1. Software and operating system used: 

Matlab R2012a/R2014b on Windows 7

2. The order in which the programs need to be run: 

Programs can be run in any order:
- lchrun.m in the folder USA gives the baseline results for the US. 
- lchrun.m in the folder Compusaus/Version 1 gives the results for Australia for Model 1
- lchrun.m in the folder Compusaus/Version 2 gives the results for Australia for Model 2
- ...

3. Expected computation time:

Running all countries and versions will take several weeks. Running one type takes a few hours (depending on the type and the country). 

4. Seed and random number generator:

At the beginning of each model the random number generator is reset to default, i.e., Mersenne Twister with seed 0. 

5. Description of other files:

Folder USA:
- aime.m: computes the pension capital for the US system
- dierrors.m: computes how many rejected DI applicants were truly disabled
- hedist.m: computes the health distribution presented in Table 3
- lch1.m: solves the model for type 1 (non-college, no health insurance, low disutility)
...
- lch12.m: solves the model for type 12 (college, retiree health insurance, high disutility)
- lchrun.m: sets the random number generator to default and runs all 12 types
- resdc.xlsx: contains the disability insurance incidence (for each type and aggregate) 
- resdi.xlsx: contains other statistics related to disability insurance (for each type and aggregate) 
- ressw.xlsx: contains the employment rates (for each type and aggregate) 
- stategrid.m: combines all state variables into one grid
- tau.m: computes the total tax rate on labor income  
- taup.m: computes the total tax rate on pension income 
- Tss.m: computes the pension benefit 
- txblinc.m: computes the taxable income
- w.m: gives the earnings profile
- weainc.m: computes the assets/income ratio

The folders PolFun and PolFunLS contain the necessary files to produce the policy and value functions presented in Figure A2. To do so, run valrun.m in the former and polfunlsrun.m in the latter. 

Country-specific folders: (similar to files in the USA folder with some additional files) 
- ocpenpoints.m: computes points for the occupational pension
- pencap.m/penpoints.m: compute the pension capital/points used to compute the pension benefit in the respective country
- Tdi.m: computes the disability pension benefit
- Tsscl.m: computes the pension benefit for the college type (Tss.m gives the benefit for the non-college types in those cases)
- Tssoc.m: computes occupational pension benefit
- Tssoccl.m: computes occupational pension benefit for the college type
- Tssochs.m: computes occupational pension benefit for the non-college type

General files:
- LIS.do: constructs life cycle earnings profiles using the LIS (Luxembourg Income Study) data set
- TauCoeff.m: computes the coefficients for the tax functions (labor income)
- TaupCoeff.m: computes the coefficients for the tax functions (pension income) 

Data Sources:

We used Luxembourg Income Study (LIS) data to construct the life cycle earnings profiles (see the paper for a description). The data can be accessed through the LISSY remote-execution data access system for LIS at: http://www.lisdatacenter.org. The accompanying Stata do file is called LIS.do. 

Income taxes over average income are from the OECD tax and benefit calculator, which can be accessed here: http://www.oecd.org/els/soc/benefitsandwagestax-benefitcalculator.htm. The corresponding Matlab code to fit the tax functions is called TauCoeff.m and TaupCoeff.m. 

The employment and disability incidence rates for the U.S. are based on Health and Retirement Study (HRS) data reported by Erosa et al. (2012). The corresponding values for the other countries are from the OECD Labor Force Statistics. The self-reported health measures are from the HRS. All of these data moments are reported in either the main paper or in the Appendix.