy 12.16.2010 34 / 37
Perspectives
On The Perspectives of DSGE Modeling and Policy Making
First, as criticized above, the new DSGE models are not
yet useful for policy analysis.
So do not be frightened and confused by these
tedious and fancy mathematical equations.The current new DSGE model is full of dubiousstructures and features, which may misleadpolicymaking.
◮ Chari, Kehoe, and McGrattan (2009)◮ Faust (2008, 2009)
Keep in mind that the success in capturing business
cycle dynamics may be simply “made up” by an
implausible framework.Shiu-Sheng Chen (NTU Econ) DSGE and Policy 12.16.2010 34 / 37
Perspectives
On The Perspectives of DSGE Modeling and Policy Making
Second, consider the inclusion of a more sophisticated
financial markets. For example,
Financial vulnerabilities and crisis
Collateral constraints and real estate markets
Choice of exchange rate regime
Shiu-Sheng Chen (NTU Econ) DSGE and Policy 12.16.2010 35 / 37
Perspectives
On The Perspectives of DSGE Modeling and Policy Making
Next, seek for better forecasting of macroeconomic
variables.
The DSGE model may be useful as a mechanism for
generating a prior which aids the estimation of a
BVAR: DSGE-VAR, see Del Negro et al. (2004).
The DSGE models need to forecast macroeconomic
variables out-of-sample well in level (rather than the
deviations from steady states) and in real time.
Shiu-Sheng Chen (NTU Econ) DSGE and Policy 12.16.2010 36 / 37
Perspectives
On The Perspectives of DSGE Modeling and Policy Making
Next, seek for better forecasting of macroeconomic
variables.
The DSGE model may be useful as a mechanism for
generating a prior which aids the estimation of a
BVAR: DSGE-VAR, see Del Negro et al. (2004).
The DSGE models need to forecast macroeconomic
variables out-of-sample well in level (rather than the
deviations from steady states) and in real time.
Shiu-Sheng Chen (NTU Econ) DSGE and Policy 12.16.2010 36 / 37