Joel Rodrigue, ``Multinational Production, Exports and Aggregate
Productivity."

This file contains details of an appendix to the paper, data files, and program files. Email enquiries to joel.b.rodrigue@vanderbilt.edu.


APPENDIX: appendix.pdf

The appendix describes the data, contains the full set of estimates,
further details on the counterfactual experiments and a number of robustness checks.

DATA SOURCE: Indonesian manufacturing census data

The main data source is the Survei Industri Large and Medium scale manufacturing survey collected by Statistics Indonesia (BPS). This contains all the firm level data in the paper. The data can be purchased by contacting:

Data Dissemination Service
Statistics Indonesia (BPS)
Attn. Mrs. Bana Bodri (bana@mailhost.bps.go.id)
Jl. Dr. Sutomo no. 6-8
Jakarta 10710
Indonesia

Phone: +62 21 350-7057 , ( +62 21 381-0291 ext. 3230)
Fax: +62 21 385-7046

PROGRAM FILES:

These are Stata (version 12.0) and Matlab (version 8.1.0.604 (R2013a)) programs. These files,
which are in do-file (Stata) or m-file (Matlab) format, are zipped in the file rodrigue-prgms.zip.

MP\_Exports.do

This file does the following:

1)  It produces Tables 1, 2 and 11 in the main text.  It also produces column
  2 of Table 5 and the top panel of Tables 5, 6 and 7 in the main text.

2)  It produces Tables 2, 3, the top panel of Table 4 in the Supplementary Appendix.

The user must set several program parameters.  The user must choose the industry for the transition probability matrix across different export and ownership status and the industry for the revenue transition matrix.

ML.m

This file does the following:

1) It estimates the coefficient vector (rho_chi, rho_x1, rho_x0, rho_m, mu_tau, c, c_x, c_m, xi, zeta_c, sigma_ah, sigma_af, phi_b, s_x, s_m, zeta_s, sigma_tau, lambda_0, lambda_1, lambda_2 from the panel data set.

This program is constructed to retrieve the coefficient vector using maximum likelihood according to the dynamic discrete choice estimator suggested in Rust (1987).

The user must set several program parameters.  The user must choose the industry to estimate, whether to include wholly owned government plants in the estimation, whether to include returning plants, whether to include plants that earn 100 percent of revenue from export sales and the fixed and sunk cost assumptions.  The subroutines called by this program include FP.m, BHHH.m and fdi_stat.m.

counterfactual.m

This file does the following:

1) This program estimates the counterfactual experiments.  The experiments are estimated relative to the base model estimated in ML.m.

The user must set several program parameters.  The user must choose the industry, the fixed and sunk cost assumptions and the counterfactual to be estimated.  There are three counterfactuals: autarky, no trade and no foreign direct investment.  The subroutines called by this program include FP_Price.m, counter.m, and fdi_stat_counter.m.

Subroutines:

FP.m

This program runs the dynamic program, given the parameter vector, calculates the conditional choice probabilities, stationary distributions and constructs the likelihood function.

BHHH.m

This program is identical to FP.m with the exception that the parameter vector is not updated.  It is used in a routine to calculate the standard errors of the parameters.

fdi_stat.m

This program displays the results for Tables 4-7.  It also retrieves the required information from the estimated model for the counterfactual experiments.

FP_Price.m

This program runs the dynamic program, given the counterfactual parameter vector, and calculates how far the the current counterfactual vector is from the counterfactual equilibrium.

counter.m

This program recovers the counterfactual results under the equilibrium counterfactual parameter vector.

fdi_stat_counter.m

This program is the counterfactual counterpart to fdi_stat.  It calculates and displays the counterfactual results.

lprint.m

This program displays results in a table.