Graciela Sanroman, "Cost and Preference Heterogeneity in Risky
Financial Markets", Journal of Applied Econometrics, Vol. 30, No. 2,
2015, pp. 313-332.

The file GSG_Internet_Appendix_20130305.pdf contains additional tables
and figures, including descriptive statistics for the whole dataset,
information about sample sizes, data about the Italian financial asset
returns, and the estimation of parameters using a narrower definition
for the risky asset. This PDF file should be readable using Acrobat
Reader or any other program for reading PDF files.

The data employed in the paper come from Historial Database of the
Survey of Household Income and Wealth of the Bank of Italy. 

The ASCII file datafile_gsg.csv contains the dataset used for the
Indirect Inference estimation. This file is zipped in the file
gs-data.zip. Unix/Linux users should use "unzip -a".

The data used to estimate structural parameters include 6377
observations. The first row of the file contains the names of the
variables including: 

id		     	Household ID
year	     		Year of the wave
age		     	Age of the Household Head
educ         		Education of the Household Head
					2= "Elementary School" 
					3= "Secondary School" 
					4= "High School"
					5= "College or Postgraduated"
blue_collar 		Occupation of the Household Head (1=Blue_collar
                        0=White_collar)		
south			Residence in South or Islands (1=Yes, 0=No)
house_owner		Owner of Principal residence  (1=Yes, 0=No)
lincome			Labor income in logs (euros, 2010 ppp)
fin_wealth		Financial Assets holdings (euros, 2010 ppp)
gov_securities		Goverment Securities holdings (euros, 2010 ppp)
other_securities	Other Securities holdings (euros, 2010 ppp)
I_ra			Own Risky Assets  (1=Yes, 0=No)	


The following files (zipped in gs-codes.zip) provide all the routines
used in the paper:

- do_eachwavedata_20120725.do (STATA code, recovers the information
  from the original *.csv files, available on the Bank of Italy web
  page, and creates one dataset for each wave)

- do_paneldata_cpi_20120725.do (STATA code, creates the panel data set)

- program_20120725.do.  This STATA code 

  * a) estimates the parameters of the labor income process, 
  * b) produces tables 1,2, 3 and Figures 1,2,3,
  * c) exports data and other inputs to the Fortran program.

- Mainprogram.EDU5.RES1.f90. This Fortran code includes:

  a- Numerical solution to the life cycle problem of consumption and
     investment in the risky asset market for given values of: the
     discount rate, expectation on asset returns, deterministic and
     stochastic parameters of labor income over a double dimensional
     grid of: the CRRA and fixed participation costs over a
     three-dimensional grid for the following variables: ratio of
     financial wealth to permanent income, permanent income and
     transitory income

  b- Evaluation of each individual contribution to the
     pseudo-likelihood and the pseudo-likelihood of different
     subsamples This code is configurated to the "College" group and
     uses specification 1 for returns, but it can be easily changed to
     other educational group or specifications.

- table4.do. This STATA code includes the routines to obtain figures of
  Table 4. A similar procedure is used to obtain Tables 5,6 and 7. It

  * loads files pseudo1_loglik`type' and pseudo_loglik`type'_IND_`type'
    produced by the Fortran code Mainprogram.EDU5.RES1.f90

  * finds values of g_coef (participation costs) and gamma (the CRRA)
    that maximize the PLL and compute the corresponding standard errors

The following files (zipped in gs-inputs.zip) are the data inputs used
by the Fortran program.

- Beta.txt - columns: 1-coefficient of age of (log) labor income process 
                      2-coefficient of age^2/10 of (log) labor income process
                      3-the constant
           - rows:    educational groups

- MINMAXY.txt - columns: 1. min of permanent income grid
                         2. max of permanent income grid
			 3. max of transitory income grid, 
              - rows:    educational groups

- SHOCKSRENTA.txt - columns 1. Variance of permanent shocks 
                            2. variance of transitory shocks
                  - rows:    educational groups

- Gauss-HermiteT and Gauss-HermiteP: - Values and probabilities for 3 and 5
                                       points Gauss-Hermite quadratures
				       (corresponding to a N(0,1)) 

- RETORNOS_ESP`XX' xx=1,2,3,4 Values and probabilities for financial assets
               returns (the first column corresponds to the riskless asset)
		   
  1. SPECIFICATION 1 IS 3 POINT GAUSS-HERMITE FOR MEAN AND VARIANCE
     1991-2011
 
  2. SPECIFICATION 2 IS 3 POINT GAUSS-HERMITE FOR Pelizzon & Weber data
     on returns
  
  3. SPECIFICATION 3 IS 5 POINT GAUSS-HERMITE FOR MEAN AND VARIANCE
     1991-2011
  
  4. SPECIFICATION 4 IS 5 POINTS FOR MEAN, VARIANCE, SKEWNESS AND
     KURTOSIS 1991-2011


- DATA_INC_A_`edu'.txt (edu=educational group) data on (real
numbers) variables about individuals characteristics including 

  1. coh1: ratio consumption(t)+financial asset at the end of t to
     labor income(t)
 
  2. coh2: ratio labor income(t) + financial asset at the end of t to
     labor income(t)

  3. ratio r1: etai/mean(Y by age2) where etai is the average income of
     i in the panel

  4. ratio r2: yi(t)/etai

  5. alpha(t)

  6. alpha(t-1) (-99 if missing)

- DATA_INC_B_`edu.txt' (edu=educational group) data on (integer numbers)
              variables about individuals characteristics including

  1. id 

  2. time
  
  3. age
  
  4. SOUTH: Regional DUMMY =1 IF SOUTH OR ISLANDS
  
  5. Doccup: OCCUPATION DUMMY =1 if BLUE-COLLAR =0 IF WHITE-COLLAR
  
  6. Young (1= younger than 45 years)
  
  7. NoHouse (1= No house owner)
  
  8. Dw (1= above median of the ratio of financial wealth to permanent
     income ratio
  
  9. Dr1 (1 = above median income)


Please address any questions to:

gsanroman [AT] decon.edu.uy

PhD. Graciela Sanroman
Constituyente 1502 Piso 6
11200 Montevideo, Uruguay
Tel: +59824106449 ext. 680