This readme file accompanies the paper, ?Asset maintenance as hidden investment among the poor and rich: Application to housing,? by Hernandez, M.A. and Trupkin, D.R. This file details the steps to generate the theoretical and empirical results reported in the paper.

-------------------------------------------

I. Theoretical section 

The theoretical analyses was all performed in Matlab version R2017a. 

The following accompanying Matlab codes should be used to reproduce the results presented in the theoretical section of the paper by computing the corresponding elasticities to generate Figures 1 through 3: 

1. 'run_figures.m' -- Main code file to run the simulations.

2. 'solve_model.m' -- Code file called in the process. 

All the results in the workspace are saved as 'results_dd-mmm-yyyy_HHMM.mat' file.

The total expected computation time may take around 12 hours. 

The following specific steps should be followed for the computation of the results corresponding to each figure:

A. To reproduce the results of Figure 1 (estimated computation time: 15 minutes):

1. Open 'run_figures.m' and set "figure=1".
2. The set of parameter values are equal to the benchmark parametrization of the paper.
3. The plot in Panel A is generated by setting "permanent=1" in the main code. 
4. The plots in Panels B and C for transitory income elasticities are generated by switching the "permanent" command to any other value. 
5. The main results are stored in vectors:

(i) "Maint_elasticity": Income elasticity of maintenance.
(ii) "Cons_elasticity": Income elasticity of consumption.


B. To generate the results needed to reproduce Figure 2 (estimated computation time: 5 hours):

1. Open 'run_figures.m' and set "figure=2".
2. The set of parameter values are equal to the benchmark parametrization of the paper.
3. Change 'minvalue', 'maxvalue' and 'stepvalue', as in Figure 2, for the corresponding range of the moving-cost parameter "phi"  (1 through 5, with steps of 0.5).
4. Set a wide income range with small steps (suggested range: 1.6*h0--3.6*h0; with 0.001 income steps). 
5. Additionally, we suggest to adapt the solution revision parameter "zeta0" to 0.01 (or lower) in the 'solve_model' file, in case the code does not converge for some ranges of the parameter values. 
6. The main results are stored in:

(i) "Maint_elasticity": Income elasticity of maintenance; MxN matrix, where M is the dimension specified for the range in "phi" and N is the income level range.
(ii) "phi_vector": M-dimensional vector according to the range specified for the moving cost.

7. Finally, compute average income elasticities for the bottom and top tertile of the simulated incomes, for each value of "phi".


C. To generate the results needed to reproduce Figure 3 (estimated computation time: 7 hours):

1. Open 'run_figures.m' and set "figure=3".
2. The set of parameter values are equal to the benchmark parametrization of the paper.
3. Change 'minvalue', 'maxvalue' and 'stepvalue', as in Figure 3, for the corresponding range of the borrowing parameter "kappa"  (0.1 through 0.5, with steps of 0.05).
4. Set a wide income range with small steps (suggested range: 1.6*h0--3.6*h0; with income steps of 0.005). 
5. Additionally, we suggest to adapt the solution revision parameter "zeta0" to 0.01 (or lower) in the 'solve_model' file, in case the code does not converge for some ranges of the parameter values. 
6. The main results are stored in:

(i) "Maint_elasticity": Income elasticity of maintenance; MxN matrix, where M is the dimension specified for the range in "kappa" and N is the income level range.
(ii) "kappa_vector": M-dimensional vector according to the range specified on the borrowing parameter.

7. Finally, compute average income elasticities for the bottom and top tertile of the simulated incomes, for each value of "kappa".


-------------------------------------------

II. Empirical section

The empirical analyses was all performed in Stata version 15.1. 

To replicate the empirical results reported in the paper (Tables 1 through 4), open the following program (do file) and accompanying dataset: 

1. empirical_results_maintenance.do -- Commented do file with all steps and command lines that generate the tables reported in the paper.

2. empirical_data_maintenance.dta -- Dataset at the unit-household-year level for owner-occupied units obtained from the American Household Survey (AHS), period 1997-2011, published by U.S. Census Bureau and the Department of Housing and Urban Development. All variables in the dataset are labeled.

The total expected computation time is 1 minute.