I am currently working on a paper to testing the impact of 'a dummy' on volatility of returns. I wish to incorporate a multiplicative dummy variable into the conditional variance equation of univariate GARCH models but the Eviews software allows addition in their variance regressors. How can this be resolved into a multiplication? I am also open to any software that can resolved this.
Thanks in advance
Can you describe more in detail what you want to study? It might be that question you asked here is too narrow, and your actual research question might be answered more easily.
Tihana Škrinjarić
adding the regressors is not the problem. The issue is eviews software seems to have a fixed structure in their variance equation and do not allow for manupulation. See for example this work:Article Political uncertainty and stock market volatility in the Mid...
Peter Molnár: I am working on the effect of political events on stock market volatility. The framework is to introduce the event dates as dummy variable in a GARCH specifications. You can see semiliar work of Article Political uncertainty and stock market volatility in the Mid...
Btw, standard GARCH model might not be the best choice for you. You can consider Range-GARCH instead, see my paper:
Article High-low range in GARCH models of stock return volatility
You are more likely to succeed with Range-GARCH, because it can better measure whether volatility was increased or decreased on a particular day. The reason is that high-low range is more informative about volatility than daily return. Confirmation that Range-GARCH works better than standard GARCH can be found also here:
Article Range-based DCC models for covariance and value-at-risk forecasting
You can do it using the following code:
'first generate returns
genr return = dlog(close)
'in case you want to conduct analysis on open-to-close returns:
genr r = log(close/open)
'standard GARCH(1,1) model can be estimated as:
equation standardgarch.arch(1,1) r
'the first step in the Range-GARCH is to generate Parkinson estimates of volatility
genr park = (log(high/low))^2/(4*log(2))
'and range-GARCH model is estimated as:
equation rangegarch.arch(0,1) r @ park(-1)
If you want to include multiplicative dummy variable D, you can consider this specification:
equation rangegarchwithdummy.arch(0,1) r @ park(-1) D*park(-1)
However, I would consider also simple additive specification:
equation rangegarchwithdummy.arch(0,1) r @ park(-1) D
Thanks Peter Molnár. Will get back on how it works on my paper
- Badges
- Science topic
- Similar topics
- Philosophy
- Esthetics
- Interpretation