The difference between VAR and SVAR
VAR
Var or vector autoregression or unstructured vector autoregression is a way of learning the relationship between one variable without structural limitations such as time.
However, the problem with VAR is that sometimes there is no particular rule on seeing both data related to each other.
For example, in monetary policy, let's say we want to control the inflation of the price; as we expect inflation to rise, the monetary authority will increase its instrument and the interest rate and expect the commodity price to go down.
It turns out that the price is not going down; instead, it is rising. Then, without any rule in conclusion, opening the tide of economic mystery, we will conclude that raising the interest rate will also increase inflation or commodity prices.
With this, we can shed a little bit of light that using vector autoregression, without specific constraints and boundaries, will be dangerous.
SVAR or Structural vector autoregression
Meanwhile, structural vector autoregression is similar to vector autoregression but with many more constraints, for example, time.
One of the differences is in the sample. For example, the government attempted to reduce inflation by increasing the interest rate.
In the unstructured vector autoregression, when increasing the interest rate raises inflation, we just read it as it is, which will create an erroneous conclusion.
Meanwhile, in structure vector autoregression, we have a time duration that can help explain when the price increases, in which period the inflation goes down or starts to take effect, and whether it's significant.
When it is better to use SVAR instead of VAR
1. Why SVAR
Svar is a fascinating development in macroeconomics from Christopher Sims.
For example, when the bank anticipates inflation while raising interest rates, turns out that the inflation still rises.
The wrong conclusion is that the interest rate hike led to inflation.
However, the monetary policy is an endogenous reaction to expected inflation.
The same issue occurred with fiscal policy; for example, we expect that there will be a reduction in private demand and, therefore, an increase in public spending, but the output will still decline.
The wrong conclusion is that public spending will cause the work to fall.
However, fiscal policy reacted endogenously with the reduction in production.
To measure the effect of policy, we need to identify or isolate purely exogenous independent movements and how the economy reacts to them.
It is called impulse reaction.
Therefore, we need to identify the structural model that isolates the exogenous variable from the model. After the shock hits the economy, getting the structural model is called identification.
According to Sims (1986), Identification is the interpretation of historically observed variation in data in a way that allows the variation to be used to predict the consequences of an action not yet undertaken.
After the structure is identified, one can predict inflation and output growth. We can expect the fiscal effect on GDP.
Examples:
How to draw SVAR in JMULTI
In the video above, I give an example of how to get the SVAR using Jmulti. This old Windows program is being used because it's the only software that is available and easy to use.
Reference:Dąbrowski, M.A., Widiantoro, D.M. Effectiveness and conduct of macroprudential policy in Indonesia in 2003–2020: Evidence from the structural VAR models. Eurasian Econ Rev 13, 703–731 (2023). https://doi.org/10.1007/s40822-023-00244-w |
0 Comments