Dynamic conditional correlation multivariate GARCH

[malagesh]

does anybody know how to incorporate asymmetry in DCC-GARCH model? i will be grateful to you if somebody can provide me asymmetric DCC-GARCH programme?
thanking you

[Norling87]

Hey!

The above given code works very well for estimating the DCC-GARCH, but as I understand it, the code uses the whole sample to estimate the correlation and uses that sample to estimate the correlation at each time t.

In my study, I want to use the DCC-GARCH in the bivariate case and roll a in-sample period of for say 500 observations to forecast an out-of-sample of, say 800 obervations. I want it to be a rolling-window forecast, shifting the in-sample 1-period ahead when forecasting the next days correlation. I tried to modify the code but wasn't even close to succeeding. Can anyone help me with this?

Regards

[charlie295]

Hi everybody,

I have managed to use the code provided kindly by hvtcapollo and produced reasonable rho values.

Probably an easy question but im not an expert in statistics, which of the series in the code gives the values of the conditional variances/st devs? (is it var_z1 and var_z2 or Z1/Z2 or something else)

please help. I need to know urgently

thanks a lot

[tg128]

You can consider this code. I used it last year for my research and you should be ok if using it for the bivariate. For trivariate u need to modify a litle bit especially for the log likelihood function.

Code: Select all

'change path to program path %path=@runpath cd %path 'load workfile containing the return series load nikkei_sp.WF1 'set sample range sample s1 1/06/1995 12/25/2007 scalar pi=3.14159 'defining the return series in terms of y1 and y2 series y1=r_nikkei series y2=r_sp 'fitting univariate GARCH(1,1) models to each of the two returns series equation eq_y1.arch(1,1,m=1000,h) y1 c equation eq_y2.arch(1,1,m=1000,h) y2 c 'extract the standardized residual series from the GARCH fit eq_y1.makeresids(s) z1 eq_y2.makeresids(s) z2 'extract garch series from univariate fit eq_y1.makegarch() garch1 eq_y2.makegarch() garch2 'Caculate sample variance of series z1, z2 and covariance of z1and z2 and correlation between z1 and z2 scalar var_z1=@var(z1) scalar var_z2=@var(z2) scalar cov_z1z2=@cov(z1,z2) scalar corr12=@cor(z1,z2) 'defining the starting values for the var(z1) var(z2) and covariance (z1,z2) series var_z1t=var_z1 series var_z2t=var_z2 series cov_z1tz2t=cov_z1z2 'declare the coefficient starting values coef(2) T T(1)=0.2 T(2)=0.7 ' ........................................................... ' LOG LIKELIHOOD for correlation part ' set up the likelihood ' 1) open a new blank likelihood object and name it 'dcc' ' 2) specify the log likelihood model by append ' ........................................................... logl dcc dcc.append @logl logl 'specify var_z1t, var_z2t, cov_z1tz2t dcc.append var_z1t=@nan(1-T(1)-T(2)+T(1)*(z1(-1)^2)+T(2)*var_z1t(-1),1) dcc.append var_z2t=@nan(1-T(1)-T(2)+T(1)*(z2(-1)^2)+T(2)*var_z2t(-1),1) dcc.append cov_z1tz2t=@nan((1-T(1)-T(2))*corr12+T(1)*z1(-1)*z2(-1)+T(2)*cov_z1tz2t(-1),1) dcc.append pen=(var_z1t<0)+(var_z2t<0) 'specify rho12 dcc.append rho12=cov_z1tz2t/@sqrt(@abs(var_z1t*var_z2t)) 'defining the determinant of correlation matrix and determinant of Dt dcc.append detrRt=(1-(rho12^2)) dcc.append detrDt=@sqrt(garch1*garch2) dcc.append pen=pen+(detrRt<0) dcc.append detrRt=@abs(detrRt) 'define the log likelihood function dcc.append logl=(-1/2)*(2*log(2*pi)+log(detrRt)+(z1^2+z2^2-2*rho12*z1*z2)/detrRt)-10*pen 'estimate the model smpl s1 dcc.ml(showopts, m=500, c=1e-5) 'display output and graphs show dcc.output graph corr.line rho12 show corr

thanks for the code!

i wonder how to compute q11,t and q22,t from the DCC model and save them seperately please?

[ruhee.mittal]

i am not able to get answers with this code.
is this code incomplete??
please help.

[ruhee.mittal]

hello,
i am not able to generate logl series with the above program. thus i am not able to reach the final results. kindly help me in solving this problem.
thnk you..

[nikolastico]

Hello to all,

I'm currently trying to using this code I copied from another post and modified it to allow assymetries (ADCC). I'm trying to testing contagion on index returns series...

Code: Select all

'ADCC Model by Cappiello et al (2006) 'Load workfile and measure length of series series obscount = 1 scalar obslength = @sum(obscount) 'Specify the return series series y1 = bovespa series y2 = igbvl 'Specify the number of iterations in the MLE (Engle & Sheppard (2001) used just one iteration) !itermle = 1000 'Specify the initial values of the parameters coef(1) alpha coef(1) beta coef(1) gamma 'Values may vary depending on which will result in the highest Likelihood value alpha(1) = 0.04 beta(1) = 0.94 gamma(1) = 0.02 'Setting the sample sample s0 @first+1 @last sample s1 @first+2 @last sample sf @first+3 @last sample sf_alt @first+13 @last 'Initialization at sample s0 smpl s0 'Each return series is modeled with their respective GARCH specifications: 'Use Bollerslev-Wooldridge QML 'Standard errors equation eq1.arch(1,1,thrsh=1,m=1000,c=1e-5,h) y1 c y1(-1) 'res_s11_1000 @ res_s11_1000 equation eq2.arch(1,1,thrsh=1,m=1000,c=1e-5,h) y2 c y2(-1) 'res_s22_1000 @ res_s22_1000 'Make residual series eq1.makeresids e1 eq2.makeresids e2 'Make a garch series from the univariate estimates eq1.makegarch h11 eq2.makegarch h22 'Normalizing the residuals from e to e* (named as "e1n" and "e2n") series e1n = e1/h11^0.5 series e2n = e2/h22^0.5 'Make residual series for asymmetries in DCC model series n1n = @recode(e1n<0,e1n*e1n,0) series n2n = @recode(e2n<0,e2n*e2n,0) 'The Q_bar=E[en*en'] components and its sample equivalent series qbar11 = @mean(e1n*e1n) series qbar12 = @mean(e1n*e2n) series qbar21 = @mean(e2n*e1n) series qbar22 = @mean(e2n*e2n) 'The N_bar series nbar11 = @mean(n1n*n1n) series nbar12 = @mean(n1n*n2n) series nbar21 = @mean(n2n*n1n) series nbar22 = @mean(n2n*n2n) 'The (en*en') matrix components series e11n = e1n*e1n series e12n = e1n*e2n series e21n = e2n*e1n series e22n = e2n*e2n 'The (nn*nn') matrix components series n11n = n1n*n1n series n12n = n1n*n2n series n21n = n2n*n1n series n22n = n2n*n2n 'Initialize the elements of Qt for variance targeting series q11 = @var(e1) series q12 = @cov(e1,e2) series q21 = @cov(e2,e1) series q22 = @var(e2) '*********************************************************************************************** 'Declare a Loglikelihood object logl dcc dcc.append @logl logl 'The elements of matrix Qt dcc.append q11 = (qbar11 - (alpha(1))^2*qbar11 - (beta(1))^2*qbar11 + (gamma(1))^2*nbar11) + (alpha(1))^2*e11n(-1) + (gamma(1))^2*n11n(-1) + (beta(1))^2*q11(-1) dcc.append q12 = (qbar12 - (alpha(1))^2*qbar12 - (beta(1))^2*qbar12 + (gamma(1))^2*nbar12) + (alpha(1))^2*e12n(-1) + (gamma(1))^2*n12n(-1) + (beta(1))^2*q12(-1) dcc.append q21 = (qbar21 - (alpha(1))^2*qbar21 - (beta(1))^2*qbar21 + (gamma(1))^2*nbar21) + (alpha(1))^2*e21n(-1) + (gamma(1))^2*n21n(-1) + (beta(1))*2*q21(-1) dcc.append q22 = (qbar22 - (alpha(1))^2*qbar22 - (beta(1))^2*qbar22 + (gamma(1))^2*nbar22) + (alpha(1))^2*e22n(-1) + (gamma(1))^2*n22n(-1) + (beta(1))^2*q22(-1) 'As input to detQQQ dcc.append q12n = ((qbar11 - (alpha(1))^2*qbar11 - (beta(1))^2*qbar11 + (gamma(1))^2*nbar11) + (alpha(1))^2*e11n(-1) + (gamma(1))^2*n11n(-1) + (beta(1))^2*q11(-1))/((abs(q11)^0.5)*(abs(q22)^0.5)) dcc.append q21n = ((qbar21 - (alpha(1))^2*qbar21 - (beta(1))^2*qbar21 + (gamma(1))^2*nbar21) + (alpha(1))^2*e21n(-1) + (gamma(1))^2*n21n(-1) + (beta(1))*2*q21(-1))/((abs(q22)^0.5)*(abs(q11)^0.5)) 'Setting up the Loglikelihood function, assuming that resid ~ N(0,H) 'The Loglikelihood function is L' = -0.5*{summation from t=1 to T of ([log of determinant of inverse(diag[Qt])*Qt*inverse(diag[Qt])] + [en'*inverse(diag[Qt])*Qt*inverse(diag[Qt])*en]) 'Taking the adjoint{inverse(diag[Qt])*Qt*inverse(diag[Qt])} dcc.append detQQQ = 1 - q12n*q21n 'Tomando el adjoint{inverse(diag[Qt])*Qt*inverse(diag[Qt])} dcc.append cofact11 = 1*1 dcc.append cofact12 =(-1)*q21n dcc.append cofact21 =(-1)*q12n dcc.append cofact22 = 1*1 'Taking the inverse{inverse(diag[Qt])*Qt*inverse(diag[Qt])} dcc.append invQQQ11 = cofact11/detQQQ dcc.append invQQQ12 = cofact12/detQQQ dcc.append invQQQ21 = cofact21/detQQQ dcc.append invQQQ22 = cofact22/detQQQ 'Taking the en'*inverse{inverse(diag[Qt])*Qt*inverse(diag[Qt])}*en dcc.append enQQQen11 = e1n*invQQQ11*e1n dcc.append enQQQen12 = e1n*invQQQ12*e2n dcc.append enQQQen21 = e2n*invQQQ21*e1n dcc.append enQQQen22 = e2n*invQQQ22*e2n 'Append the loglikelihood function 'Instead of log(detQQQ) use log(abs(detQQQ)) dcc.append logl = -0.5*(log(abs(detQQQ)) + (enQQQen11+enQQQen21+ enQQQen12+enQQQen22)) 'Specifies the sample data where the estimation will be made smpl s1 'Estimates the parameters now using BHHH algorithm dcc.ml(b, showopts, m=!itermle, c=1e-5) 'A detQQQnpd = 0 indicates detQQQ is positive definite series count = (detQQQ<=0) scalar detQQQnpd = @sum(count) 'Display the estimated parameters show dcc.output 'Forecast the q's by initializing them series q11f = 0 series q12f = 0 series q21f = 0 series q22f = 0 'Specify the sample period to be forecasted smpl sf series q11f = (qbar11 - (alpha(1))^2*qbar11 - (beta(1))^2*qbar11 + (gamma(1))^2*nbar11) + (alpha(1))^2*e11n(-1) + (gamma(1))^2*n11n(-1) + (beta(1))^2*q11(-1) series q12f = (qbar12 - (alpha(1))^2*qbar12 - (beta(1))^2*qbar12 + (gamma(1))^2*nbar12) + (alpha(1))^2*e12n(-1) + (gamma(1))^2*n12n(-1) + (beta(1))^2*q12(-1) series q21f = (qbar21 - (alpha(1))^2*qbar21 - (beta(1))^2*qbar21 + (gamma(1))^2*nbar21) + (alpha(1))^2*e21n(-1) + (gamma(1))^2*n21n(-1) + (beta(1))*2*q21(-1) series q22f = (qbar22 - (alpha(1))^2*qbar22 - (beta(1))^2*qbar22 + (gamma(1))^2*nbar22) + (alpha(1))^2*e22n(-1) + (gamma(1))^2*n22n(-1) + (beta(1))^2*q22(-1) 'To plot the time-varying conditional correlation smpl sf_alt series rt = q12f/(@sqr(q11f)*@sqr(q22f)) graph rtgraph.line rt show rtgraph delete e1n e2n delete q12n q21n delete n1n n2n delete n11n n12n n21n n22n delete e11n e12n e21n e22n delete qbar* delete xbar* delete cofact* delete enqqqen* delete invqqq* delete enqqqen* delete eigenvect* delete invqqq* show detqqqnpd series eigmins scalar eigmin for !j = 4 to obslength 'Input the hijsim to the elements of a 2x2 Q matrix sym(2) Q Q(1,1) = q11f(!j) Q(2,1) = q21f(!j) Q(2,2) = q22f(!j) 'Take the eigenvalues of Q vector(4) Qeigenval Qeigenval = @eigenvalues(Q) 'Collect the min eigenvalues in each Qt eigmins(!j) = @min(Qeigenval) next eigmin = @min(eigmins) show eigmin

I don't understand the following error message that Eviews 6 shows:

Missing values in @LOGL series at current coefficients at observation 1/24/1993 in "DO_DCC.ML(B,SHOWOPTS,M=1000,C=1E-5)"

Could anyone help me with this?

Thanks in advance.

tesis_dcc.wf1

Thesis Workfile

(78.54 KiB) Downloaded 1570 times

[trubador]

Carefully place your sample periods. For instance, you should use "smpl sf" instead of "smpl s1" before calling the estimation algorithm...

[nicktz]

I am having trouble extending the DCC code to the tri-variate case... Please help!!

[snowdeer]

Dear trubador,
I used the codes you provided for DCC-GARCH for eviews 5.0 and got the warning message "unmatched parenthesis" after typing the code"dcc.ml(showopts, m=500, c=1e-5)". Would you please enlighten me how to solve the problem? Thank you very much

[trubador]

The code is not mine.

I am not sure if anything specific to version 5 is going on, but the message is pretty clear: There is at least one missing bracket somewhere in the code that you are currently using.

[nicktz]

Hi Trubador,
I have trouble running the ADCC code on the attached data set.
It keeps saying missing values at 11/09/2007. This is a random date in the middle of my workfile.
I cannot understand what the problem might be! I have checked the starting values yet nothing seems to help.
Any ideas? Much appreciated!

[trubador]

There is a typo in your likelihood specification, which leads to explosive behavior:
dcc.append q21 = (qbar21 - (alpha(1))^2*qbar21 - (beta(1))^2*qbar21 + (gamma(1))^2*nbar21) + (alpha(1))^2*e21n(-1) + (gamma(1))^2*n21n(-1) + (beta(1))*2*q21(-1)
Just replace the red part with ^

[nicktz]

Thanks so much!!!!

[Rosen89]

I have seen that there are some other posts about this but none of the answers seem to make any difference for me: the problem is that when I run the LOGL I get up that I have missing values....
If anybody could help that would be brilliant as I have not much experience in this area (which is also why the code I am using is not mine but one I found in this forum)
For info purposes: I am running a spot price against a portfolio -- both weekly returns



'change path to program path
%path=@runpath
cd %path

'load workfile containing the return series
load spot_pfh.WF1

'set sample range
sample s1 1/07/2000 6/09/2006
scalar pi=3.14159

'defining the return series in terms of y1 and y2
series y1=spot
series y2=pf_h

'fitting univariate GARCH(1,1) models to each of the two returns series
equation eq_y1.arch(1,1,m=1000,h) y1 c
equation eq_y2.arch(1,1,m=1000,h) y2 c

'extract the standardized residual series from the GARCH fit
eq_y1.makeresids(s) z1
eq_y2.makeresids(s) z2

'extract garch series from univariate fit
eq_y1.makegarch() garch1
eq_y2.makegarch() garch2

'Caculate sample variance of series z1, z2 and covariance of z1and z2 and correlation between z1 and z2
scalar var_z1=@var(z1)
scalar var_z2=@var(z2)
scalar cov_z1z2=@cov(z1,z2)
scalar corr12=@cor(z1,z2)

'defining the starting values for the var(z1) var(z2) and covariance (z1,z2)
series var_z1t=var_z1
series var_z2t=var_z2
series cov_z1tz2t=cov_z1z2

'declare the coefficient starting values
coef(2) T
T(1)=0.2
T(2)=0.7

logl dcc
dcc.append @logl logl

'specify var_z1t, var_z2t, cov_z1tz2t
dcc.append var_z1t=@nan(1-T(1)-T(2)+T(1)*(z1(-1)^2)+T(2)*var_z1t(-1),1)
dcc.append var_z2t=@nan(1-T(1)-T(2)+T(1)*(z2(-1)^2)+T(2)*var_z2t(-1),1)
dcc.append cov_z1tz2t=@nan((1-T(1)-T(2))*corr12+T(1)*z1(-1)*z2(-1)+T(2)*cov_z1tz2t(-1),1)

dcc.append pen=(var_z1t<0)+(var_z2t<0)

'specify rho12
dcc.append rho12=cov_z1tz2t/@sqrt(@abs(var_z1t*var_z2t))

'defining the determinant of correlation matrix and determinant of Dt
dcc.append detrRt=(1-(rho12^2))
dcc.append detrDt=@sqrt(garch1*garch2)
dcc.append pen=pen+(detrRt<0)
dcc.append detrRt=@abs(detrRt)

'define the log likelihood function
dcc.append logl=(-1/2)*(2*log(2*pi)+log(detrRt)+(z1^2+z2^2-2*rho12*z1*z2)/detrRt)-10*pen

'estimate the model
smpl s1
dcc.ml(showopts, m=500, c=1e-5)



This is when I get up that I have missing values -- again, if anybody has time to help I would very much appreciate it :)