Transcript Slide 1
Building and Utilizing Macroeconomic Modeling
for Policy Purposes: An Overview
by
Atchana Waiquamdee
Deputy Governor
Bank of Thailand
Presented at the SEACEN-CCBS/BOE-BSP Workshop on
Dynamic Stochastic General Equilibrium Modeling and Econometric Techniques
November 23–27, 2009
Manila, Philippines
Outline of Presentation
1.
2.
3.
4.
5.
Introduction
Macro models and DSGE model
Building DSGE model
Challenges in macro modeling
Conclusion
2
1. Introduction
3
4
The Long History of Economic Models
Analogies – clear your thinking
– Adam Smith’s pin factory
Early analytical tools – simplify the issues
– Alfred Marshall’s demand/supply, Edgeworth’s box,
Cournot’s model of competition, Hicks’ IS-LM analysis
Formal use of mathematics – retain the coherence
– Micro: Game theory, General equilibrium
– Macro: Dynamic models with optimizing and rational agents
– Empirical: Econometrics, Time-series
Two macro revolutions
(1) Lucas critique
Role of monetary policy: anchoring inflation expectations
Expectations in previous macro models
– Purely adaptive: based on what happened in the past
Rational expectations Agents do not make systematic errors
when predicting the future
– Model parameters in fact depend on agents’ expectations of policy
– Would be naive to evaluate policies based on past aggregate data
– To predict effects of policy experiment, we must model
deep parameters that govern individual behavior
5
Two macro revolutions
(2) New Keynesian macroeconomics
Nevertheless, even in model with rational agents, role of
monetary policy is limited
– In reality, we do observe impacts of monetary policy in short run
Hence, emerging role of market imperfections
– Imperfect competition
– Nominal frictions: price and wage rigidities
Rational expectations and New Keynesian framework
– Give rise to crucial role of macroeconomic stabilization policies
– Form the basis for modern macroeconomic models
6
2. An overview of macro models
7
8
Generic uses of macro models
Examination of dynamics of macro variables
– Output and its components, employment, prices
Macro models:
–
–
–
–
Articulate and expound theoretical ideas
Test different theories
Construct “what if” scenarios
Produce forecasts
9
Using macro models in policymaking process
Policymaker must base decisions on limited understanding of
the economy, a large and complex system
Models aim to simplify complex system but keep all salient features
– Deep parameters capturing preferences, technology, and institution factors
Good models, for policy making, should…
– Be backed by a coherent and logical theory
– Replicate the statistical properties of the actual time series
Both criteria are necessary
– Lack of a coherent theory leads to ‘Lucas critique’, spurious relationships,
data mining, over-fitting, and model misspecification for example
– Lack of statistical congruence leads to poor forecasting
Portfolio of macro models in most central banks
Driven by theory
Different models have different
comparative advantage in
what is required by policymakers
DSGE model
■
Small New-Keynesian
■ model
■
Macroeconometric model
■
• Those mainly driven by data:
forecasting
• Those mainly driven by theory:
policy analysis
VAR
Driven by data
Adapted from Pagan (2003), “Report on Modeling and Forecasting at the Bank of England,” Bank of England
Quarterly Bulletin, (Spring), 60-88
10
VAR
Use econometric procedures to obtain model parameters
Emphasize fitting historical data
Rely very little on theory
11
Macroeconometric Model
Fit data to equations that are loosely based on theory
Typically use error correction mechanism, specifying
– Long-run relationship: based on theory
– Short-run relationship: guided by factors driving short-run dynamics
While linked to theory, long-run values
– Do not depend on supply side
– May not satisfy stock-flow consistency
Furthermore, model does not emphasize microeconomic agents
and their expectations
– Susceptible to Lucas critique
12
Small New-Keynesian Model
This model borrows key results from New Keynesian economics
that emphasizes market imperfections, especially, price rigidities
Model usually specified in terms of gaps—how much each
variable is deviated from its long-run, equilibrium value
While model focuses on key aggregate relationships
–IS relation, Phillips curve, Taylor rule, UIP condition
they lack structural features that underline aggregate dynamics
13
14
DSGE Model
New Neoclassical Synthesis
Neoclassical RBC
models
Optimizing agents
Perfect
competition
No rigidities
New Keynesian
imperfections
Monopolistic
competition
Real and nominal
rigidities
DSGE Model = Sticky-Price RBC Model
Long run
Driving forces = real factors
(e.g., technology shocks)
Short run
Monetary policy does have real effects
because of price rigidities
15
Why DSGE Models?
DSGE models characterize the workings of the economy
Rational expected utility-maximizing agents
and forward-looking central bank
Both demand and supply sides of the economy
Stock-flows consistency between various variables
in order to explain
Short-run business cycle fluctuations
Long-run growth path of the economy
16
Why DSGE Models?
DSGE models provide coherent framework
for policy analysis and forecasting
Provide counterfactual experiments/simulations
Answer questions relating to structural changes
Identify sources of shocks in the economy
Forecast and predict the effects of policy changes
Evaluate alternative policies
17
What can DSGE offer the policymaker?
It aims at ‘understanding’ rather than just ‘forecasting’
– A rigorous structural model is required to uncover the underlying
driving forces of economic development
It imposes logical discipline on the policymaker
– Effects of monetary policy works in complex ways especially for
small open economies; Interdependence of economic variables
must be acknowledged
It is conceptually a natural tool for policy experiments
– The framework is free from ‘Lucas critique’, and is therefore
more stable for alternative policy experiments
18
Policymaking process
Traditional process
Inputs
Assumptions &
risk assessment
With DSGE
Inputs
Assumptions &
risk assessment
Priors
Standard
Macroeconometric
model
DSGE model
Scenarios
evaluation
Policy
experiments
Transmission
linkages
Tunings
Forecasts
Tunings
Forecasts
Judgments
Balance of risk &
policy decision
Judgments
Balance of risk &
optimal policy
DSGE models at policy institutions
Institution/Model
Federal Reserve Board
SIGMA
IMF
GEM & GIMF
Central Bank of Chile
MAS
Riksbank
RAMSES
Bank of Thailand
BOTDSGE
Key applications
Increased government spending, rising home
consumption demand, falling currency risk
premiums, changes in foreign demand, permanent
productivity growth, reductions in labor and capital
tax rates
Multi-country open economy;
Fiscal analysis (GIMF)
Contribution of shocks to business cycle,
Effects of copper price shocks under different
fiscal rules
Scenario analysis
given alternative interest rate paths
Macro-financial linkage:
financial accelerators
19
20
Applications of DSGE Models at policy institutions
Examples of policy analysis
A. Effectiveness of alternative monetary policy rules
(Laxton and Pesenti, 2003)
B. Effects of structural reform policies
(Bayoumi, Laxton, and Pesenti, 2004)
C. Natural rates of output and interest
(Edge, Kiley, and Laforte, 2008)
21
A. Effectiveness of Alternative Monetary Policy Rules
Two-country DSGE model used to assess effectiveness
of Taylor rules in Czech Republic
Key findings
Rules that perform well in closed economy also perform well in another
closed economy
Yet, maybe inefficient when applied to small open economies
Such rules respond too weakly to forecasts of inflation and
too strongly to movements in output gap
In small, open, emerging economies, a modified rule that responds
strongly to inflation produces better macroeconomic performance
22
B. Effects of structural reform policies
Two-country DSGE model where price and wage markup parameters
of euro area are lowered to U.S. levels
to simulate impact of increasing product & labor market competition
Key findings
Euro area output and welfare rises significantly
There are positive spillovers to the rest of the world
Structural reforms ease the task of monetary policymakers in euro area
Increased competition reduces nominal rigidities in euro area
Costs of stabilizing output reduced, thereby making it easier
to use monetary policy in a counter-cyclical manner
23
C. Natural rates of output and interest
A disaggregated DSGE model for U.S. economy studying
historical evolution of natural rates of output and interest
Potential output
Output that would prevail absent wage and price rigidities and
abstracting from shocks to markups
– Different from a more traditional view of potential output
as a smoothly evolving series
In short, DSGE models facilitate assessing how structural
features of the economy affect its responses to shocks
Reduced-form models If someone disagrees with predicted
dynamics, it may be difficult to discuss why
VARs impose very little restrictions (just regressing on past data),
making it difficult to identify what is behind model dynamics
Structural models are superior as they offer precise and
coherent framework of where restrictions come from
If policymakers disagree with the results, it is possible for staff
to identify what is behind them and fix them
Two-way interaction between policymakers and staff
24
25
3.
Building models for policy purposes
General Approach and Objective
Identify key features/stylized facts of economy
Balanced growth path, steady-state ratios
Slow adjustments in prices and wages
Slow adjustments in GDP components, e.g. consumption,
investment
Mimic those features by incorporating in models:
Calibrate key stylized facts, e.g. C/Y = 0.60
Nominal rigidities: contractual price/wage setting processes
Real rigidities: habit persistence, investment adjustment costs
26
Model environment of a baseline DSGE model
Agents
Households
Firms
Government
Central Bank
What they do
Features
Consume and invest
Consumption habit persistence;
Supply labor to firms and set wage
Monopolistic competitive labor
market; wage rigidities
Hire inputs and produce
Invest
Monopolistic competitive market
price rigidity
Investment adjustment costs
Spend according to fiscal rule
Set interest rate according to
monetary policy rule
Fixed proportion of nominal GDP
Inflation targeting
27
27
28
3.1. Model structure
Model consists of
a) First-order conditions Fully/partially
micro-founded
and policy rules
behavior
b) Exogenous processes
c) Market clearing conditions
d) Steady-state conditions
Transition
dynamics
N equations
in N unknowns
Transition
dynamics
&
terminal
conditions
Dynamic path of
model variables
can be solved
(Think of solving a system
of differential equations)
29
3.1. Model structure
a) First-order conditions
1. Firm’s input demand problem
Optimal inputs demanded
micro-founded
behavior
30
3.1. Model structure
2. Firm’s pricing problem
micro-founded behavior
Optimal price setting
Price rigidity parameter characterize
Markup parameter
underlying
structure of
economy
Price in long run = markup over marginal cost
Price in short run depends on how difficult it is for firms to change prices
3.2. Model parameterization
Two broad methods
Calibration Select parameters based on empirical findings
that result in model that best characterizes Thai economy
Estimation Potential difficulties with likelihood
– limited data
– structural breaks
– multiple local maxima
32
4. Challenges and New Directions
33
Potential weaknesses in DSGE models
Reliance on individual rationality
Near-absence of heterogeneity
Simplifying assumptions:
– Assuming perfect financial market: missing macro-financial linkage
– Assuming Ricardian equivalence: no role of fiscal policy
Approximate model solution
– Missing nonlinearities originally built in model
34
4.1. Models need to strike balance
between simplicity and complexity
Not enough emphasis on financial market imperfections
– Standard models focus on real economy exclusively
– Probably missing real-financial linkage in the economy
While some models study liquidity problems in financial markets
– They do not explain why some assets that previously
were very liquid suddenly stop being traded
Possible improvement
Integrate financial imperfections in addition to standard nominal and
real rigidities in macro models—recent study by Bank of Thailand
More on last day of Workshop
35
4.2. Model solutions need to incorporate
risks and uncertainty
Although DSGE models themselves admit nonlinearities,
model solution relies on linearizing the model
– We focus on small shocks in the neigbourhood of steady state
– Buiter (2009): We took these models “into the basement and beat
them with a rubber hose until they behaved”
In such linear quadratic framework equations describing economy are
linear and objective function specifying policy goals is quadratic
– Minimizing inflation and output volatility subject to linear model
36
4.2. Model solutions need to incorporate
risks and uncertainty
Under these assumptions, monetary policy is certainty equivalent
– Policy responses do not depend on variances or any other aspect of
probability distribution of shocks
– Mishkin (2008): In such an environment, monetary policy “does not
focus on risk management”
Possible improvement
Solving model using higher-order approximation methods that retain
information discarded by linearization (first-order approximation)
37
4.3. DSGE models need to have
good forecast performance
It was only until recently that DSGE models can
Track/forecast time series as well as VAR (Smets and Wouters, 2003)
Incorporate extraneous predictions and judgments consistent with
underlying structure of the model (Benes et al., 2009)
Predict future path of nonmodeled variables (Schorfheide and Sills, 2009)
These improvements will help
Utilize information on high-frequency (monthly) indicators in
(quarterly) DSGE models
Gain more accuracy of initial condition of forecast
More on last day of Workshop
38
We need macro models for policymaking process
Success of central banks rests in appropriate and timely
response of monetary policy to disturbances to anchor
inflation expectation
Role of modelers and forecasters is to provide clear framework
of the working of economy and monetary policy transmission
Models help policymakers take appropriate action and
communicate their decisions to public
– Provide consistent framework to think about forecast and
to evaluate alternative monetary policy
39