Homework Assignment No. 2: due February 14, 1997

Use the programs and data in the accompanying self-extracting ZIP file [F533HW2.EXE available on the web page and on the student fileserver] to replicate some of the results from John Campbell's paper "A Variance Decomposition for Stock Returns," Economic Journal, March 1991, pp. 157-179. The disk contains all of the data and programs used to perform the computations in that paper, along with some brief documentation. Make sure you understand what the GAUSS programs are doing in calculating the GMM estimates and variance-covariance matrix.

Documentation for Campbell Assignment

 

The basic program is VAR.PRG. This runs a vector autoregression, using method of moments to compute standard errors for the variance-covariance matrix of the innovations as well as the regression coefficients. Various significance tests for blocks of coefficients are computed. The program uses real stock returns and information variables.

 

NLVAR.PRG treats the variance decomposition and persistence statistics as nonlinear functions of the parameters (including the innovation variance-covariance matrix). It computes these functions and their standard errors. It is #INCLUDEd in VAR.PRG.

 

ACVAR.PRG can also be #INCLUDEd in VAR.PRG. It computes the univariate autocorrelations of stock returns implied by the estimated VAR system.

 

BVAR.PRG and NLBVAR.PRG work with larger VARs for excess returns rather than real returns.

 

VARRAT.PRG and ACVARRAT.PRG do the variance ratio calculations reported in Figure 1 of "A Variance Decomposition...".

 

VARR2 and ACVARR2 do the implied R-Squared calculations reported in Figures 2a, 2b, and 2c of "A Variance Decomposition...".

 

DATAMAKE.PRG creates data sets called PSIST88.ASC (ASCII format) and PSIST88.FMT (Gauss matrix format). These contain real returns and forecasting variables, monthly.

 

 

DATBMAKE.PRG creates data sets called BPSIST88.ASC (ASCII format) and BPSIST88.FMT (Gauss matrix format). These contain excess returns and forecasting variables, monthly.

 

DATQMAKE.PRG creates data sets called QPSIST88.ASC (ASCII format) and QPSIST88.FMT (Gauss matrix format). These contain real returns and forecasting variables, quarterly.

 

 

---

Note: You do not have to use GAUSS. You may use any computer program that you know how to use. My understanding is that GAUSS is installed on the Pentium in the Ph.D. office and on a couple of the computers in the SSCC (ask Jarrad for more information).

---

This assignment is available in Acrobat's portable data format (.pdf). The file is about 11K and can only be viewed (and printed) using a copy of Acrobat Reader. Click here to download this document.

---

Return to FIN 533 Home Page