Ricardo Mora, "A Nonparametric Decomposition of the Mexican American 
Average Wage Gap", Journal of Applied Econometrics, Vol. 23, No. 4, 
2008, pp. 463-485.

There are three zip files. The file <rm-stata-dta.zip> contains 
<datos01.dta>. The file <rm-data.zip> contains <datos01.txt>. The file 
<rm-macros.zip> contains a large number of Stata .do files. Note that 
<datos01.txt> and the .do files are ASCII files in DOS format.

The original source of the data corresponds to extracts of the Merged
Outgoing Rotation Groups file of the Current Population Survey (CPS)
prepared by the NBER for the 1994:2002 period. The data used in the
paper is obtained from the original morg files in STATA format after
running <datos00.do> and <datos01.do>, two batch files for STATA
included in <rm-macros.zip>. The result is <datos01.dta>, a file dataset 
in STATA format.

The data used in the paper can also be directly read from <datos01.txt>, 
a plain (ASCII) text file included in <rm-data.zip>. Each line is an 
observation. Variables are stored in columns in fixed format. Missing 
variables (wages for non-participants) are represented by ".".

The data set contains 75,949 observations and 27 variables. Here is a
brief description of the variables. The first variable corresponds to
the first column in <datos01.txt>, and so on.
(see also notes in Table 2 in the paper)

periodo     Month of interview; 1: Jan 1994.
state       State (74: Texas, 85: New Mexico, 86: Arizona, 93: California)
year        year of interview
salarios    log-hourly wages
age         years of age
age2        age*age
exp         years of potential experience
exp2        exp*exp
agexed      age*educa
expxed      exp*educa
educa       years of education
voca        dummy for vocational studies
etnia       ethnic status (0 non-hispanic 1: mexame)
veterano    veteran status
marital     marital status
calif       state=93
arizo       state=86
texas       state=74
newme       state=85
wagesmpl    dummy for participation in work-for-pay market
doccphis    percentage of Hispanic population in occupation
dindphis    percentage of Hispanic population in industry
hisppor     percentage of hispanics in local area
educcon     spouse's years of education if marital=1
agecon      spouse's age if marital=1
agepad      father's age
agemad         mother's age

(variable "Parents" can be obtained as marital=0 and (agepad>0 and/or
agemad>0))

In addition, the following STATA batch files are included in 
<rm-macros.zip> to facilitate the replication of the paper.

<table02.do> gives the summary statistics from Table 2;

<partic01.do> obtains results from the parametric participation models
and prepares the data sets for the participation models with tree
structures;

<partic02.do> computes for the tree participation models the
likelihoods, betas, inverse mill-s ratios, the residuals, and tests for
normality (the program calls <parall1.do>, <parall2.do> and <parall3.do>)

<partic03.do> computes the marginal effects for the participation models;

<salari01.do>, <salari02.do>, and <salari03.do> prepare the wage
equations data (line 70 calls an external programm to transfer the
dataset from a STATA format into a GAUSS format)

<regre01.do> estimates the parametric wage equations in STATA

<regre02.do> takes the nonparametric wage structure and computes
marginal effects

<salari04.do> presents decompositions from wage equations results  

The estimation of the tree structures was performed with a gauss macro
available from the author upon request.

Ricardo Mora