This file provides a guide to the replication files for "Sectoral vs. Aggregate Shocks: A Structural Factor Analysis of Industrial Production," by Andrew Foerster, Pierre-Daniel Sarte, and Mark Watson, September 2008.

All programs related to the factor analysis and the computation of the canonical model with input-output (I0) linkages are in GAUSS. The programs for the factor analysis are in the directory ...\ipMWW.  Data files are Excel files and are in the subdirectory ...\data. The programs that generate the filtering matrices from the calibrated model with IO linkages are in the subdirectory ...\Models.


Data:

1. The IO tables for the benchmark years 1977 and 1998, for different levels of disaggregation,  are in the files IO77.xls and IO98.xls respectively.

2. The files IP77.xls and IP98.xls contain the IP data (sectoral indices and shares using different levels of disaggregation) broken down by SIC and NAICS code respectively.


Programs:

(a) IPload_77.gss and IPload_98.gss load the IP data in Excel format and produce corresponding .fmt data saved  in the subdirectory ...\fmt.

(b) Principle component estimates for the statistical factor analysis are computed using factor.gss.

(c) The filtering matrices from the calibrated model with IO linkages are computed using Resolkw_77.gss and Resolkw_98.gss (in the subdirectory ...\Models) depending on the data vintage of interest. Results from the structural factor analysis using these matrices are computed using factor_io.gss, factor_io_correlation.gss, and factor_io_mc.gss.


Tables 1 and 2: Results in these tables are computed in UnivariateSummary.gss (user supplied parameters at the top of the program govern the specifics of the tables).

Tables 3, 4, and 5: Results in these tables are computed in factor.gss (user supplied parameters at the top of the program govern the specifics of the tables).

Table 6: Results are computed in factor_io_correlation.gss.

Table 7: Results are computed in factor_io_mc.gss.

Table 8: Results are computed in factor_io_correlation.gss

Table 9: Results are computed in factor_io_mc.gss.

Tables 10 and 11: Results for the structural factor model in these tables are computed in factor_io.gss. Results for the statistical factor model in Table 10 are computed in factor.gss.

Table 12: The results in this table involve 2 steps. First, set delta=1 near the top of Model_Ext_98.gss (or Model_Ext_77.gss depending on the data vintage of interest) and run Resolkw_98.gss (Resolkw_77.gss) to obtain new filtering matrices. Results are then computed in factor_io_correlation.gss and factor_io.gss using the new matrices.

Table 13: The results in this table involve 2 steps. First, set the level of disaggregation at the top of Model_Ext_98.gss (or Model_Ext_77.gss depending on the data vintage of interest) and run Resolkw_98.gss (Resolkw_77.gss) to obtain new filtering matrices. Results are then computed in factor_io_correlation.gss and factor_io.gss using the new matrices.


Figures 1 and 2: These are computed in UnivariateSummary.gss.

Figures 3 and 4: These are computed in factor.gss.






 



 

     







