paper_cpsearn.lyx~

#LyX 2.1 created this file. For more info see http://www.lyx.org/
\lyxformat 474
\begin_document
\begin_header
\textclass article
\begin_preamble
\input{/Users/lutz/Dropbox/data/software/lyx/paper_preamble.tex}
\end_preamble
\use_default_options false
\maintain_unincluded_children false
\language english
\language_package default
\inputencoding auto
\fontencoding global
\font_roman default
\font_sans default
\font_typewriter default
\font_math auto
\font_default_family default
\use_non_tex_fonts false
\font_sc false
\font_osf false
\font_sf_scale 100
\font_tt_scale 100
\graphics default
\default_output_format default
\output_sync 0
\bibtex_command default
\index_command default
\paperfontsize 12
\spacing single
\use_hyperref true
\pdf_bookmarks true
\pdf_bookmarksnumbered false
\pdf_bookmarksopen false
\pdf_bookmarksopenlevel 1
\pdf_breaklinks true
\pdf_pdfborder true
\pdf_colorlinks true
\pdf_backref false
\pdf_pdfusetitle true
\papersize default
\use_geometry false
\use_package amsmath 2
\use_package amssymb 2
\use_package cancel 0
\use_package esint 1
\use_package mathdots 1
\use_package mathtools 0
\use_package mhchem 1
\use_package stackrel 0
\use_package stmaryrd 0
\use_package undertilde 0
\cite_engine natbib
\cite_engine_type authoryear
\biblio_style plainnat
\use_bibtopic false
\use_indices false
\paperorientation portrait
\suppress_date false
\justification true
\use_refstyle 0
\index Index
\shortcut idx
\color #008000
\end_index
\secnumdepth 3
\tocdepth 3
\paragraph_separation skip
\defskip medskip
\quotes_language english
\papercolumns 1
\papersides 1
\paperpagestyle default
\tracking_changes false
\output_changes false
\html_math_output 0
\html_css_as_file 0
\html_be_strict false
\end_header

\begin_body

\begin_layout Title
Constructing Age-earnings Profiles From CPS Data
\end_layout

\begin_layout Author
Lutz Hendricks
\begin_inset Foot
status collapsed

\begin_layout Plain Layout
University of North Carolina, Chapel Hill; lutz@lhendricks.org
\end_layout

\end_inset


\end_layout

\begin_layout Date
\begin_inset ERT
status open

\begin_layout Plain Layout


\backslash
today
\end_layout

\end_inset


\end_layout

\begin_layout Standard
\begin_inset ERT
status collapsed

\begin_layout Plain Layout


\backslash
thispagestyle{empty}
\end_layout

\end_inset

 
\end_layout

\begin_layout Abstract
This document describes the construction of age-earnings profiles from CPS
 data.
 It serves as documentation for the data used in a number of my papers.
\end_layout

\begin_layout Standard
\begin_inset Newpage newpage
\end_inset


\end_layout

\begin_layout Section
Introduction
\end_layout

\begin_layout Standard
This document describes the construction of age-earnings or age-wage profiles
 from March CPS data.
 The same procedures are used in a number of my papers.
 Documenting them in one place avoids duplication.
\end_layout

\begin_layout Standard
Details, such as the age ranges or cohorts included in the sample, differ
 across papers.
 This document therefore replaces specific values of the parameters governing
 these choices with variables.
 The values of these variables are specified in the papers that use the
 data.
\end_layout

\begin_layout Section
Sample
\end_layout

\begin_layout Standard
The data are taken from the March Current Population Survey 
\begin_inset CommandInset citation
LatexCommand citep
key "ipums_cps"

\end_inset

.
 The data waves used and the sampling criterisa (age and birth year ranges,
 etc.) vary across applications and are documented in the papers that use
 the data.
 See 
\family typewriter
import_cpsearn.filter
\family default
 and 
\family typewriter
import_cpsearn.import
\family default
 (this is a pointer to the program file that executes this step).
\end_layout

\begin_layout Section
Individual Variables
\end_layout

\begin_layout Standard
The construction of individual variables is based on 
\begin_inset CommandInset citation
LatexCommand citet
key "Bowlus:2012"

\end_inset

.
 
\end_layout

\begin_layout Standard
Dollar values are deflated using the Consumer Price Index (all items, U.S.
 city average, series Id: CUUR0000SA0; see bls.gov).
\end_layout

\begin_layout Paragraph
Schooling:
\end_layout

\begin_layout Standard
Individuals are assigned one of 4 schooling levels: high school dropout
 (HSD), high school graduate (HSG), college dropout (CD), college graduate
 (CG).
 See 
\family typewriter
var_cpsearn.school_create
\family default
.
\end_layout

\begin_layout Standard
As discussed in 
\begin_inset CommandInset citation
LatexCommand citet
key "jaeger1997"

\end_inset

, the coding of schooling changes in 1991.
 I use the coding scheme proposed in his tables 2 and 7 to recode HIGRADE
 and EDUC99 into the highest degree completed and the highest grade completed.
\end_layout

\begin_layout Standard
Prior to 1992, we have information about completed years of schooling (variable
 higrade).
 During this period, we define high school graduates as those completing
 12 years of schooling (higrade=150), college dropouts as those with less
 than four years of college (151,...,181), and college graduates as those with
 16+ years of schooling (190 and above).
 Beginning in 1992, the CPS reports education according to the highest degree
 attained (educ99).
 For this period, we define high school graduates as those with a high school
 diploma or GED (educ99=10), college dropouts as those with "some college
 no degree," "associate degree/occupational program," "associate degree/academic
 program" (11,12,13,14).
 College graduates are those with a bachelors, masters, professional, or
 doctorate degree (15,16,17,18).
\end_layout

\begin_layout Paragraph
Hours worked:
\end_layout

\begin_layout Standard
Hours worked per year are defined as the product of hours worked last week
 and weeks worked last year.
 
\end_layout

\begin_layout Standard
Weeks worked per year are intervalled until 1975 (WKSWORK2).
 For consistency, I use the intervalled variable for all years.
 Each interval is assigned the interval midpoint.
\end_layout

\begin_layout Standard
Until 1975, hours worked per week are only available for the previous week
 (HRSWORK).
 For consistency, I use this variable for all years.
\end_layout

\begin_layout Paragraph
Income variables:
\end_layout

\begin_layout Standard
Labor 
\series bold
earnings
\series default
 are defined as the sum of wage and salary incomes (INCWAGE).
 In some cases, I consider a broad measure of labor income that includes
 a fraction of self-employment income (INCBUS).
\end_layout

\begin_layout Standard

\series bold
Wages
\series default
 are defined as labor earnings divided by weeks worked.
 
\end_layout

\begin_layout Standard
I construct 2 versions of earnings and wages.
 For the 
\series bold
full sample
\series default
, I retain all observations with non-negative earnings or wages.
 For the 
\series bold
wage sample
\series default
, I drop individuals who work fewer than a minimum number of weeks or whose
 wages are outside a fixed range around the median (presumed to be outliers
 or persons not truly working for pay).
\end_layout

\begin_layout Standard
Income variables are top-coded.
 As discussed in 
\begin_inset CommandInset citation
LatexCommand citet
key "Bowlus:2012"

\end_inset

, the frequency of top-coding and the top-coded amounts vary substantially
 over time.
 In addition, top-coding flags contain obvious errors.
 In most years, fewer than 2% of labor earnings observations appear to be
 top-coded.
 Following 
\begin_inset CommandInset citation
LatexCommand citet
key "Bowlus:2012"

\end_inset

, I prefer to use median rather than mean log wages to avoid this problem.
 When using mean log wages, I follow 
\begin_inset CommandInset citation
LatexCommand citet
key "AutorKK2008"

\end_inset

 and replace top-coded incomes with 1.5 times their top-coded values.
\end_layout

\begin_layout Standard
See 
\family typewriter
var_cpsearn.wage_create
\family default
.
\end_layout

\begin_layout Paragraph
Weights:
\end_layout

\begin_layout Standard
Observations are weighted by WTSUPP which excludes the armed forces and
 the Hispanic oversample.
\end_layout

\begin_layout Paragraph
Work status:
\end_layout

\begin_layout Standard
Individuals who report working for wages (CLASSWKR) are classified as wage
 earners.
\end_layout

\begin_layout Section
Aggregate Statistics
\end_layout

\begin_layout Standard
All aggregate statistics are computed for the full sample and the wage sample.
 For each [age, school, year, sample] cell, I compute (
\family typewriter
aggr_cpsearn.age_school_year_stats
\family default
):
\end_layout

\begin_layout Enumerate
number of observations
\end_layout

\begin_layout Enumerate
mass
\end_layout

\begin_layout Enumerate
for earnings and wages: median, mean log
\end_layout

\begin_layout Enumerate
mean weeks worked
\end_layout

\begin_layout Enumerate
mean years of schooling.
\end_layout

\begin_layout Standard
I compute similar statistics (
\family typewriter
aggr_cpsearn.aggr_stats
\family default
)
\end_layout

\begin_layout Enumerate
by [schooling, year], pooling a fixed age range
\end_layout

\begin_layout Enumerate
by [age range, schooling, year].
 This is mainly for computing the college wage premium for young and old
 workers as in 
\begin_inset CommandInset citation
LatexCommand citet
key "CardLemieux2001"

\end_inset

.
\end_layout

\begin_layout Section
Cohort Lifetime Earnings
\end_layout

\begin_layout Standard
For each [schooling, cohort] cell, I compute the present value of lifetime
 earnings over a fixed age range (that depends on schooling) (
\family typewriter
profiles_cpsearn.cohort_earn_profiles
\family default
).
\end_layout

\begin_layout Standard
Since full age profiles are not observed for most cohorts, some imputation
 is needed.
 This involves the following steps:
\end_layout

\begin_layout Enumerate
Regress log median wages on time dummies and age dummies.
\end_layout

\begin_layout Enumerate
Smooth each cohort's observed age-wage profiles using a Lowess filter.
\end_layout

\begin_layout Enumerate
Linearly interpolate interior missing values (these are rare and typically
 due to small numbers of observations in some cells).
\end_layout

\begin_layout Enumerate
Impute missing observations at the start or at the end of the cohort's career
 using the estimated age-wage profile.
 This is scaled to match the mean of 5 overlapping periods.
\end_layout

\begin_layout Section
Code Documentation
\end_layout

\begin_layout Standard
Much of the code documention exists as program comments.
\end_layout

\begin_layout Standard

\family typewriter
run_all_cpsearn
\family default
 runs all routines in sequence.
 As inputs, this expects:
\end_layout

\begin_layout Enumerate
CPS data files and code to load / recode them.
\end_layout

\begin_layout Enumerate
A 
\family typewriter
filterCpsearn
\family default
 object that specifies sample selection rules.
\end_layout

\begin_layout Enumerate
A 
\family typewriter
profile
\family default
 object that specifies how cohort wage profiles are to be constructed.
\end_layout

\begin_layout Enumerate
The program directory and some general purpose code need to be on the Matlab
 path.
\end_layout

\begin_layout Standard
\begin_inset CommandInset bibtex
LatexCommand bibtex
bibfiles "/Users/lutz/Documents/econ/References"
options "econometrica"

\end_inset


\end_layout

\end_body
\end_document