Readme: Figures & Tables for
“Repatriation Taxes”
by Chad Curtis, Julio Garín, & Saif Mehkari

Required Software/Hardware:

Software:
o Stata
o Excel
o Matlab
o Fortran Compiler (e.g. Microsoft Visual Studio) 
We have also provided the compiled code as an exe file too in case someone does not want to recompile themselves.

Hardware:
o At least 28 cores
o At least 64Gb memory 
The file may run on lower memory computers too, but have not been tested on such.

Figures: 
Runtimes are included for files that take a significant amount of time to run. These runtimes are for a computer with dual Intel Xeon E5-2680 v4 processors and 64GB of RAM.

Figure 1:

Gives the timeline of decisions and is provided as a pdf file called Figure1.pdf.

Figure 2:

Step 1: Run theCashModel.exe in the Effective Repatriation Tax folder. This file solves and simulates our baseline
model for seven different values of τ and generates the model simulated data we need to create the figure.
Alternatively, you can compile theCashModel.exe yourself using the Microsoft Visual Studio Solution
theCashModel.sln or the raw Fortran files (.f90) in the subfolder called theCashModel. Runtime for the
theCashModel.exe file is approximately 21 hours.
Step 2: Run the Stata file Figure2.do. This file reads in the files containing the simulated data and calculates the
effective repatriation tax rates and plots them against the Foley et al. (2007) estimates.

Figure 3:

Gives the policy functions and is provided as a pdf file called Figure3.pdf.

Figure 4:

Step 1: Run theCashModel.exe in the BaselineModel folder. This file solves and simulates our baseline model and
generates the model simulated data. Alternatively, you can compile theCashModel.exe yourself using the Microsoft
Visual Studio Solution theCashModel.sln or the raw Fortran files (.f90) in the subfolder called theCashModel.
Runtime for the theCashModel.exe file is approximately 3 hours.
Step 2: Run the Matlab file Figure4.m. This file reads in the files containing the simulated data and constructs the impulse responses.

Figure 5:

Step 1: Run the Matlab file Figure5.m. This file contains the data and constructs the time series of transfers.

Figure 6:

Step 1: Run theCashModel.exe in the BaselineModel folder. This file solves and simulates our baseline model and
generates the model simulated data. Alternatively, you can compile theCashModel.exe yourself using the Microsoft
Visual Studio Solution theCashModel.sln or the raw Fortran files (.f90) in the subfolder called theCashModel.
Runtime for the theCashModel.exe file is approximately 3 hours.
Step 2: Run theCashModel.exe in the Model – 0 Period folder. This file solves and simulates our baseline model
but now with 0 news periods to the repatriation tax shock and generates the model simulated data. Alternatively,
you can compile theCashModel.exe yourself using the Microsoft Visual Studio Solution theCashModel.sln or the
raw Fortran files (.f90) in the subfolder called theCashModel. Runtime for the theCashModel.exe file is
approximately 3 hours.
Step 3: Run theCashModel.exe in the Model – 1 Period folder. This file solves and simulates our baseline model
but now with 1 news period to the repatriation tax shock and generates the model simulated data. Alternatively, you
can compile theCashModel.exe yourself using the Microsoft Visual Studio Solution theCashModel.sln or the raw
Fortran files (.f90) in the subfolder called theCashModel. Runtime for the theCashModel.exe file is approximately
3 hours.
Step 4: Run the Matlab file Figure6.m. This file reads in the files containing the simulated data and constructs the cumulative responses.

Figure 7:

Step 1: Run theCashModel.exe in the Model – Psi Values folder. This file solves and simulates our baseline model
for a three different psi values and generates the model simulated data. Alternatively, you can compile
theCashModel.exe yourself using the Microsoft Visual Studio Solution theCashModel.sln or the raw Fortran files
(.f90) in the subfolder called theCashModel. Runtime for the theCashModel.exe file is approximately 9 hours.
Step 2: Run the Matlab file Figure7.m. This file reads in the files containing the simulated data and constructs the cumulative responses.

Figure 8:

Step 1: Run theCashModel.exe in the Model – 20 Period folder. This file solves and simulates our baseline model
for 20 periods with a different news structure and generates the model simulated data. Alternatively, you can
compile theCashModel.exe yourself using the Microsoft Visual Studio Solution theCashModel.sln or the raw
Fortran files (.f90) in the subfolder called theCashModel. Runtime for the theCashModel.exe file is approximately
15 hours.
Step 2: Run the Matlab file Figure8.m. This file reads in the files containing the simulated data and constructs the
cumulative responses.

Figure 9:

Step 1: Run theCashModel.exe in the BaselineModel folder. This file solves and simulates our baseline model and
generates the model simulated data. Alternatively, you can compile theCashModel.exe yourself using the Microsoft
Visual Studio Solution theCashModel.sln or the raw Fortran files (.f90) in the subfolder called theCashModel.
Runtime for the theCashModel.exe file is approximately 3 hours.
Step 2: Run theCashModel.exe in the Model – IF Materializes folder. This file solves and simulates our baseline
model with a news structure where the tax holiday may or may not occur and generates the model simulated data
with it occurring. Alternatively, you can compile theCashModel.exe yourself using the Microsoft Visual Studio
Solution theCashModel.sln or the raw Fortran files (.f90) in the subfolder called theCashModel. Runtime for the
theCashModel.exe file is approximately 3 hours.
Step 3: Run theCashModel.exe in the Model – IF Does Not Materializes folder. This file solves and simulates our
baseline model with a news structure where the tax holiday may or may not occur and generates the model simulated
data with it not occurring. Alternatively, you can compile theCashModel.exe yourself using the Microsoft Visual
Studio Solution theCashModel.sln or the raw Fortran files (.f90) in the subfolder called theCashModel. Runtime
for the theCashModel.exe file is approximately 3 hours.
Step 3: Run the Matlab file Figure9.m. This file reads in the files containing the simulated data and constructs the cumulative responses.


Tables:

Table 1:

Step 1: Get the following variables from the Compustat Industrial Database for the years 1990-2014: gvkey, fyear,
cusip2, fic, txfo, pifo, pidom, at, dt, and emp. Save this Compustat file as Firm_level.dta in Stata format.
Step 2: Get the following US marginal tax variable from John Graham (see Graham, J.R. (1996): "Prixies for the
Corporate Marginal Tax Rate," Journal of Financial Economics, 42, p.187-221): tax2, fyear, and cusip2. Save the
variables as Tax_rates.dta in Stata format.
Step 3: Run theCashModel.exe in the BaselineModel folder. This file solves and simulates our baseline model and
generates the model simulated data we need to compare to the real-world data. Alternatively, you can compile
the CashModel.exe yourself using the Microsoft Visual Studio Solution theCashModel.sln or the raw Fortran files
(.f90) in the subfolder called theCashModel.
Step 4: Run the Stata file Table1.do. This file first calls the files Firm_level.dta, Tax_rates.dta, and Deflator.dta (included in the folder) and calculates the relevant moments from the data. The description of these calculations are described in Section 3.1 in the main text. The results are exported to an Excel file titled Calibration.xlsx. Next, this code reads in the files containing the simulated data and calculates the corresponding model moments. The results are also exported to the Excel file Calibration.xlsx. The results are clearly labeled in the Excel file.

Table 2:

Step 1: Run theCashModel.exe in the AJCAModel folder. This file solves and simulates our baseline model with an
AJCA type tax holiday. Alternatively, you can compile theCashModel.exe yourself using the Microsoft Visual
Studio Solution theCashModel.sln or the raw Fortran files (.f90) in the subfolder called theCashModel. Runtime
for the theCashModel.exe file is approximately 9 hours.
Step 2: Run the Matlab file Table2.m. This file reads in the files containing the simulated data and constructs the
table.

Table 3:

Step 1: Run theCashModel.exe in the Shadow Tax\p1CertainT1 folder. This file solves and simulates our baseline
model with the tax holiday occurring with 100% probability, 1 news period, and a shadow tax of 5.49%. The code
outputs the change in transfers between steady state and the period of the news. You will see the change is
approximately zero implying that the shadow tax calibration is correct. Alternatively, you can compile
theCashModel.exe yourself using the Microsoft Visual Studio Solution theCashModel.sln or the raw Fortran files
(.f90) in the subfolder called theCashModel. Runtime for the theCashModel.exe file is approximately 3 hours.
Step 2: Repeat step 1 for the remaining subfolders in the Shadow Tax Folder:

Subfolder       Probability News Periods Shadow Tax Uncertainty
p1CertainT1       100%          1          5.49%        No
p1CertainT2       100%          2          4.58%        No
p1CertainT3       100%          3          3.61%        No
p1CertainT4       100%          4          1.95%        No
p1UncertainT2     100%          2          4.68%       Yes
p1UncertainT3     100%          3          4.51%       Yes
p1UncertainT4     100%          4          3.72%       Yes
p05CertainT1       50%          1          2.67%        No
p05CertainT2       50%          2          2.08%        No
p05CertainT3       50%          3          1.47%        No
p05CertainT4       50%          4          0.90%        No
p05UncertainT2     50%          2          2.73%       Yes
p05UncertainT3     50%          3          2.67%       Yes
p05UncertainT4     50%          4          2.24%       Yes