Scribd
Upload
460 views

Partial Adjustment Model

The partial adjustment model describes how the desired level of a variable is determined and how it adjusts over time. It consists of two equations: one for the desired level and one for the partial adjustment process. The error correction model (ECM) is derived from the autoregressive distributed lag (ARDL) model and models both short-run dynamics and the long-run equilibrium relationship between variables. It contains an error correction term that represents the deviation from long-run equilibrium and ensures convergence over time. The ECM can be used to estimate both the short-run coefficients and long-run elasticities between variables.

Uploaded by

nai09
Copyright
© © All Rights Reserved
Available Formats
Download as DOC, PDF, TXT or read online on Scribd
460 views

Partial Adjustment Model

The partial adjustment model describes how the desired level of a variable is determined and how it adjusts over time. It consists of two equations: one for the desired level and one for the partial adjustment process. The error correction model (ECM) is derived from the autoregressive distributed lag (ARDL) model and models both short-run dynamics and the long-run equilibrium relationship between variables. It contains an error correction term that represents the deviation from long-run equilibrium and ensures convergence over time. The ECM can be used to estimate both the short-run coefficients and long-run elasticities between variables.

Uploaded by

nai09
Copyright
© © All Rights Reserved
Available Formats
Download as DOC, PDF, TXT or read online on Scribd
You are on page 1/ 3

Partial Adjustment Model

The partial adjustment model comprises two parts, a static part to describe how the
desired amount is determined and the dynamic partial adjustment process:

y t*   0  1 xt  u t
y t  y t 1   ( yt*  y t 1 )

Where y* is the desired level of y. By substituting the expression for y* into the other
equation we obtain the following estimating equation:

yt   0   (1   ) yt 1   1 xt  u t

We can estimate this equation as a general ARDL model as follows:

y t   0  1 yt 1   2 xt   3 xt 1  vt

In this case the following restriction would be imposed if partial adjustment occurred:

3  0

In addition we could get estimates of the parameters in the original equation


containing the desired level of y, as well as the adjustment parameter λ. In the above
case:

1  (1   )    (1  1 )
 2   1
 0   0

The adjustment parameter λ measures the speed of adjustment and lies between 0 and
1. The closer it is to 1 the faster the speed of adjustment. An example of this model is
the Lintner Dividend-Adjustment Model.

The Error Correction Model

The error correction model (ECM) is an example of a short-run dynamic model,


which is used to model both economic and financial time series. It is expressed in first
differences ( yt  yt  yt 1 ), except the error correction term. It is based on the
ARDL model again and can be derived from this model, with the addition of a
specific restriction.

y t   0  1 yt 1   2 xt   3 xt 1  vt

To produce the ECM, firstly you have to subtract yt-1 from both sides of the ARDL
equation:

yt  yt 1  yt   0  1 yt 1   2 xt   3 xt 1  yt 1  vt
The next step is to express the x in difference form. This involves adding and
subtracting  2 xt 1 from the right hand side of the above equation.

y t   0  1 y t 1   2 xt   2 xt 1   3 xt 1  y t 1   2 xt 1  vt

Collecting terms gives:

yt   0   2 xt  ( 1  1) yt 1  (  2   3 ) xt 1  vt

In order to produce the ECM, we need to assume that the coefficient on yt-1 is equal to
minus the coefficient on xt-1. That means:

1  1  (  2   3 )
1   2   3  1

The sum of the coefficients, excluding the constant, must sum to one in the ARDL
model, for the ECM to apply. The ECM is usually written with τ as the coefficient on
the error correction term, to give the following:

yt   0   2 xt   ( y t 1  xt 1 )  ut
  ( 1  1)  (  2   3 )

The above ECM is the short-run relationship between y and x. The long-run
relationship can be formed in the usual way, although instead of assuming all
differenced terms equal 0, we assume that they grow at a constant rate g. This gives:

g   0   2 g   ( y *  x*)
 ( y *  x*)   0  (  2  1) g
  (  2  1) g
y*  0  x*

If the original model was: y *t  kx *t which in logs is log y *t  log k  log x *t .


Then in terms of the above long-run expression, when antilogged:

 0  (  2  1) g
k  exp[ ]

The term k can be interpreted as the long-run relationship between y and x. i.e. if y is
consumption and x is wealth, k would be the average propensity to consume from
wealth.

You might also like