Markov Switching Regime: volatility value in the regimes?

[evasileiou]

I recently start to use the Eviews-8.

I apply the Hamilton's (1989) model as you present in the Manual, but I have a crucial question to make: Hamilton presents in his results the volatility in Regime 1 and the volatility in regime 2. How can I find the specific values using Eviews?

Thanks in advance.

[EViews Glenn]

Can you be specific very about which calculation you want?

[trubador]

LOG(SIGMA)s in the output denote the logarithms of volatilities for estimated regimes. Just compute the exponential values of the related parameters to obtain volatilities (or standard deviations) for each regime.

[EViews Glenn]

Ah, the volatilities were based on the content of the variables, and not derived from the estimates. Thanks Trubador.

[evasileiou]

My results regarding the log(sigma) are the following:
Log(Sigma) Coefficient Std.Error z-Statistic Prob.
regime 1 -4.68 0.025 -189.78 0.0000

regime 2 -3.75 0.024 -159.40 0.0000


My question is the following: how can I calculate each regime's
(i) volatility

(ii) the volatility's std. error

and the

(iii) volatility's p-value.

Thanks in advance!

[EViews Glenn]

Is there only the one coefficient in each regime? Since I don't know the details of your application, it's hard to say. You might want to post the full results for the estimation or perhaps your workfile. It might be as easy as computing the Wald test for each of the two coefficients.

[evasileiou]

In Hamilton (1989) the writer Reports the two regimes' results. This report includes: each regime's mean return and each regime's volatility. The regime's using the Hamilton approach may be distinguished in low and High volatility regimes. How can I report each regime's volatility using E-views?

[EViews Glenn]

I'm a bit slow on the uptake on this one. I wasn't thinking about the error variances as volatilities and don't have Hamilton's paper close by. Sorry.

Trubador is right, you simply want to compute the exponential of the estimate of the log(sigma). The easiest way to do this and get your standard errors and t-stats, etc. is to do a separate Wald test for each of the coefficients where you test the exponential. I don't recall what the coefficient numbers are but the representations view will tell you. Suppose the first log(sigma) coefficient is C(4) and the second is C(5). Then do a Wald test with

Code: Select all

exp(c(4))

and

Code: Select all

exp(c(5))

Note that you have to do two separate Wald tests since you don't want to do the joint test.

Lastly, note that this form of the test will test the sigmas against 0. If you want to get the variances, you'll have to square things.

[smile]

hi，two question：
1，what's the meaning of “log（sigma）” in the estimation output？
2，In Hamilton (1989) the writer Reports the two regimes' results. This report includes: each regime's mean return and each regime's error variances,namely volatility. How can i get the error variances in eviews 8?
Thank you in advance

[EViews Glenn]

log(sigma) is the estimate of the log of the square root of the error variance.