The following instructions explain how to use the replication programs for Minchul Yum (2018) ‘On the distribution of wealth and employment’.

Enclosed with these instructions are the two subfolders.
1. code 
2. data

Folder 1: CODE

This sub-folder contains simplified programs that can be sufficient to solve the model numerically in the paper, which will allow users to replicate results in the paper. The programs are written in Matlab based on Windows 10 OS system, and have been tested with many recent versions since 2015a. The program requires the parallel toolbox to execute them faster.

1) Mainsolve.m solves the incomplete markets model, and returns the value functions and equilibrium prices that are necessary for computing all results based on the model. This file should be executed first.

The following files 2)-7) are required to run 1).

2) GSS.m performs Golden section search for optimization. 
3) rouwenhorst.m returns the discretized Markov process using Rouwenhorst method.
4) markovss.m finds the stationary distribution given Markov transition matrix.
5) nltr.m contains the income-dependent nonlinear transfer schedule.
6) Tmatinv.m computes the T matrix required for evaluating cubic spline with knot-a-knot end point conditions
7) csieval.m evaluates cubic spline given T matrix computed using 6).

When market-clearing is achieved in the outer loop, Mainsolve.m is complete. In the meantime, the inner-loop performs value function iterations and it stores distribution on the grids. To give some estimates of total computation time, it takes roughly 30 minutes using na = 100 and nk = 1000 and 4 hours using na = 200 and nk = 2000 on the personal computer with octa-core AMD Ryzen 7.

8) statsbywealth5.m computes statistics by wealth quintile. This file must be executed after 1) is completed. It returns the main statistics, employment rates by wealth quintiles, shown in Figures 1 and 2. 

Note that the specifications other than the baseline specification are simply nested specifications. Hence, one can easily impose the parameter restrictions manually. The above files are sufficient to replicate Tables 3-4, 8 and Figures 2-5. 

However, the other figures and tables using the model require simulation. To do so, the following file should be run after completing 1). For random number generation, the algorithm of Mersenne twister is used with the seed number 820406. It has been tested that the results do not change much with other seed numbers and other algorithms. The execution time for simulation is only several seconds.

9) Simul.m simulates the model economy using the equilibrium output from 1) such as decision rules and equilibrium prices.

Folder 2: DATA

This sub-folder contains the data sets that can be used to replicate results in the paper. The program used for the analysis was Stata 15.1. There are two data files associated with two do-files. Each can be run separately.

1) scf_code.do first loads scf_data.dta containing the 1992, 1995, 1998, 2001, 2004, and 2007 waves of the Survey of Consumer Finances, which can be freely obtained at https://www.federalreserve.gov/econres/scf-previous-surveys.htm. The file then creates key variables such as employment based on data on labor supply, and imposes sample selections. The program then generates statistics reported in Figure 1 and Tables 1,3,4. Relevant explanations can be also found in the program along with the commands.

2) sipp_code.do first loads sipp_data.dta containing the first to ninth waves of the 2001 Survey of Income and Program Participation, which can be freely obtained at https://www.census.gov/programs-surveys/sipp/data.2001.html. Both core and topical module data are included. The file then constructs key variables such as total transfers and employment. The program then generates the statistics by income quintile reported in Tables 3 and 5. Next, the program conducts several regressions, which generate the results reported in Table 7. Relevant explanations can be also found in the program along with the commands.