请教Bernanke&Mihov（1998）structural VAR的方法

是 否 +2 论坛币

k人 参与回答

经管之家送您一份

应届毕业生专属福利!

求职就业群

赵安豆老师微信：zhaoandou666

经管之家联合CDA

送您一个全额奖学金名额~ !

立即领取

感谢您参与论坛问题回答

经管之家送您两个论坛币！

+2 论坛币

各位大侠，小弟有一问题请教。最近在看Ben Bernanke和Mihov 1998年QJE的那篇“Measuring Monetary Policy”，对里面的方法不是很清楚。他们用?个变量估计不同的货币政策，如fed fund rate targeting，Nonborrowed reserves targeting。小弟从Bernanke在princeton的网站上搞到了数据和RATS的代码。http://www.princeton.edu/~bernanke/data.htm由于对RATS不太熟悉，小弟尝试用Eviews来replicate他们的结果。但是发现与论文中的大相径庭。以下是设定structural VAR的部分代码

************************************************************************
* MEASURE. SRC
*
*  Procedure to accompany the calulations of the strutural parameters
*    for Model A as described in Bernanke and Mihov (1995)
*
*  The procedure takes the variables:
*    NVAR [integer] - the number of endogenous variables in the VAR
*    FFM[integer] - the location index for the FF market (TR, NBR, FFR);
*                   effectively should contain the index for TR
*    VMAT[symmetric] - the variance-covariance matrix of residuals
*
* The following two variabes are defined prior to the procedure
*       and are altered by it:
*    ZET[vector] - a vector of period zero shocks
*    SP[vector] - a vector which stores the structural parameters
*       in the following order: A, B, Fd, Fb, std(Vd), std(Vs), Std(Vb) 
************************************************************************
PROCEDURE measure nvar ffm vmat zet sp

     TYPE symmetric vmat
     TYPE vector *sp *zet

     DEC rect vmat2 v1
     DEC symmetric vmat1

     DIMENSION vmat2(3,3) v1(3,3) vmat1(nvar,nvar)
     COMPUTE vmat1 = tr(vmat)

     compute chol=%decomp(vmat1)
        compute chol(ffm  ,ffm)   = 0
        compute chol(ffm+1,ffm)   = 0
        compute chol(ffm+1,ffm+1) = 0
        compute chol(ffm+2,ffm)   = 0
        compute chol(ffm+2,ffm+1) = 0
        compute chol(ffm+2,ffm+2) = 0
    compute cholsq = chol*tr(chol)
    compute vmat1 = vmat1 - cholsq

     DO j2 = ffm,ffm+2
             DO i2=ffm,ffm+2
                   compute VMAT2(1+i2-ffm,1+j2-ffm) = vmat1(i2,j2)
             END DO i2
     END DO j2

     COMPUTE v1 = vmat2

     compute sd = v1(1,1)
     compute fd = v1(2,1)/sd
     compute b = (1-fd)*sd/v1(3,1)

     compute dd = fd*fd*sd - v1(2,2)
     compute ee = (1-fd)*fd*sd - b*v1(3,2)
     compute ff = (1-fd)*(1-fd)*sd - b*b*v1(3,3)

     compute sb = dd - ff - 2*(dd+ee)
     compute ss = -dd - (dd+ee)*(dd+ee)/sb
     compute fb = (dd + ee)/sb

     compute std = sqrt(sd)
     compute sts = sqrt(ss)
     compute stb = sqrt(sb)

     compute sp(1) = 0
     compute sp(2) = b
     compute sp(3) = fd
     compute sp(4) = fb
     compute sp(5) = std
     compute sp(6) = sts
     compute sp(7) = stb
  
     DO i1=1,nvar
        compute zet(i1) = chol(i1,ffm+1)
     END DO i1

     compute zet(ffm+1) =  sts   
     compute zet(ffm+2) = -sts/b
end

在之前估计的VAR中，6个变量的依次顺序是真实GDP，GDP平减指数，价格指数，总储备，非借贷性储备，联邦储备利率。他模型的假设是后三个政策变量的扰动项和前三个变量的扰动项是正交的。也就是说所有对于structural VAR的restriction都在后三个政策变量上。以上的代码应该是针对后三个政策变量的structural VAR。最后估计的是a，b，fd，fb四个参数。

问题在于，如果用Eviews做，对应的A矩阵和B矩阵是怎么样的？

[此贴子已经被作者于2008-2-11 11:38:04编辑过]

扫码加我 拉你入群

请注明：姓名-公司-职位

以便审核进群资格，未注明则拒绝

分享0 收藏1 回帖

关键词：Structural Bernanke struct Mihov Ber 请教 VaR Bernanke Structural Mihov