Calibrating the Model
It was Kydland and Prescott (1982) who first demonstrated that a general
equilibrium real business cycle model, driven by exogenous technological
shocks, was capable of generating time series data that possessed the statistical
properties of US business cycles over the period 1950–79. However,
real business cycle theorists have not generally attempted to provide models
capable of conventional econometric testing but have instead tended to
focus on providing numerical examples of a more general theory of
fluctuations. In order to examine the quantitative implications of their mod
els, real business cycle theorists have developed a method known as ‘calibration’
or ‘computational experiments’. Cooley (1997) defines calibration
as ‘a strategy for finding numerical values for the parameters of artificial
economies’ and involves a ‘symbiotic relationship between theory and measurement’.
The calibration strategy consists of the following steps (see
Kydland and Prescott, 1982, 1991, 1996; Plosser, 1989; Backhouse, 1997b;
Abel and Bernanke, 2001):
1. Pose a question relating to a specific issue of concern, for example an
important policy issue such as ‘What is the quantitative nature of
fluctuations caused by technology shocks?’
2. Use a ‘well-tested’ theory, where ‘theory’ is interpreted as a specific set
of instructions about how to build the imitation economy.
3. Construct a model economy and select functional forms. Kydland and
Prescott (1982) utilize the basic stochastic neoclassical growth model as
the cornerstone of their model.
4. Provide specific algebraic forms of the functions used to represent production
and consumption decisions. For example, a specific Cobb–Douglas
production function is used by Plosser (1989).
5. Calibrate the model economy using data from pre-existing microeconomic
studies and knowledge of the ‘stylized facts’. Where no information
exists select values for parameters so that the model is capable of mimicking
the real-world behaviour of variables.
6. The calibration exercise then involves simulating the effect of subjecting
the model to a series of random technology shocks using a computer.
7. The impact that these shocks have on the key macroeconomic variables
is then traced out so that the results can be compared with the actual
behaviour of the main macroeconomic time series.
8. Run the experiment and compare the equilibrium path of the model
economy with the behaviour of the actual economy. Use these types of
simulations to answer questions relating to the important issues initially
identified under (1).
In their seminal 1982 paper Kydland and Prescott use the neoclassical growth
model and follow the calibration/simulation procedure to see if the model can
explain aggregate fluctuations when the model economy is subject to technology
shocks. As Prescott (1986) recalls, ‘the finding that when uncertainty in
the rate of technological change is incorporated into the growth model it
displays business cycle phenomena was both dramatic and unanticipated’.
The simulations carried out by Kydland, Prescott and Plosser produced some
impressive results in that their models are able to mimic an actual economy
with respect to some important time series data. These simulations indicate
that a competitive economy hit by repeated technology shocks can exhibit the
kind of fluctuations that are actually observed.
On the negative side, one of the problems with calibration is that it currently
does not provide a method that allows one to judge between the
performance of real and other (for example Keynesian) business cycle models.
As Hoover (1995b) notes, ‘the calibration methodology, to date, lacks
any discipline as stern as that imposed by econometric methods … Above all,
it is not clear on what standards competing, but contradictory, models are to
be compared and adjudicated.’ Nevertheless calibration has provided an important
new contribution to the methodology of empirical macroeconomic
research. While initially the calibration methodology was focused on business
cycle research, more recently calibrated models have been used to
investigate issues in public finance, economic growth, industry, firm and
plant dynamics and questions related to the choice of economic policy (Cooley,
1997). For more detailed discussions and critiques of the calibration methodology
see Kydland and Prescott (1991, 1996); Summers (1991a); Quah (1995);
Hoover (1995b); Wickens (1995); Hansen and Heckman (1996); Sims (1996);
Cooley (1997); Hartley et al. (1998).
It was Kydland and Prescott (1982) who first demonstrated that a general
equilibrium real business cycle model, driven by exogenous technological
shocks, was capable of generating time series data that possessed the statistical
properties of US business cycles over the period 1950–79. However,
real business cycle theorists have not generally attempted to provide models
capable of conventional econometric testing but have instead tended to
focus on providing numerical examples of a more general theory of
fluctuations. In order to examine the quantitative implications of their mod
els, real business cycle theorists have developed a method known as ‘calibration’
or ‘computational experiments’. Cooley (1997) defines calibration
as ‘a strategy for finding numerical values for the parameters of artificial
economies’ and involves a ‘symbiotic relationship between theory and measurement’.
The calibration strategy consists of the following steps (see
Kydland and Prescott, 1982, 1991, 1996; Plosser, 1989; Backhouse, 1997b;
Abel and Bernanke, 2001):
1. Pose a question relating to a specific issue of concern, for example an
important policy issue such as ‘What is the quantitative nature of
fluctuations caused by technology shocks?’
2. Use a ‘well-tested’ theory, where ‘theory’ is interpreted as a specific set
of instructions about how to build the imitation economy.
3. Construct a model economy and select functional forms. Kydland and
Prescott (1982) utilize the basic stochastic neoclassical growth model as
the cornerstone of their model.
4. Provide specific algebraic forms of the functions used to represent production
and consumption decisions. For example, a specific Cobb–Douglas
production function is used by Plosser (1989).
5. Calibrate the model economy using data from pre-existing microeconomic
studies and knowledge of the ‘stylized facts’. Where no information
exists select values for parameters so that the model is capable of mimicking
the real-world behaviour of variables.
6. The calibration exercise then involves simulating the effect of subjecting
the model to a series of random technology shocks using a computer.
7. The impact that these shocks have on the key macroeconomic variables
is then traced out so that the results can be compared with the actual
behaviour of the main macroeconomic time series.
8. Run the experiment and compare the equilibrium path of the model
economy with the behaviour of the actual economy. Use these types of
simulations to answer questions relating to the important issues initially
identified under (1).
In their seminal 1982 paper Kydland and Prescott use the neoclassical growth
model and follow the calibration/simulation procedure to see if the model can
explain aggregate fluctuations when the model economy is subject to technology
shocks. As Prescott (1986) recalls, ‘the finding that when uncertainty in
the rate of technological change is incorporated into the growth model it
displays business cycle phenomena was both dramatic and unanticipated’.
The simulations carried out by Kydland, Prescott and Plosser produced some
impressive results in that their models are able to mimic an actual economy
with respect to some important time series data. These simulations indicate
that a competitive economy hit by repeated technology shocks can exhibit the
kind of fluctuations that are actually observed.
On the negative side, one of the problems with calibration is that it currently
does not provide a method that allows one to judge between the
performance of real and other (for example Keynesian) business cycle models.
As Hoover (1995b) notes, ‘the calibration methodology, to date, lacks
any discipline as stern as that imposed by econometric methods … Above all,
it is not clear on what standards competing, but contradictory, models are to
be compared and adjudicated.’ Nevertheless calibration has provided an important
new contribution to the methodology of empirical macroeconomic
research. While initially the calibration methodology was focused on business
cycle research, more recently calibrated models have been used to
investigate issues in public finance, economic growth, industry, firm and
plant dynamics and questions related to the choice of economic policy (Cooley,
1997). For more detailed discussions and critiques of the calibration methodology
see Kydland and Prescott (1991, 1996); Summers (1991a); Quah (1995);
Hoover (1995b); Wickens (1995); Hansen and Heckman (1996); Sims (1996);
Cooley (1997); Hartley et al. (1998).
No comments:
Post a Comment