NLS
Discussion
Papers
.
.
.
The Effects of Unemployment
on the Unemployment
Compensation
of Youths
.
A. Colin Cameron
University of California at Davis, and
Economics Research Center/NORC
R. Mark Gtitz
University of Washington, and
Economics Research Center/NORC
Thomas MaCurdy
Stanford University, and
Economics Research Center/NORC
September 1989
~s find paper was funded by U.S. Depment of Labor, Bureau of Labor Statistiti under
contmct nusnber J-9-J-7-0092. Opinions smted in MIs paper do not necessafly representtie
officifl position or policy of me U. S. Depament of Lahr.
FINAL ~PoRT
DOL/BLS #J-9-J-7-0092
WE EFFEcTs OF UNE~W~NT
COMPENSATION
ON TEE .~LO=NT
OF YOUTHS
by
A. Colin Cmeronl
R. Mark Gritz2
~omas MaCurdy3
September
1989
lAss is tant Professor;
Department
Davis , CA and Research
Associate,
IL
of
California,
University
of Economics,
Economics Rese=ch
.Center/NORC,
Chicago,
2A~~is ~ant
of
Seattle,
IL
3F=ofe~~or,
Institution,
Economics
professor,
WA and Rese=ch
Department
Department
Associate,
Economics,
Economics
Univers itY Of Washington,
Resezch
Center flORC, Chicago,
Of Economics
Stanford
University,
Research
Center/NORC,
and Senior FellOw, me HOOver
Stanford
CA, and Research Associate,
Chicago,
IL
Department
~is
project
was funded by the U.S.
Statistics
under
Contract
Nwber
#J-9-J-7-0092.
doc-ent
& not necessarily
represent
the official
U.S. Department of Labor.
of ~bor
Opinions
position
Bureau of bbor
stated
in this
or policy
of the
CONTENTS
Page
Section
Executive
I
Summary
Introduction
. . . . . . .
. . . . . . . . . . .
Linking UIEntitlements
3
Data Requirements
2.2
Data Sources Used inthe
2.3
Features of the YNLS
A Description
5
6
. . .
. .
. . .
1-111
. . . . . . . . . . . . . . . . . . . . ..1
and Unemployment
2.1
Experiences
4
. .
Experiences
to Imput UIEligibility
.
md Benefits’
Previous Literature
.
. .
. . . . .
.
. . . . . . . . . . . . . . . . . . . . . .
.-
5
.
5
. .
7
. ..8
of the Earrrings and Employment
of~ouths
. . . . . .
. .
Sample Compositions
3.2
Measures of Annual Earnings
3.3
Reliability
3.4
Characteristics
of Weekly md Hourly Earnings
17
3.5
Characteristics
of Employment
Activities
20
and Use Among Young Workers
22
Me=ures
of Eligibility
4.2
Me=ures
of Utilization
4.3
A Data Set Integrating
4.4
Patterns of UI Eligibility.
4.5
Patterns of UI Use.....
4.6
Cornpmison
AGenerd
.
of Earnings Me=ures
4.1
Statistics
.
.“
. ..11
3.1
UI Eligibility
and Descriptive
.
.
.
.
11
. . .
.
.
. .
13
.
14
. . .
..22
.24
UI Eligibility
and Utilization
. .
Approach..
. .
.
.
.
.
27
. ..28
.
29
. . . . . . . . . . . . . . . . . . . .
32
Alternative
5.2
Assessing the In6uence of UI on Unemployment
34
5.3
Me=uting
36
5.4
An Alternative
the Imp~t
S&emes
.
.“.
5.1
A Spedfication
Estimation
25
. . .
with Findings in the Literature
Estimation
.
. . .
. .
of Shifts in UI Pohcy
Formulation
Characterizing
.
. . . .
,
A Sample Liting
38
the Duration of
UnemploymentB etweenJobs
6.1
32
40
UI Entitlements
1
and Unemployment
Duritions
41
?
8
6.2
Defining Variables in the Empirical Specifications
6.3
Repr-entative
Cases
The Influence of UI Programs
Duration Distributions
7.2
Exploratory
Data Analysis
7.3
An Empirid
Specification
7.4
Estimation
7.5
Imprecations of the Empirical Findings
Results
Estimation
10
Imphcations
The hpact
10.1
11
Comparing
Comparing
.
49
.
50
55
S7
. . . . . . . . .
57
. . . . . . . . .. . .
58
. . . . . . . . . .
61
of
.
Classification
a Specification
.
.
Results
. . . . . . .
. . . .
for the Recipiency
of the Empirid
.
Popdations
6.4
. . .
64
. . . . . . . . .
65
.
6i
. . . . . . . . . . . . . . . . . . . . . . .
67
Unemployment
11.2
Comparison
11.3
Pokey Impkmtions
Durations
Across Pohcy Regimes
.
. .
68
. . . .
70
. . . . . . . . . . . . . . . . . . . ..70
with Results in the L]teratue
. . . . . . . . . . . .
71
. . . . . . . . . . . . . . . . . . . . . . . ...74
Construction
of a Data Set from the YNLS Describing
Earnings md Employment
Al
.
Durations for UI and non-UI
F]ndings and Closing Remarks
Summary of the Findings.
62
. . . . .
Distribution
Resdts
.
. . .
Unemployment
ll.l
A;
. .
. . . . . .
Distribution
md Recipiency
A Synthesis of the Empiticd
APPENDIX
.
of UI Policies on the Duration of Unemployment
Recipient
10.2
48
Results for the Division of the
UI Entitlements
9.2
: . . . .
. . . . . . . . . . . .
Proportions
Proportions
ImpHcaticms of the Empiticd
Estimating
.’.
Results for the Broad Classification
Some Unemployment
9.1
.. .
48
. . . . . . . . . . . . . . . . . . . . . . . . ..53
8.2
Relating
.
. . . .
for Spdl Lengths in Nonemployrnent
Specifying, a Time-Proportion
8.4
45
. . .... .. ..,. . . . : . . .
8.1
Estimation
. .... . . . . . . . . . . .
and Survivor Functions
The Effects of UI on Unemployment
8.3
42
on Nonemployment
7.1
Unemployment
9
.
. . . . .
General Considerations
A.2 Work History Data
Experiences
. .
... . . . .
.
. . . . . . .
. . . . .. . .
, . .
. _ . . . .
76
76
i7
._,
.—
.
A.2.1 Data on each job in each week
A.2.2 WeeMy emnings
A.3 Annual Reported
A.4 School Attendance
A.5 Variables
APPENDIX
B:
Construction
. . . . . . . . . . . . . . . . .
Earnings
of Missing Emnings
Sdection
.
“.”
. .
. . . . .
86
.
Level
.
. .
87
.
88
. . .
95
.
. .
. . . . .
98
98
.
100
. . . .. . . . . . . . . . . . .
101
. . . .
of Work History Variables
of Imputed
md Entitlements
98
. . .
. .
of UI Entitlements
B.3 Construction
References
.
84
of a Data Set Linking UI Entitlements
B.2 Imputation
B.4 Accuracy
.
md Education
and Unemployment
B..l Smple
earnings and work
and Samples Used in Section 3
A.6 Imputation
78
83
md work =perienms
A.2.3 Annual mmputed
experiences
.
. . . . .. . . . . . . . . . . .
. . .
Measures of UI Eligibility
. . . . . . . . . . . . . .
104
TABLES
Page
Table
3.1-M
Summary
Statistics for Measures of Annual Earnings (Men)”
3.1-W
Summary
Statistics for Measures of Annual Emnings [Women)
3.2-M
Me=urement
(Men)
Me=urement
Emnings
. . . . . . .. . . . . . . . . . . . . . . . . . . . . ..l6a
(Women)
. . . . . . . . . . . . . . . . . . . . . . . . . ..16b
3.3-M
Summary
Statistics for Me=ures
3.3-W
Summary
Statistics for Measures of Weekly Eunings
3.4-M
Summary
Statistics for Mmsures of Hourly Etinings
3.4-M’
Summary
Statistics for Me=ures
3.5-M
Summa~Statistics
for Weeks Worked ?nd JObs pe!Xear(Meq).
3.5-M’
Summmy
Statistic
for Weeks Worked arrd Jobs Per Year (Women)
3.6-M
Summary
Statistics for Annual and Weekly Hours of Work (Men)
21.
3.6-M’
Summary
Statistics for Annual and Weekly Horrrsof
21b
4.1-M
Meas_ures of Unemployment
4.1-W
Categories
Me.wures of Unemployment
Age-Education
Categories
of Weekly Eartings
(Men)
.
18a
(Women)
18b
(Men)
19a
of Hourly Earnings (Women)
..19b
.. 20a
20b
Work (Women)
Insurance Eligibility “by
(Men)
.
.
. .
27a
Insurance Eligibility by
(Women)
. . . .. . . .
. 27b
4.2-M
Measures of Unemploymmt
Insurance Eligibility by Year (Men)
27c
4.2-W
Measures of Unemployment
Insuruce
27d
4.3-M
Measures of Unemployment
Insurance Utilization by
Age-Education
4.3-W
Categories
Measures of Unemployment
Age-Education
Categories
EhgibiIityby
Year (W’omen)
. . .
(Men)
Insurance Utibzation
(Women)
.
. .
.
28a
by
. . .
.
. . . . . 28b
4.4-M
Memures of Unemployment
hsurmci"Utifization
by Yem (Men).
28c
4.4–w
Memmesof
Insurance Utfization
by Year (Women)
28d
6.1-M
Summary
Spds
.
14b
Error Variance Estimates for “Log Annual
Age-Education
..-
14a”
Error Variance Estimates for Log “Anngd
Etitings
3.2-W
“.
6.1-W
Unemployment
Statistics of Demographics
and Nonemployrnent
for Men . . . ....... . . . . . . . . . . ... . .. . . . . . . . . . ...41a
Summary Statistics of Demographics
Spdls for Women
arid Nonemployment
. . . . . . . . . . . . . . . . . . . . . . . . ..41b
6.2-M
Summary
Statistics of Work History and WEntitlements
for Men
41c
6.2-W
Summmy
Statistics of Work History and~
for Women
41d
iv
Entitlements
7.1-M
P=.meter
Estimates of NOnemplOyment Duration PrObablEties (Men)
7.1-W
Parameter Estimates of Nonemployment
8.i-M
P=ameter
Estimates of Tme
DuratiOn”TrObabiEties (Women)
Proportion
of No, Some, and All Unemployment
8.1-W
Pmameter
8.2-M
Par=eter
Parmeter
(Women)
. . . . . . 60a
.
;
. . . 60c
PrObabfities
. . . . . . . . . . . . . . . . . . ....62a
Estimates of Time Proportion
OfktefiOr
.
Probabilities
Estimates of Time Proportion
Of InteriOr CeHs (Men)...
8.2-W
(Men)
All Unemploymat
CWs(WOmen)
PrObabiEties
. . . . . . . . . . . . . . . . . . . . --.62c
8.3-M
Predictions
of Time Proportion
Probabilities
(Men)
8.3-W
PredictiOm
of Time Proportion
Probablhties
(Women)
9.1-M
P=meter
Estimates of the UI Receipt Probablhty
(Men)
9.1-W
Parmeter
Estimates of the UI Receipt Probabihty
(Women)
9.2-M
Predictions
of the Probabihty
9.2-W
Predictions
of the PrObabihty of UI Receipt (Women)
1O.I-M
1O.1–W
of UI Rectipt
Pre&ctiOns of the Distribution
by Recipiency
Predictions
. . . . . . . . . . 62e
. .
. . . 62g
. .
. . 65a
(Men)
. .
. .
. . 65b
. .
. . . . .
Status (Men)
.
. . . . . . . -–
of Weeks of Unemployment
Status (WOmen)
. .
. . . . . . . . . . . . . 67b
Predictions
of the Distribution
of Weeks of Unemployment
(Mm)
1O.2–W
Predictions
of the Distribution
of Weeks of Unemployment
(Women)
A.1-M
Sample Sizes (Men)
A.1-W
SamPle Sizes (Women)
A.2
Definitions
A.3
Samples used fOr Tables 3.1-3.6
B.1
Effect of S-pie
B.2
Bracket Definitions of AWE,
B.3-M
Defition
of Work H:story Controls for Men
B.4
Frequency
Table of Imputed Eh@bihty
B.5
Sumuy
WBA
65d
. .. . . 67a
1O.2-M
. . . 69a
. 69b
. . . . . . . . . . . . . . ... . . . . ...----91a
Of Samples
BOth Petitions
65c
of We&s of Unemployment
of the Distribution
by Recipiency
54b
Probabilities
Estimates of Time Proportion
of No, Some, ad
54a
. . . . . .
. . . . . .. . . . ..-
. . . ..glb
. . . . . . . . . . . . . . . . . . .. . . . . ..
Selection Crittia
HQE
Of Ehgibifity
. . . . . . . . . . . . . . . . . . . .
on Sample Size
and BPE
.
. . . . . . . . .. . - -
. . . . . .
100a
. .
10la
. . . . . . . .
102a
md ~
Rec&pt for
.
.
of Ehgiblbty
93
. . . . . . . . . . 98a
Statistics for the Difference between Reported
by Yem for Broad Definition
g2
and Imputed
. .. . . . .
103.
.
v
FIGURES
Figure
i.1-h!
.
Pag?
Empirical Hazard Rates for NorremploymeI,t
Status (Men),
7.1-W
Empirical
, .. . . . . . . . . .
7.2-M
Empirical
Hward
Empiriid
Huard
Suvivor
50a
Rate for Weeks of UI Receipt During a
. . . . . . . . . . . .. . .. . . . . . . . . ..50b
. . . . . . . . . . . . . . . . . . .. . . . . . . . . . .55a
. . . . . . . . . . . . . . . . . . . .. . . . . . . ...55a
. . . . . . . . . . . . . . . . . . . . . . . . .. . . . .
55b
Functions for Work History Hm Under Various
Regimes (Women)
7. 5-M
. . . . . . . . . . . .. .. . . . . ..
Survivor Functions for Work History Hm Under Various
Regimes (Men)
7.4-iv
During a
Survivor Functions for Work History Hz Under Various
Regimw(Womm)
7.4–M
49b
Survivor Functions for Work History Ht Under .Various
Regimes (Men)
7.3-IV
.,,....
49a
Spells by UI’
Rate for Weeks of. U] Receipt
Benefit Year(Womm)
7.3-M
. . . ..”
. . . . . . . . . . . . . . . . . . . . . . . ..... . .
Benefit Year (Men),
7.2-W
. . . . . . . . . ..”
Hazard Rates for Norremployment
Status (Women)
Spells by UI
. . . . . . . . . . . .. . . . . . . . . . . . . . ..55b
Survivor Functions for Work History Hk U rider Various
Regimes (Men) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55c
vi
ACKNOWLEDGEMENTS
This project
Washington
was funded by the U.S. Depmtment
D.C. 20210; under contract
ument do not necessarily
Labor.
.
number J-9-J-7-0092.
UI benefits whid
fuUy a&nowledge
Opinions
stated in this doc-
represent the official position or policy of the U.S. Department
We thank David Card for generously
extended
of Labor, Bureau of Labor Statistics,
from the U.S. Department
on dl aspects of the project.
providing us with a data set that he collected on
we utifize in constructing
the contributions
of
some’ of our variables.
We dso grate-
of Cindy Ambler who provided us with unpublished
of Labor and of Bart Hamilton
who provided
apert
data
assistance
Executive
Summary
Objectivm
This report examines
of unemployment
on determining
the role of unemployment
that youths
experience
how the weekly benefit
insurmce
between
mounts
jobs.
the fraction
that anindividud
veloping
apicture
exploiting
mary
objective,
The first of th=e
by UI
length oftime
spent
as unemployment;
and the Ekelihood
episode.
by the Youth
Cohort
involves the computation
work and earnings
experienc=
gods
concerned
of. youths,
with de-
of UI programs
of National
Longitudind
of a comprehensive
and the second
sum-
consists
Of
are ehgible for UI and the degree to which they draw.
on UI entitlements.
The aim is to identify
across demographic
&aracteristics;
covered
offered
in the labor mmket and utilization
gods
the extent to which youths
the analysis focuses
thetotd
we pursue two intermediate
of youths’ pmticipation
of the weeMy
=sessing
of this time reported
the rich source of data provided
Survey (YNLS).
activities:
collects UI during a nonemployment
En route to this primmy
Specifically,
the amount
and the weeks of efigibifity
progrmns influence three aspects ofnonemployment
in nonemployment;
(UI)pqlici:sp_:
two sets of patterns:
and those capturing
chang=
those describing
differences
over the period
1979–1984
by the data.
hIethodology
The mrdysis
dependable
constructs
measures of their UI eligibihty,
on a random s-pie
and nonemployment,
data on profiles of we~y
=d
force ad
unemployment
residency
and UI-program
classifications.
CPS-type
the w=kly
me=ures
atilable
benefit mount
upon leafing
mines the fiects
nonemployment
UI programs
time between
out-of-the-labor-
with supplementary
data on State of
on the lagths
benefit mount
of UI poficies
complement
with the resulting
of UI entitlements
spds.
.
ml
durations
model that jointly
which in combination
One component
of nrmemployment
of unemployment
m econometric
of behavior
(i.e. to both
md to UI co~ection.
on the distribution
this study devdops
of in&viduds.
wsociated
data set is unique in that no other source relates
to the fti
of UI cm three =pects
activities
statuses,
on episodes of both em.ploymezlt
with the chmacteristics
and to weeks of ehgibllity)
jobs,
to
to each person at the time of every job separation;
The constructed
of unemployment
=perienc-
labor-market
rules of that State, the analysis infers the wee~y
To assess the influence
that occur
of each person’s
emnings,
k conjunction
md it furthm combines this information
episode.
unemployment
and use. The YNLS offers information
on the division of nonemployment
the weeks of UI &gibility
nonemployment
benefits
of youths with detailed bistori=
alongwith considerable
md
a data set that finks individuds’
chmacterize
of the model describes
A second evaluates
deterthe
the role of
the effects of UI
on the classification
of these spe~s as unemployment.
between UI recipients
the generosity
and non-recipients,
of UI progr~s
In specifying
a third component
influences
these components,
the IikeEhood
to UI entitlements
of UI programs
that individuals
in a nonuniform
Further, the analysis t~es
coUect UI benefits.
m~er
of UI benefits rmd
varying according
great care to avoid bimes ii wtimating
by ensuring that vmiation
rather thm movements
for distinctions
of the model analyzes whether
this study accounts for dl the dimensions
allows these benefits to affect un~ployment
ration length.
Finally, to account
in benefits reflect differences
du-
raponses
in the generosity
along U] scheddes.
Findings
For men, the empirical
coUects UI typicdy
haustion
experiences
of UI benefits,
nonrecipient.
resrdts presented
in this study indicate
that an individud
a Iongm spell of nonemployment,
md reports a larger fraction
In total, UI recipients
who
at least up to the W-
of this spell as unemployment
report more weeks of unemployment
bdore
than a
returing
to
jobs.
Regarding
individuals’
the influence of UI entitlements
activities
benefit =ount
fraction
h=
through
essentially
only the population
this po~cy
spd
Concerning
the resdts
of men, these benefits alter
the effect of a rise in the weeMy
show slight incre~es
fisted as unemployment;
irrespective
of whether
shift induces only a relatively
case for an incre=e
in wee~y
of W=MY benefits,
=
benefit
extension
of unemploym~t
md for the popdation
at lWge.
or unemployment,
the highest dorations
behg
m extention
of &gibifity
of w=ks
much as 8 WAS
The findings
benefits
speUs or on the number
one considers
duration
b
the popdation
the weeks of &gibiiity
amounts.
However,
that occurs
Ttis =tension
of
at large or
jobs
does not influence
individud,
ody
both
as is the
to the effects
nonemployment
both for UI recipients
short durations
of either
of the longer durations
stretched out the most. In partictiar,
for the median
of recipiency,
lengthens
between
from 26 to 39 generates
by a program,
in sharp contrast
but it leads to an expansion
for those persons experiencing
summarized
off=ed
minor rise in the hke~ood
of weeks of UI ehgibifity
speUs and the amount
of unemployment
and in the
of UI recipients.
to the effects of an increme
nonemployment
k recipiency
but this rise in weMy
no effect on either the length of nonemployment
weeks of unemployment,
Turting
several rout=.
ptid by a program,
of a nonemployment
on the =periences
the findings indicate
about a 1 w=k
but unemployment
tith
that
lengthening
lengthens
by m
the longer durations.
above for young men dso
two receptions.
First, while female UI recipients mperience
cipients at le=t
up to the point of benefit Aaustion,
apply for young
women
more unemployment
there is some -bigtity
with onIy
thm nonreas to whether
~
a similar rdationship
tists
for women when comparing
lengths of. non:mployment
Second, the weekly benefit amount is not a factor at dl in influencing
In contrast to men, chmges
ployment
spe~ reported
co~ects WI benfits.
spells.
women’s experiences.
in weekly benefits have no effect on the fraction of a nonem-
as unemployment,
nor do they affect the likelihood
that a women
Whereas total WI benefits serve as the primary measure of WI entitle-
ments determining
~
recipiency
weeks of eligibility matter.
status for men, the result: for women indicate
Other than these two relatively minor exceptions,
of WI poficies on women’s experiences between jobs in nonemployment
that only
the influences
md in unemployment
follow the same pattern as those orrtfined above for men, rdthrmgh the magnitudes
of the
various effects differ.
Implications
The findings of this report suggest several implications
on the =ount
of unemployment.
of WI programs
employment,
At the most basic levd, the resdts
indicate that features
features that alter the amout
for those individuals
in the m=mum
In contr=t,
ptid by a program
ben~t
to have no
programs can be
with a more uneven distribution
of experiences
persons.
At a more subtle level, these implications
cations in ~
Thus, chuges
can be expected
of extended
un-
are fikely to shift
the longer durations.
the introduction
to lead to greater unemployment
across nonemployed
of w=ks of &gibifity
who experience
level of weekly benefits
effect on unemployment.
expected
the role of W pohcies
that change the size of weekly benefit amounts are not Ekely to tiect
where=
unemployment
concerting
progrms.
highlight the importance
A casual compmison
of ehgibility
quUfi-
of WI regimes across stgtes reveals that those
progrms
paying higher benefits dso apply more stringent qrrdifimtio.n requirements.
progrms
in effect offer higher wee~y
benefit mounts
Such
to those persons who qrrdlfy md at
the same time assign zero weeks to eligibility to a greater &action of the nonemployed
rdation.
Consequently,
these programs are fikely to indum less unemployment
the findings of this report because the higher we~y
bmefit mount
pop-
according
ptid by a progra
yields
no change and the lowering to weeks of ehgibifity reduces the amount of unemployment.
x
to
1. Introduction
Over the past 20 years there hss been a steady flow of empirical research on e>-aluatillg tbe
eflects of unemployment
demographic
programs
groups.
insurance
Regardless
requires empirical
labor-market
decisions.
the unemployment
whether
(UI) programs
on the unemployment
of the group considered,
knowledge
assessing the full impact
Most ob~.ious, one needs to know how changes in a UI system alter
such changes induce nonworking
individuals
h addition,
to become
more. indirect effects, an arralyst dso requires information
future
UI benefits.
to policy
Finally,
Existing
changes in adjusting
if program
one needs some determination
behavior.
UI can influence unemployment,
source
other sourcm,
that permits
activities
The central empirical
policies
on the mount
one to overcome
a flexible econometric
commonly
question
in finacing
across studi=
to collect
features,
w-hicb
make it very
rehable estimates of comprehensive
of the factors noted above.
of such effects.
many
of the shortcomings
fimework
This paper
The analysis exploits
inherel~t in
for assessing the role of LTI-
of individuals
accounting
in this analysis concerm
that irrdividuds
of interest corresponds
for both their
experience
the influence
between
where
which is a popular
To study the effects of UI entitlements
occurs during spells of nonemployment,
on both the length of nonemployment
1
jobs,
of UI
to a CPS type measure of the sort most
statistics rather tharr weeks of UI coHection
choice for other research on this topic.
these entitlements
activities
responses
in UI programs.
investigated
of unemployment
cited in national
unemploymmt
experiences
and participation
the concept of unemployment
alterations
bcompatibilities
for combinations
system features on the nonemployment
unemployment
the potential
that firms alter their hiring and separation
analysis that pro~-ides estimates
and it develops
Considering
but available studies tend to consider effects in isolntion due
that account
presents an empirical
concerting
involve
to determix~e
UI recipients.
their employment
results for the purpose of developing
effects of UI policies
one i~ds
aspects of each of these possible routes through
or’ methodology cal problems.
difficult to integrate
a new data
changes
of the likehhood
research exsmines
to data limitations
of these
of the way in w-hich UI policies influence a variety of
duration of recipients of UI benefits.
of w,orking individrrds
acti>-ities of various
one needs to =amine
on how much
the influence of
spe~s and the division of these spells
between
unemployment
and out-of-the-labor
the argument that the distinction
for many people
time doug
when they are not working.
decision
Ebgibilit y to ret=+e
are potentially
The current body
at best to infer the influence
UI benefits
important
of empirical
factors
choice
during this
in explaining
instead of as OLF during an
research provides
of UI on the distinction
OLF; in fact, most of this work does not even recognize
One often encounters
and OLF is an arbitrary
to report himself or herself as unemployed
of nonemployment.
evidence
acti~-ities.
between being unemployed
with the levels of these benefits
a person’s
episode
force (OLF)
between
only indirect
unemployed
OLF as a possible
and
status in the
labor market.
Data limitations
aIld unemployment
have been a major
experiences,
individuals
over this horizon,
during
order
to draw inferences
the U.S. population.
F:r~her,
the sample
from its results about
Data” sources
malyzed
if one is interested
compensation;
measures
the impact
in that concept
of unemployment.
Further,
to create reliable
dso commody
proties
progmm
avtilable
durations
measured
status.
composition
in
on segments
of
of stationarity
2
on UI
a random
and utilization.
of UI recipients,
but only
as time spent co~ecting
of UI polities
a basis for evaluating
Survey data, on the other
of information
assumptions
UI
on CPS
key benefit vnriables in specifications
The shortage
to
experiences
information
eligibihty,
data do not provide
to include
w,ork.
hterat ure do not meet these
UI benefits,
to recipiency
for those included.
forces the imposition
of UI pohcies
for an analysis of the tiects
of policies on shifting inditiduds
often lack sufficient information
A stud> of
or uses survey data, v,hlch provides
of unemployment
these data do not Wow
a r~dom
data, w-hich offers accurate
the unemployment
UI
of these benefits
unemployment
requires
the effects
to infer individuals’
data permit one to malyze
UI benefits
the utilization
in tbe existing
only for samples of UI recipients,
sample with sparse information
bud,
to determine
Past research either uses program
entitlements
Program
to infer the potential
time horizon,
time frame.
linking
group considered.
and to relate these items to the indlviduds’
tile relevant
demands.
the relationships
demands ,of any sample used in the empirical
sufficient information
over an extended
in analyzing
regardless of the demographic
the full effects of UI makes substantial
A sample must include
obstacle
and
in survey data
in statistical models, whi&
resdts
in specifications
This paper
National
random
.
develops
Longitudinal
the influence
known to be grossly inconsistent
ad
analyzes
Survey (YNLS),
of UI programs
a new- data set based on the Youth
u,hich constitutes
on youths’
sample of youths with dettiled
a period
covering
earnings
and on episodes
supplementary
provides
information
benefits ad
link the unemployment
their UI eligibility,
durations
mid-
employment
Exploiting
Yh’LS
offers a
data available
In conjunction
rules of that State,
assessment
the unique opportunity
on
with
the YNLS
of an individud’s
UI
offered by the YNLS to
to dependable
measures of
of UI benefits on the unemployment
of young people,
this topic, this paper develops a comprehensive
in UI progras
the extent
to which youths
entitlements.
nonemployment
Further,
and their ncmemployment
it links this information
activities,
including
unemployment
a comprehensive
durations
that jointly
model describes
to a vmiety
of measures
between UI and unemployment,
picture
rmd evaluating
of time spent in
the role of UI programs
In addition
activities
many findings.in
this paper proposes
of individcrds.
the li terature.
which in com-
One component
on the lengths of nonemployment
spds.
of these spells as unemployment.
account
and non-recipients,
model
analyzes
viduds
collect
whether
UI recipients
the generosity
UI benefits.
In specifying
of UI programs
influences
these components,
3
of
an econometric c
assesses the effects .of UI on the classification
betwem
to
the analysis presented
the effects of UI on three aspects of behavior
the nonemployment
for distinctions
assesses
of the influence of UI policies on the distribution
that occur upon leaving jobs,
determines
characterize
The analysis
time in insured and total unemployment.
here furnishes a natural setting for integrating
To develop
experiences.
picture of youth’s
are eligible for UI and the degree to which they draw on UI
offering insights into the connections
.
of the
labor market statuses over
and nonemployment.
this analysis explores the importance
involvement
bination
The
1985, with considerable
an accurate
Cohort
source for studying
experiences,
histories of a random sample of individuals
En route to examining
model
market
and UI program’
to construct
utilization,
an unparalleled
histories of each person’s
data on State of residency
sufficient
eligibility,
labor
the years 1978 through
of both
x,ith the facts,
of the
A second
Finally, to
a third component
of the
the likelihood
that indi.
this study accounts
for dl the
dimensions
of LTI benefits
in financing
programs.
and incorporates
controls for the t=-rate
f~tures
faced by firms
Further, the analysis takes great care to avoid biases in estimat il]g
responses
to UI entitlements
generosity
of UI programs
by ensuring
that mriatiorr in benefits reflect differences
rather than differences in w,orkers’ attributes
in the
which dso determine
benefits
The remtinder
of the YNLS
of this report consists of ten sections.
over other sources for constructing
arrd unemployment
md
employment
for a random
experiences
YNLS.
Section
during
the first hdf
for analyzing
investigate
4 providm
of the 1980’s decade.
the problem
youths
7 investigates
Section 8 proceeds
and the fraction
empirical
determine
account
of youths’
Section
between jobs,
the effects of UI programs
to the next step ud
of nonemployment
link between
UI recipiency
the comprehensive
occurs after job separations.
examines
ehgibility
5 pr=ents
and Section
of whether the generosity
experience
work histories
provided
m econometric
of U] programs
by the
of UI
framework
this framework
influences
As a first step in mswering
to
the amount
of
th]s question,
on the lengths of spells in nonemployment.
the relationships
time classified w unemployment.
and UI benefits.
effects of UI pohcies
Finally, Section
the earnings
for and utilization
6 ttilors
.
UI entitlements
Section 3 characterizes
using the weekly
the effects of UI on unemployment,
unemployment
Section
~ dettiled
the advantages
a data set that integrates
sample of individrrds.
of youths
Section 2 outhnes
Section
Section
10 combines
on the duration
11 summmizes
between
UI entitlemexlts
9 ~plores
the
the findiz~gs to
of unemployment
that
the r~rrlts.
.
4
2. Linking
Data limitations
.
UI Entitlements
have severely
and
curttiled
our ability
scription
of the links between unemployment,
obstacles
stem from an incapacity
benefits
and the co~ection
requires sufficient
uninsured
dso
unemployment
sourcm
to provide
experienced
stantial compromises
in accounting
form of heroic assumptions
they also commonly
need for detailed
between unemployment
2.1
Data
knov,ledge.
entitled.
Given the deficiencies
often take the
statistical
structures
and
to avoid the
of data sources used in pre~.ious v-ork,
of the empirical
experience=.
relationships
The availability
duration
between
linking U1 eligibility,
of the YNLS pro~.ides an
vary substantitiy
an individud’s
is needed
nriables
of benefits
~gibility,
whether
an individual
to whi~,
w=kly
he or she is
benefit
across states and are characterized
earnings
of the UI benefits individuals
pattern of wages over the relennt
to determine
determining
history,
requires a complete
in almost every state depend
The entitlement
and Benefits
and the amount
roles and regulations
One =sentidly
me=ures
of information
UI compensation
The spetific
requirements.
UI Eligibility
to Impute
amount
to receive
relationships
ehgibility
or to make sub-
to pro}.ide an assessment of the degree to which these com-
unemployment
Requirement
and potential
accurate
of data
research either to focus on
These compromises
of restrictive
but
to begin such an msessment,
A considerable
is ebgible
time horizon,
and UI programs
for data shortcomings.
promises ha~,e clouded our understanding
OppOrtunity
of insured and
The inadequacies
ha~re forced previous
involve the introduction
md
over this horizon.
UI
Such a match
the amounts
over an extended
de-
The mtin
that permit the creation of proxies for missing information,
there has been no opportunity
UI participation
and UI utilization.
between
by an individual
of information
narrow, aspects of the relationship
a comprehensive
to measures Qf unemployment.
UI entitlements
this levd
to formulate
UI eligibility,
not only to distinguish
to infer the individud’s
Experiences
using current data sources to reliably match potential
of these be~efits
information
Unemployment
benefit
scheddes
amounts
by complex
and qualification
time series of weekly e=nings
to obtain
are entitled to receive; the roles determining
not only on the mount
of income,
but dso on the
time period.
associated
with UI programs consist of an assigned weekly ben5
efit. amount
(li’&).
and the number of u.eeks of ebgi bilit y during w-hich benefits are avtilable
(WBA)
An individual
of qualification
requirements
me not unimportant,
of eligibility
must meet certain criteria to qualify for benefits and then satisfy a set
on a week by week basis. While the weekly qrrdification
the program features of central importance
criteria.
These conditions
benefits as well as the amount
determine
whether
of benefits atilable
While
determine
there is a large amount
entitlements,
an individrrd
week. period
of diversity
among
States in the eact
,from the most recent employer.
criteria.
resulting from direct involvemmt
in a labor dispute.
a worker to kave an employment
history that demonstrates
red/or
The evidence
for such m attachment
earnings
a permanent
in any quart=
of the base period
received
(HQE),
than some multiple
(usuWy
eligibility
1.25 or 1.5) of HQE
of the State express their ehgiblfity
criterion requires
attachment
to tke
level of earnings
during a recent fifty-two
of the States determine
ehgibihty
work with wages greater than some nominal
five States reqtire
wag=
in the base period
to receive UI payments.
to become
requirements
and hdf of these States add a requirement
The remtinder
unemployrneut
(BPE),
and total weeks of work during
imately hdf of the States require a worker to have a mitimum
BPE,
rules used to
ad
consists of a mitimum
of total earnings
the base period (UW7 ) to establish an individuals
one-fourth
HQE
along
with BPE
efigible for benefits.
in tams
of a minimum
Approxgreater
Another
level of
of wages in more than one calendar quarter.
b=ed
amount.
in more than one calendar
upon a reqtired
Whether
explicit
number of weeks of
or implicit,
quarter for an individud
dl but
to be judged
ehgible for UI payments.
Once
“usudn
-
termed the “base period. ”
All States use some combination
high=t
initiated
The first concerns
The second ehgibility
a minimum number of weeks of work in covered employment
week period,
to any
All States have disqualification
pro~,isions for lea~-i:g work without good cause, d]scharge for misconduct
labor force.
is entitled
.
year, ”
every State appfies two types of ehgibility
the reason for separation
for this analysis are the set
during a fifty-two
witk the fifing of a UI claim, termed the “benefit
tests
deemed
efigible,
an individud’s
earnings in covered employment
W’BA
is detetined
up to some mtimum
6
w a fraction
of his or her
level such that approximately
-
hdf
of the usual weekly
calculate
a person’s
T~BA typically
w,age is replaced
IVB.4. The most common
equal to 1/25 of HQE.
weekly earnings
by UI payments.
(AWE)
over
States using this procedure
method
the 1~’BA ranges
(i.e.,
as the appropriate
to qualified individuds.
AM’E
1.5 percent
for determining
=
Among
BPE/Wli’),
of AWE.
of BPE,
Finally,
The first approach,
determines
calculates
adopted
a fraction
up to a maximum
The third scheme determines
of weeks, people
individuals
(usually
by about ten Stat=,
of an individud’s
of benefits
1/3) of BPE
with a ratio between
HQE
common
these two extremes
All
weeks of benefits.
md tl~l..
(TBA)
and then cdcdates
The most
to an individual
WE
method
in the
by the ratio
resigns U ‘E as
again up to some maximum.
based on the ratio of BPE
with a ratio above 3.5 is assigned
with a ratio close to 1.5 are allotted
provides
work mperiences
avtilable
in the base period,
M7E by using a schedule
an indi}.idual
on BPE,
number of weeks. Another
ranging from 1/2 to 4/5 of lilt:
Under this regime,
defines
who is eligible for UI payments.
that use information
the total amount
over the benefit year as a fraction
of TBA /WBA
as a function
ti’E
base period by one of three methods
method
the third
which implicitly
but a few of these uniform duration States pro$.ide everyone ~,ith twenty-six
prevalent
the ten
the number of weeks of benefits (li’-E)
the same number of weeks of benefits to every individual
The second approach
with
w HQE
measure of usual eunings.
States apply two basic approaches
atilable
to
defines the usual w,age as average
from 1/2 to 2/3
regime sets the lJJBA equal to approximately
BPE
defines usual earnings
A second approach
the base period
States use three methods
the minimum
the mtimum
to HQE.
number
numbez of weeks, and
me given an intermediate
number
of
weeks.
2.2
Data
There
Sources
Used in the Prem”ous
are printipdly
issues of UI eligibility
(CWBH).l
potential
two types of data analyzed
and utilization.
tion Offices of UI programs,
in existing
First, there are data atilable
su& as that provided
While these program
duration
Litemture
by the Continuous
studies
the individuals
Wage Benefit .History
on the amount and
making up these samples are observed
] Studies uting such data sources include Newton and Rosen (1979), Cl;sen
7
the
from State administ ra-
data sets offer very rehable information
of UI compensation,
to examine
(1979), and Moffitt (1985).
only as long m they are actually
information
co~ecting
benefits.
Consequently,
on only a very select group of the nonv,orking
Second, there are data from ~.arious representative
the Panel Study of Income
Dynaics
these data sets include
population
(i.e. UI recipients).
sur~’eys of individuals
such as the CPS,
(PSID ), and the earEer National Longitudind
Sur\,eys
(NLS ).2 In contrast to the first type of data, these survey data conttin insufficient information
to impute
individuals’
potential
UI compensation
without
relying on wsumptions
not credible.
Previous studies (e.g. Clark and Summers (1982a),
Card (1988))
have attempted
m out-of-work
her emnings
indi~-idud’s
history.
to infer a person’s
Further, one often cannot
in these surveys
unemployment
during a y=
2.3
of the
The YNLS
comparable
ad
or on single episodes
that enables one to overcome
to infer an individud’s
UI ehgibllity
accurate
over a relatively
short
these data conttin
were received.
.When
picture integrating
encountered
State, the YNLS protides
and atilable
comprehensive
weekly benefit
combined,
and employment
with supplementary
crdendar yem information
the average
but it supplies an inactivities of youths
with the data sets used in
statuses and earnings OV= a period co~,ering the
mid- 1985. In conjunction
of that
benefits
data on accumulated
representati~,e sample of youths with comprehen-
labor-market
and the UI benefit rules
UI payments,
of unemployment
on the unemployment
many of the problems
sive histories on each person’s
rehable
provide
and uninsured
in progress.
past work. The YNLS includes a nationally
protiding
sources
by treating
measure of his or
between insured
among the second type of data fisted above,
source of information
In addition,
benfits
Blank ad
J’NLS
cl=sifies
years 1978 through
and atilable
distinguish
at best th=e
w~ith some spdls interrupted
Features
Topel (1985), ad
previous annual labor income as the appropriate
unemployment
time horizon
ehgibllity
that are
data on State of residency
sufficient employment
benefits
information
information
duting times of unemployment.
on the receipt
of UI benefits,
on the total number of weeks a youth received
amount
over the year =d
these data permit
UI entitlements,
the utilization
the month.
one to construct
in which
a re=onably
of these entitlements,
and the
2 Studies uring such data sourcesinclude Ehzenberg md OUaca (1976), Clark and Summers (1979, 1982a)
and Katz (1986).
8
.
labor market acti%-ities of individuals.
The
development
of this picture
histories of il~difiduds,
unemployment,
initially
not only dating
but dso identifying
requires
the construction
their “periods of employment,
of complete
work
nonemploymellt
and
the precise time pattern of their wee~y
level of dettil is needed to infer UI benefits and to determine
the availability
during episodes
of describing
employment
when individuals
experiences
do not work.
The t=k
earnings.
This
of these benefits
the earnings
of youths is the topic of Section 3, which immediately
and
follows the
current discussion.
Using this information
extent
to which young
activities.
compensation.
ingredient
To take ad}-antage
the collection
UI benefits,
workers me eligible for UI benefits
they draw on available
UI is an important
on work histories to impute
Knowledge
4 examines
the
along with the degree to whick
of youths’
eligibility
in assessing the role of UI progras
of the richness of the information
of UI compensation,
Section
and utilization
on their labor
provided
the analysis focuses
on calendar
the data set characterized
in Sections
of
market
by the YNLS
on
years as the periods
of
obser~-ation.
The later sections exploit
influence
of UI policy on the nonemployment
of concern in this analysis is to determine
amount
of unemployment
done without
entitlements,
While
the opportunity
UI collection
the YNLS
YNLS rdewnt
that occurs
experiences
whether the generosity
between jobs.
provided
by the YNLS
offers this unique
opportunity,
to construct
a contiguous
sequence
famili=
models
of UI programs
unemployed
analysis cannot
be
a data set that links UI
of duration
of unemployment
reports himself or herself as being
spe~s roles out the possibility
of applying
analysis and has lead us to focus on predicting
on the total number
during a nonemployment
on the Iagths
of the
of weeks in which the youth does not work. This
on the timing of unemployment
dects
problem
affects the
there are three shortcomings
First, data are not provided
iack of information
statistical
of UI programs
the
and labor market activities.
to this mdysis.
dufing
The particular
Such an empirical
speUs, but only on the number of weeks that an individud
unemployed
of youths.
3 md 4 to ~amine
of weeks a youth reports
episode.
9
the
himself or herself as
A second
detailed
limitation
questions
of the YNLS
about extraneous
arises because
jobs,
survey
Specifically,
wage information
that were not the mtin job held at the date of the interview,
trtining
program,
though
income.
asked
is missing for jobs
were not part of a government
time series,of wee~y
we impute a wage rate for those individuals
the eanings
from these small jobs
This imputation
interviews
were not
and were held for less than nine weeks or required less than twenty hours
of work per week. To obtain uninterrupted
as possible,
respondents
procedure
account
e~nings
for as many people
with missing wage information
for a negligible
utilizes wage data atilable
.
fraction
in preceding
even
-
of total labor
and subsequent
= well as earnings on other jobs held during the current interview year. Appendix
A contains
a description
of the procedures
used to impute
wages rates for those jobs w-ith
missing information.
.,
Third,
benefits
while the Yh’LS
than are atilable
UI utilization
able calendar
the average
..
received.
contains
more comprehensive
in other data sources,
within single none.mployment
year information
wee~y
However,
each nonemployment
benefit
th=e
episodes.
information
data preclude
Specifically,
on the receipt of UI
a detailed
the YNLS
on the total number of weeks a youth received
amount
over the year and the months
this annual information
spell, udess
is insufficient
an indi~~idual happens
analysis of
pro~,ides reliUI payments,
in which benefits
to determine
to experience
v,hat occurs
were
w.itkin
only one nonemploy -
ment spell that starts and ends w-ithin the year. We can infer whether UI receipt takm place
within speUs, but not how much.
.
10
3. A Description
of the
Earnings
Experiences
The imputation
and volatility
of these activities
Population
horizon
(termed
Longitudind
ixlformation
volatility
Setiors.
from the Current
These data essentially
either in the context
of a short
crdendm year at a level of detail indicating
usual hours w,orked per week, md
annual earnings.
the
lVhile such
it f~ils to capture much of the
that occurs within a year.
a source for constructing
a weekly history of both earnings and hours
of work which one can use to summarize the patterns of these qumtities
The follov,ing
summmizing
earxlings.
discussion
begins by summarizing
information
Compositions
The following
towards describing
time. The 12-month
1979-1984.
individuals
ad
weekly measur=
The smples
of youths within annual horizons,
considered
in this =dysis
used to describe
ea~
wriable
in the YNLS for which data are avtilable
The use of dl atilable
This analysis
Statistics
both the earnings
with the focus directed
how these vmiables vmy across md within age-education
horizons
of hourly
of hours worked.
considers a variety of variables to characterize
experiences
then considers
the patterns
experiences.
In
about weeks worked and jobs held w-ithin a year, and it
and Descmptive
discussion
the employment
to the topic of employment
next turns
then takes up tke topic of amud
Sample
the arrdysis begins with annual measures,
of weekly earnings o~,er the year, and finally examines
The discussion
over annual periods.
describes these work histories at decreasing levels of aggregatiml.
earnings experiences
characteristics
md
Current knowledge
Surveys of Young Men and of Young W’omen,
conveys the broad outlines of anrrud experiences,
The YNLS provides
3.1
on i,ldivid-
little is knov,n about
on information
acti~,ities of individuals
of survey weeks or over a previous
of weeks worked,
Remarkably
Survey of 1972 High S&ool
depict the earnings and employment
number
earnings information
over s.ud horizons.
rests primarily
Survey, the National Longitudinal
and the National
sequence
a base period).
of labor market activities
in the case “of youths
Employment
of Yollths
of tlI benefits requires comprehensive
uds during a 12-month
the patterns
and
data mems
to the 6 calendar years
consist of dl the observations
on
for the year considaed.
that different
11
correspond
groups and over
sample
compositions
are exploited
depending
on the particular
compositions
(i.e.
see Section
criteria incorporate
in the military
to infer emnings
the avtilabifity
labor-market
of grade 8
data for these individuals
on a wide rmge of variables needed
experiences,
with the most demmding
wage information
data requirement
held during the ye=
for intermittent
jobs.
to constrict
In dl,
– recall that the
the analysis of this
the various descriptions
of interest for describing
vary among
youths:
bow- the }=rious mewures
age-education
the first involves
a comparison
across
different
both of these dimensions.
let the mriable
yit denote
T
Yii = z
each measure into
k and = O otherwise;
age-education
cme the coefficient
t. Consider the regression
?k d2k +
by W groups in yea
= 1, . . . . K),
j(j
The
age-education
25-27
for grades 8-11;
equatiOn
md
d2k = 1 if individu~
i is a member
in Other words, the d] j ‘S me time
bpose
the id~tification
ad
the @j’s represent
dummies
cOnWtiOn Z;=l
and the
@j = 0, in =hich
with age-edumtion
the common
Of age-
group k over
deviation
experienced
= 1,.. .T).
empirical
1984.
or
e~zo~,
~k represents the average of y associated
1 to T(k
In the fo~owing
dummies.
wtitb all earnings
k=l
where dlj = 1 if t = j and =. O other~’ise,
group
associated
K
63 dlj + z
j=]
the period
an observation
measure for the t.~h indi~,idual in yea
(3.1)
education
It does so by decomposing
This
and time effects using a simple regression framework.
In particdar,
hours-of-work
of
of earnings
and age groups; and the second focuses on the time path of these me=rrres.
considers
me
sample selection
wperiences.
and work activities
dzk ‘S
Tke least stringent
for the
to work some time during the year
of wage rates for dl jobs
There are two dimensions
nalysis
with ezch composition
criteria require individuals
section rdies on twelve distinct sample compositions
education
these samPle
who are age 18 or more in March of the calendar year, not
and employment
does not supply
youths’
A describes
and not in school at any time during the year, and with education
and for there to be non-tissing
YNLS
A.5 and Tables Al).
all individuals
The more stringent
involving
Appendix
in detail and reports the sample sizes wsociated
years 1979–84
or more.
l.ariable and year analyzed.
work the periods
categories
considered
(2) ages 18-19,
20-22,
12
1...
T refer to the calendar
below
23-24,
uc
25-27
(1) ages 18-19,
years 1979, . . ..
20-22,
23-24,
for grade 12; (3) ages 20-22;
—
23-24,
25-27 for grades 13–15; and (4) ages 23-24,
25-27 for grades 16 and abo~,e (i.e. 16+).
Tke term grade here refers to the highest year of educatiOn
Tables A.1-M
and A.1-W
sizes associated
alternative
3.2
in Appendix
with these various
A list respecti~.ely for men and vromen the sample
age-education
categories
for each of the years and the
from” the YNLS
earnings of individuals
The first variable
item directly
collected
year.
permits
ARE
=
annual reported
ACE
=
annrrd computed
to a CPS-type
by the YNLS
The construction
provided
ACE, along with the sample compositions
“M” attached
present
summary
to the numbering
men, while the dmignator
“W”
of a table prments
for a mriable
resdts
signifim
associated
statistics
w-ith =ch
describing
age-education
the variation
the top number
and over time.
refer to women,
listed at the top of the column.
Each column
Moving
down a
appear in each box: the top number is the estimate of the coefficient
row at the far left of the table and with the mriable
(3.1),
with k identifying
taken m the dependent
this estimate
with dl obsermtions
of these
that the results refer to
the corresponding
below
variable.
categories
of a table indicates
that the statistics
to create values of both
equation
reported
tke second
amount of detail.
~k from regression
of the column
which is a data
One calculates
A.2, A.3 and A.5) describes the steps fo~owed
across and writhin the various
three numbers
year.
which involves a considerable
ARE =d
The designator
are:
of ACE requires use of all the wage and employment
A (Sections
measures
the
received on all jobs held in those weeks making up
by the YNLS,
and 3.1–W
measuring
earnings.
measure of annual earnings,
Appendix
Tables 3. I-M
of two nfiables
earnings;
for each calendar
the weekly eartings
history information
the construction
over the period of a calendar year. These quantities
corrmponds
variable by summing
the calendar
Earnings
of Annual
Information
column,
by an individual
sample compositions.
Jfeasures
earnings
cOmpleted
fdfing
variable
of yk represent
the age-education
y in the regression;
demographic
gives the average for an age-education
13
group,
group
md
fisted at the top
and the two numbers
the lower and the upper
into the designated
group listed flong
quartiles
associated
in all yems.
Thus,
the two lower numbers
describe
the dispersion
eack column
tiitkin the group.
are the estimates
effects occurring
of tl?e 6j‘s from regression
in each. of the caleudzr
Tables 3.1-M
The six numbers
reported
(3.1),
in the bottom
w-hich capture
the miation
in the ainud
measures ARE and ACE. The second column of these tables, with the’ v~iable
matches that used in constructing
properties
=cludm
of ACE is smder
individuals
data are unavailable
the description
who hold intermittent
of ACE.
ARE_
A comparison
listed
which
of ACE requires nonmissing
the sample available
for chwacterizing
for summtizing
ARE.
data for these persons
u-age
do not go into
of the results in the first and second
columns
are higher in the sample used to construct
u.ith the view that annual earnings are lower for individuals
the
The ACE sample
jobs at my time during the year - because
jobs - so the eartings
that the average mlues of ARE
consistent
in the YNLS,
than that avtilable
for th=e
earnings
for the variable ARE using a sample composition
ACE. Because the calculation
data on a wide range of vmiabl~
the period
years.
and 3. 1–W report statistics describing
at tke top, presents results computed
rows of
ACE,
show-s
which is
who hold intermittent
jobs.
Three
patterns
emerge from the results in Tables 3.1.
First, average
dispersion within demographic
annual earnings
iucrease w,ith age and education.
Second,
categories
typicrdly
increases v.ith age md education,
Finally, a~,erage earnings generally dec~ned over the period
1979-1984.
3.3
Reliability
In carrying
of Earnings
Memures
out the empirical
work on the effects of UI discussed in the later sections of
this report, we infer UI entitlements
same sort of information
for U] benefits
imputed
wage information
UI information
employment.
stings
done below.
benefits is much less restrictive
tissing
using weMy
that goes into the construction
of ACE u a measure of amud
imputations
of individrrds
prorides
data on canings.
of ACE. Thus, =sessing
some guidance
than the ACE sample composition
A and B for further
14
considered
here because
the samples containing
who hold intermittent
discussion
of our
used belov, to impute UI
is assigned where possible in constructing
(See Appendices
the accuracy
as to the refiablkty
The sample composition
to avoid deleting individuals
This is the
jobs in covered
of this issue. ) In any case,
TABLE 3.1-M
Summary Sfa(islks for(~8~$~
of Annual Earnings”
ARE
(Slm)
VAWMLE
EDUC, AGE
8+
1$-19
7.6
I
I
25
I
4,0
10.7
~
1
23-24
J
I
8.7
3.8
ACE
($ I,w)
I
11.5 I
8.9
4.5
12.I
""""--""T"
""""""" ---~i:2---""-"""---`r-"-"--"-"-""""~6--""-----""""--
~---"""--"""~<--"
mzz
ARE_
($ I ,m)
I
14.0
4,8
I
15.6
I,
15.8
;
5,4
14.6 !
~"---"------"-~:o""
53
I
13.4
4.2
25.21
I
5.9
15,2
.......... ............ .... ..... ~......... ............... ...-..! ......... .... ..........___-_-__
I
I23
13.1
11.7
16.0 1 7,4
143
5.0
I
16.7 , J.8
....-____
.......... .. .
15.9
14.8
18.1
"--"""""""""i"
---"-""--;05"-""""-----"-l"" ""`""""-":112-"""-------"----""--"---"
13,8
I
J.8
,"---::!"-"~j:~"--"""---"-""~"--"""""-"""~3n"-"--!"4:!"-!i"":n"""""""l44----!4:-----""-------""-"""---'
2a22
23.24
$
I
/
I
25-27
~..!of
13+
13
17.4
I
I6.J
8S
23.24
~
103
-
!8.3
i
21.9
I
, 12.0
17.3
I
I
l&
f
X-27
I
14,6
30.3
/
2.0
0,2
-1.2
O.J
I
1
17.7
9.4
18.8
, 10.1
23.0
I
1 12.0
18.7
19.3
18.9
23.5
I
19.0
19,2
18.7
10.8
23.2
I
11.0
24.4
23.5
i 10.6
... ..-..-..94
... .... ......+ ......... .... .... ... . ........ ........ .... ........ ..-- —-----(
21.1
20,6
11,8
26,0 1 13.6
26.1
25.0
i 14.1
23-24
Yea E~m
79
80
81
82
83
84
17.!
18.6
21.1 ~10.8
20.8
I 10.3
20.8
...... ..... ..... ............... . ............ ........ .... . . .. .. ...... .......... ....... . ... .. . .... ... . . . . ........ ... .
t
t
19.1
19.0
18.1
24.0 [ 12.4
24.0
~ 12,1
.....i.4
....... .....1 ..... ........a.{ .... ..... ....i .. ...... .... ..Ki ... . . ..6 . . ..... . ... . . . .. . . .... ...
1
,
I
18.2 , 8.8
8.3
9.?
2a22
E-21
,
23.8
25.4
}/
I
~ 16.0
31.2
i 16.5
;
2.3
[
.!)
0.9
4.7
-!.0
I
I 3.2
,4.1
,.1.9
25.7
30.5
1.0
4,9
-1.3
.. . . . ... . .---
..—.
Summary Statid&
ARE
(Slm)
EDUC. AGE
n+
18-19
.
1
I
I
2a22
I
I
I
3.9
4.1
4.5
5.7
I
1.1
1.0
7.1 I 1.2
1.4
.............. ....................4 .. ... ...... ............ .. .....- + ................... .... ...... . . ......
5,4
5.9
5.0
I
7,8
, 1.4
1.3
8.3 ! t.u
9.0
163
Z-24
I
I
25.27
~
12
1,6
9.6
I
I
2%22
I
3.6
23.24
;
3.6
13+
m22
I
I
23.24
27.27
I
I
1
6.9
10.0
12.3
I
4,8
I
12.I
~ 4.3
11.8
15.0
12.5
~ 4.8
9.0
121
8.7
9“2
9’5
127
.............................1...... ... . . ..... ... . . . .-----------------T-””- 10.0
10.2
,3,5
9.5
4,0
4.8
I 5.0
13.7
13.8
... ...... ........... ........ ...... ._ . ................. ... .......--------------------------------------------r
10.7
9.7
10.3
12.9
I 6.6
14.0
5.4
13.6
~ 6.8
I
I
9.9
8.8
11.4
~-----------------------------‘
I
6.8
2,0
I
8.1
I
1
I
I
I
6.2
5.6
5.8
I
1.3
I 1.4
9.3
9.7 I 1.s
.................................. ............................'. ....l .........................-9. . . .............
1
7.8
1,4
6.9
I 3.1
10.0
10.3 , 3.8
10.7
2.9
18-19
Z-27
TABLE 3.1-W
for Measures of Annual Earnings’
(1984 $)
ACE
ARE:’
(Sl,m)
(Sl,w)
1
11.6
15,0
I 63
15.2
15.0
; 6.7
‘“”
...... ......-..T . ..... . ..... ................. 4.... ..... .. ....... ..... . ..-.. -------—
12.1
11.8
15.9
15.0
I 7.4
15,5
~ 6.9
10.8
6.2
............tii
6.9
. . . . . . . . . .. . . . . .. . .. . . .. . . . . . . . . . . .. .. . . . . .. . . . .. .. . . . . . .. .. .. -. .. . . . .. . .. ---------- .. -----—
i
14.5
13.6
14.4
I
I@
23.24
/,
17.0
I
9.6
17.9
: 9.6
18.1
I
15,2
15.4
14.9
20.2
20.0 \ 8.9
19.2
I 9.7
9.3
I .. . . ......... .... ... .. .......... .. . .+ ........... .. ............. .......... ..... . ....... ..... ..... .... . . .
I
1
I 0,7
02
I
0.6
0,4
0.4
I 0.8
Q.6
0.0
.0.6
(R
~
;;;
.0,3
.0.!
~
.0,3
i
I
I
I
25-27
... .. ... . ..------Yem Eff@
19
80
81
82
83
a4
9.2
--------------------
I
-----
.
w-bile the sample composition
exploited in the current analysis of ACE differs somewhat from
that used in the subsequent study of UI benefits, e~-aluating the degree of measurement error
contaminating
for judging
ACE for the samples considered
the reasonableness
A comparison
here pro~~ides a ~~aluable source of evidence
of our subsequent imputations
of the ~-=iables ARE =d
of UI entitlements.
ACE offers a simple approach
relative accuracies for these quantities as measures of annual eartings.
for assessing the
inspection
of columns
2 and 3 of Tables 3. I reveals a great deal of agreement in the averages md tbe dispersions
imphed by these two mriables for similar sample compositions.
A more sophisticated
approach
for detecting
the extent of me=urement
variables ARE and ACE involves the implementation
of a multiple indicator model.
that both ARE and ACE are imperfect indicators of “true anual
}.ariable TE.
observed
Follov-ing
error in the
a classical errors-in-the-variable
earnings”,
framework,
earnings quantities ARE and ACE relate to the unobserved
Suppose
denoted by the
assume that the t~~.o
quantity
TE via the
relationships:
tnARE
= b + lnTE
+ c,
(3.2)
tn ACE = –b + lnTE
+ e.,
where the coefficient b is an intercept, and c, and ec are mutually independent
error terms which are distributed
which possess zero mems
total mriances
and wriances
of the natural
where the symbOls Uim,
in ARE, In ACE, =d
log of true earnings
equal to a? and a: respectively.
of the two observed earnings variables decompose
~~cE = C$E + c:,
qumtities
independently
measurement
as: aim
Accordingly,
ax~d
the
= u~.E + u; and
~~CE, and U$E denote the variances of the
fn TE r=pectively.
The pmameters
c? and a: determine
the dispersion of measurement error in the two mnual earnings =iables,
with a larger c: (a:)
signifying more noise in ARE (ACE).
relate to one another according
The =pected
dues
to the relation E(tnTE)
of
the mrious eartings quantities
= [E(tn ARE) + E(ln ACE)] /2.
Use of a single cross-section of dat a on ARE and ACE - i.e. @ven a sample of observations
on AREi
md
ACE~ for inditiduais
sufficient information
i = 1, . . . . N for a specific calendar
year - provides
to estimate the parameters a:, rr~, a~E1 band ~(fnTE)
In particular,
one can estimate the first and second moments of in ARE =d
15
Ln ACE from the cross-section
data a~ld then develop estimates of structural parameters exploiting
(3.3)
~~ = ~~RE – cO~ (/n ARE,
In ACE)
o: = a~c~ – COI-(fn ARE,
tn ACE)
u$fi = co>, (&n ARE,
the relatioIlsllips:
en ACE)
b = [E(trr ARE) - E(tn ACE)]/2
E(lnTE)
MaCurdy
(1985)
d~cribes
= [E(trr ARE)+
B(fn ACE)] /2.
the dettils of the estimation
both to crdculate pwameter
estimates md to compute
procedure
apphed k this analysis
the standard errors wsociated
with
these estimates.
Tables 3.2-M and 3.2-V’
obttined
report estimated mlues for the vtiances
for six cross-sections
corresponding
of measurement error
to the calendar yems 1979-1984,
set of pooled estimates that combines the data for W years. The designation
title signifies results for men, and “W”! indicates estimates for women.
sample composition
nonzero.
includes individuals
Rows 1 ad
findings gerrerdly
of u~cE than a: contributes
support
to me~urement
two conclusions.
First, the extmt
u: accounts
ACE is tiny except in the year 1979, when it is still small.
to the view that computed
picture of individuds’
e=nings
Cameron md MaCurdy
experiences
specifications
error in the two annual eatings
study ~ploits
the pud
for a smaller
error
Such evidence
over the period of a year.
discussion of these findings,
to detect the magnitude
nriables
the autocorrelation
we~. In addition,
considers a miety
16
and the properties
of
ARE and ACE. This more tiaustive
feature of the YNLS to Id=
of model (3.2) and to ~atine
this mdysis
of mea-
wee~y earnings data provides m accurate
(1988) provide a more sophisticated
along with richer statistical
empirical
error for
to 61RE. Second, the amount of mewurement
lends some confidence
me=urement
and
2 present estimates and standard errors for the parameters a: and u:,
surement error is less of a problem for ACE than it is for ARE;
contaminating
Each cross-section
vaiables.
These empificd
proportion
“M” in the table
for v,hich both ARE and ACE ~e nonmissing
and rows 3 and 4 show the fraction of total variance attributable
the two mrtings
along w-itll a
characteristics
several of the restrictions
of memurement
of sample composition
error m
issues. Wrhile this
TABLE 3,2-M
Measurement
Error Variance Estimates For Log Annual Earnings’
(sbndard errors in parentheses)
d
.274
.138
(.037)
(.109)
.129
(.053)
.022
(.023)
djo;,
.274
.413
dJd*cE
.260
.054
$
.206
(.039)
.136
(.030)
.187
(.034)
.198
(.025)
.189
(.018)
(:M)
.054
(.031)
.053
(.027)
.016
(.026)
.038
(.012)
.232
.159
.215
.249
.234
.wa
.070
.072
.026
.059
~
0 EslimaIm
bawd on Smplc
L
descnbd
in Ap~tiix
A,
TABLE 3.2-W
Measurement
1
I
Variance
Estimates
For Log Annual wings’
w
Parameter
m
m
H
m
m
@
.204
.134
.114
.167
.249
(.046)
(.041)
(.025)
(.028)
(.035)
.150
(.033)
.171
(.014)
.025
(.039)
.147
(.077)
.058
(.026)
.047
(.025)
-.008.
(.021)
.056
(.042)
.048
(.016)
djaIRF
.243
.182
.115
.138
.213
.140
.162
&Ju;c~
.038
.196
.062
.043
-.008
.057
.052
d
zw
Error
m
‘Estimates based On sample L described in Appendix A,
,
other analysis indicates a number of qualifications that need to be kept in mind in e}.dusting
the validity of the tu.o main conclusions
noted above, the main thrust of these conclusions
survives.
3.4
Characteristics
of Weekly
The atilibility
provides
youths’
and Hourly
of weekly histories on earnings and hours of work supplied by the YNLS
the opportunity
to examine the pattern of a mriety
labor mmket experiences
and over time.
in wrious
Earnings
tiithin annual periods,
The current sub-section
exploits
wage measures less aggregated
characterizing
both across demographic
this information
to explore
than total annual earnings,
discussion intiestigates aspects of employment
Beginning “with the topic pfy:ekly
of dimensions
groups
the variation
while the following
experiences.
earnings, there are three measures O<avemge weekly
wages that one can associate with an indi~.idual during a calendar year using data from tke
l-NLS,
For each week during a year, one can infer the variables:
U’EZ =
weekly earnings from dl jobs in week ~ and
WHf
wee~y
=
hours from dl jobs in week t,
witk t = 1, . . . . 52 signifying the length of a cflendar
A1t’W
one can compute
=
for each individud:
reported earnings;
ARE/AWW
=
weetiy
ACEfAWW
=
weekly computed
AVE(WE)
The first two quantities
merdy
= ~
f=]
the quantity
weeks worked,
annual
the follow-ing three wriables
year. 3 Upon calculating
earnings; md
WEZ/AWW.
represent familiar me=ures
obtained
by dividing
annual
earnings by weeks worked. .The latter quantity denotes a simple average of an individual’s
weekly earnings
over
a ye=.
3 In constructing the annualme=ure ACE, we use the actual calendar year which involves slightly more
than 52 week. See Appendix A for further discussion.
17
Table 3.3-M
the mriation
for men md Table 3.3-M’ for women present summary statistics describing
in the three measures of a~,erage weekly earnings across persons of %,arious age-
education categories over the calendm years 1979-1984.
presented in this sub-section,
ha~-e mactly
These tables, ad
the remtining ones
the same structure as Tables 3.1. The top of each
column lists the variable to which the numbers refez, and the three estimated vdum in each
box represent the m-cient
estimate #j of regression modd
with the 25th and 75th percentiles msociated
category
(the lower two nmbers).
are the estimated
Comparing
(3. I ) (the top number)
with observations
along
in the relemrrt demographic
The six numbers reported at the bottom
of each column
year effects ~k associated with regression model (3.1).
summ=y
statistics for the alternative measur=
of average weeMy earnings
listed in the first three columns of Tables 3.3 reveals general agreement among the r~ults
associated
with these measures.
quite sitilar.
reported
For the older and more educated
groups, the findings are
For the younger and less educated categories, there is tendency
for the hourly
earnings to indicate slightly Iov,er weekly wages than the other two memures.
light of the results presented in Tables 3.1, this lower tendency
differences in the sample compositions
These &dings
no doubt p~tidly
with intermittent
jobs - whose eartings
scribing the measur=
their counterparts.
reflects
used to compile the statistics in the wrious
me consistent with the view that the younger md less educated
observations
are not included
columus.
individuals
in the statistics de-
md A17E(1fi”E) - tend to ha%-elower weekly evnings
~E/Aiiqi’
The estimates for yea
In
effects reported in the bottom
than
rows of the table
indicate that wmkly earnings declined over the period 1979-1984.
In addition to these mrious averages, Tables 3.3 dso present restits to capture the extent
to which a person’s
Min(WE)
own w=kly
u the mtimum
t = 1, . . . . 52 for whi~
wage ties
due
within a calendar year. Define M=(E’E)
md the mitimum
both WEf
snd WHf
value, respectively,
of WEf
over
have nonzero and nonmissing dues.
md
w=ks
Form the
quantities:
=4.
AR(WE)
= M=(WE)
The mewure
miable
(Mux(WE)/Min(WE))
RR(WE)
RR
= Mu(ZnH’E)
- Min(tnWE);
and
– Min(WE).
ceptures the notion of a “rdative
rangen (or percentage
difference) for the
WE over the year, and AR represents an “absolute range” for W’E.
18
As in the -e
.-
Summary
Mw
ACUAWW
$
MAW
$
VARIABLE
229
210
Age
TAnLE 93-M
for Measures of Weekly Farnings “
(1984 $)
AVE(WE)
RRm)
$
StaIislks
I
I
8+
..........!!!?...l.!!!
. . . ... ..!!
.............?l..~.!!
...
.. . . .......\..!' .......
....!.! ......\...!. . ................!!!...~ ......!
......... ..!4..l..!.!6 . ... .
...3!! . ......!...!.. . ......
....342.......'. .!.!!......... .....?!...i
..............!...'..!
I
--------..?Z!...!7878
....?!s ....-J..!!o
- ...._.=m
........- ... . . ..~. 2’8
25-27
.... —----- -----
......./..!.! .... ..........!! .....l...!`4
.-..---...;.----2:.-
!..-.-::.-.:-.--o4:...4.-.-.-o---.-..;--:!~------------------
... ....iG
. ............~.?Y
307..... ...?....... . .. .
361
3ti..::
I 239
~-... .-----
... .... ........2? ....... ~..!!'
........ .
—- ....... ... ~...
.................!...l
29a
. .... ..3. .....'?
..............!! ....~......! ...... . ... ..! . .....................
....!! . ....
. . . ... . ...?! . .......j...!! ......3 .....!? .......\...!.!!......3. .....?!!...~ .....!
226
147
249
12
18-19 I 139
........ ... ..........~-... .-----
AR~)
$
....
.....!! . ...{......!..._ ... .._...!?_ ........................
.....!!! .....!.......!.......... ....!!
....!" . ............ ....!
.... .. ..?!'
.. ................._
0,41
....
.. ..l;4 ....... ... ..*...!*
. .......!...t
417 I
0.02
. ....3 . ... ....?........+ ......... .....3i ..............i .. .......... ..02i ......!..+
. ... ................. ..
. .. . . . . . ....!!
.. .. ... .... .. .... ...
..... . . ..i4...!'
. .... ..... ...... .. .. .
----.3;;-.-~6------j---2:!.------.3i----:2--l.-.-o:---o1i.-:~--+-:----<j---!~--------------
!.?.---?Y.-l?w
---"";:--:n"-"--"+"2rn""-";io""""3~"----":-22!""-"-"-";G-""--3n-"";"---Q:--""ojo"::--;-:---;m--':-""-----""----"-I
........--?*.-t-T...-.!..!1.4...-.l.4’
".----'"3i;-----:6!-"-""""+--~:--"-"-3G"""":----:"--:"-"-"o;"-"----`---~-----"--;6-""-!!:"--""""-"---------""
S.27
--------------
--------—!m
23-24
I&
~27
...... ----------
I 261
~ - --__’w ---
4W
--------
1237
4W
j-y
. . ....... ...........
0.29
10
lW
........ . .... . .... ....—. ... .... ........ ...... .. . .. ... . . . ..--------------------I
I
4’9
‘m
,269
403
497,.... .. ..............
0.07 ‘“32
0.43
~
20
.,35 ..--.----—- -------—.. w
.......... . ~:.. . . . ........................
. ..........................
.........
....---------------- ‘m
=- —--
.... ..... . ... ...... .. ....... .. . . .. ......... ... .
4 m
4n
59’
l-::.----:-.--::-.--..!l.3!:------:----.:~:-.-+--."o'...-:-.-.o:9----}----.!o.-..--:::----!!6--------------------------YurW79
—m
81
83
137
12
1$$
:;:
~~
,:8
;;
:
l;[
------------
“
[:~
I
.28
I
.21
I
::
O.m
O.w
I
,9
:,
I
I
!10
-1
z
t-
:
I g:.
~i~
:,
81
I
]
0:;
901
vol.
I
;:j, {
97 I
~
~til
09
~
Zu
[*
18
‘lw&
""...'....""`-..""--"""""-.--"----.--.-..;--t--i;""""""""-""""&o--`i-""";6;---"----"----";zl"-"""iv---"--"""--;;]"----kT"""----;~"--;:;*
81s
,
;;----------I
.... ...!...
...... . ... .. ..... ... .. . . . . . . ...... . . ..l . . . .........!..! .... ..,00
"""-"--""----"--J~`---"~;-t""---"-.-.-~~"-"--"-+------.-""-";;
61Z ]
L9E
I
Ost
~,
w,
11
I
f~’o
LOJ
6L
~ i---......_!~o
.-.. ---.. —---..—--------..--"-----;o"---l---;;--f6z----"-;g;l---";;----!o~--;;~--;;------;,t-;;;--""-"---8C0
6L
6S
I
--"--"-"----"------------"`""-"-----"----"--""gzo---Go--{""";6z"u:-"--""-;;""!"--L;---s9~-""""viF+"-"-;i;-~~-""--ii+-ii:i------"-"LZO
S.1
*ZSZ
-"""-"---------:--:~+.---i--i--;;o"-::--;o.-."l"-"-""-""-"-::"---.--zi.""l-.-"--""."-"--fl:----"-"----l-"--------f~---;;t--;i7--
$5
. ... .......... . ... . . . . . . . . . ... ..1
.. . ........ ....... . . ..AE
. ... ..J
.... ....... .. ... ..... . ... ... . ....... ...
.. ......-. ......-..- . ....i_ . ........j .
I. . ..o
L91 I
I
‘“
~z
I ‘z
Ozz
..........io . ......21 ................ . . ........ijz . . . ........il..F...j
. . . .........!!...io
....~...iz!!
. ......il...~..._
1
....
..LE.-Ej~.~[.~"--.---g''l-iiiz
I ?Iz
9ZI
.-..---.. -::!. . . .......... .
}
Lvo
‘z
.
~Z
I
. . . ... ..i .. ..i..
--------.
n
m
.
~----~:~-----;l--l----;;--~!~----;;;l-;z;---------.... ..... ..
...........................
.....
.
Wo
01
..... ..... ....................~.s. .. .. ..... ‘. . . I. . . ‘Fo
......o... . . .......
o
....-..-.- ....-----1--
]
Ozz
0s1 I
mz
.-.- ... .._-~!..-- .......--_-._-._--.~9!..e
I
1
S6
I
~Z
101 I
..................
6[-$1
Zl
. . ...+------------------
--"-"""""
"".-".".-----"":"--.--"6~----""--u]---"i:o-"~:o-"".;-""-"l--";""-"-:~!..-"-..;.-""""{"`"'-::----f~!"":;;--t"---------"-":f!---;-:-;:---"-""--..... ............... . . . ...!! . .......i.7..G..!?Q
....i.o
....{...i .
...!!! ........[
... . .. ..iz
i
. ...!!! .......i ....tii
‘z
69,
. . . . ...Qi...uoz
..........
Li
.......... ..... . . ..... ... ...........o. ... ~- .....-.---::o.-...--.-.--.-.l-.-.....--.-:y!....--.;l.-+-.----.-..---v!.-..-.;o;-4---;i,---.~!---zi-;---------------0s
Sro
Bf
Lzo
I
S8I
‘o
I
ml
9PI
LV[
I
j 6!;;
LZI
I
mi
of the averages discussed above, one can calculate a value for RR and AR for each individual
ill each calendar year and use these data as dependent
The fourth and the fiftk columns of Tables 3.3-}1
ated with these variables.
According
percent,
with a quarter of individuals
for absolute
ranges (i.e.
AR)
and 3.3-IV report the findings associ-
to these r~ults,
in weekly earnings within calendar years.
indicate
~-ariables in regression model (3.1).
indi~,iduds experience
Percentage
experiencing
changes (i.e.
=ound
average about 30
40 percent or more.
average changes of about
weekly wages, with a quarter of individuals experiencing
RR)
large variati611
Estimates
$100 (in 1984 dollars) in
chariges of about $10 to $20 above
the averages.
One can construct-”a set of measures for hourly earnings that are completely
to those formulated
above for weekly earnings.
analogous
Given the quantities
52
AH
= ~
;I’H1
=
hours at all jobs
anIlud
f= 1
and
AH_
=
annual ho,urs at all jobs for which wage data are nonmisiing,
the three measures of a person’s average hourly wages mrned over a year are:
AREJAH
=
hourly reported earnings;
ACE/AH_
=
hourly computed
earnings; and
A~ZE
(Ii:E/WrH)
=~ 1
M’Ez/Ii;H1
/AWW
=
the average of hourly earnings for those
[f=l
where in the cdcdation
Also similar to the r=ge
w~ks
introduced
= in (Mu(U’E/WH))
AR (WE/WH)
= M=(H’E/WH)
these mriables
= 0.
above, one can calculate:
RR(WE/WH)
Respectively,
= O when either t$rEt = O or lVHt
of this average, tirEz /WHt
wriables
in which a person n-orks,
– tn(Min(WE/WH));
and
– Min( WE/WH).
mpture the relative ud
the absolute r=nges of an individual’s
hourly wages earned within the yeai.
Tables 3.4–M and 3.&W’ present summary statistics for the five variables formed above
to characterize
the hourly earnings of youtks.
19
The first three col~ns
report findings for
VAR14BLE
Table 3.4-M
Summary Statistics for Measures of Ilourly
(1984 $)
ACE/AII_
AVE(WwH)
(ccnLs)
(ccnls)
ARWAH
(wnE)
RR(WEMH)
AR(WH)
(cenk)
I
1
I
552
I
558
0.29
....4m
... ..6!-.. -.!. --.--.m. ......-.!5-..--.-..-... -..n..-..-..-..03 a..=.-.!-.-o.----=.--...-. -!94
J
I
6%
632
0.25
I
I
EI)UC AGE I
520
8+
.......h..!a.!9...--...2R9..- .......-..626------612
I
I:arnings”
154
I
151
............2n22..l ....?4' ....654 ....72! ........j ...... ..4.30.............7!.?......L ......4!! .............!! . .....................o.. ............!?.............................!.?
695
6W
0,27
178
f
I
7W
I
433
23-24 1 342
7W
434
am
0.02
0,35
I
10
1
25-21 I
721
383
719
792
1
I
S56
12
18-19 I
354
435
613
I
595
426
124
1
7M
714
436
I
,
O,w
I
598
425
I
0.25
I
785
716
I
0.31
0,29
0.03
0.39
201
0
I
,
194
215
1(D
8
215
.
w
w
12a
I
721
I
720
I
I
I
o.2a
I8a
!? ..........!
....1
.......!
.....n.!!! ......J
.........!!
.....n3i
....!!
........{
.........!
......iir
....!....................!
......G..!...............?
......i
. ......?
23.24 1
527
1
IW2
I
557
I
917
1025
928
I
I
556
1026
929
,
I
0.05
0.36
0.22
I
;
27
257
207
,
....5.? . ............!.* . ...... ..........!3..............!on2 .............??4............!.!.4 .........! ...........!.* ............?!! .. ..........!.?..................
9m
902
929
0.29
22a
23-24 1 601
1134
601
i Ioa
i
610
It]a
0,31
‘
40
2a5
0.06
I
...! ............ .......... ..
I
.............52...4
16+
lwa
25-21
.. . .. . .. ...
YearEfrttis
7Y
RI
83
::
84
6a2
113s
1334
!
7~
-"-"l""------"-"--"-"--"-----"-`""""""-----;"-"-'-"-""-"-'----"--------------"""";""--""-----"--""-"-"'""-255
0.25
1122
1349 [
729
[361
;
0,05
0,31
,
54
29 I
z
..... . ..
Wo
I
I
le-
9,&
I
Zlc
6;;
WI
,8
~
........o
..................ii
....1
....i%
.............\.... ....... .j...2..!!..f
..E.~
......o
[i
.;
I
~
,
;:g-
g-
i
;:
.............._
......!!------J-.-.... .i6.-Y!
....
.o
....j.-“-;;--gzn----;i;--------;i-Gf--ii;--;ziI ...____..__?.!
..--.xL-___-!.-__..._-::-_:.”.......
Szl
Zvz
K
..
:!:c::g:
9Z0
WI
--.--.!-!..........-..............J.................fl..............t................!~
-..---i;
....-?.........--?..-.l--.?
......{ ....___!!_
0s1
811
;j . ......!~----
;,;...:!::-+__.::_3zo:_3zo
------..-‘O”O
....-.-..--...------i9...9.~--;;---.-4--.-.ii.-.2`t.-.zl;....;..-..-..;--!2--:G~---..--..-...::
88
, ~__ ‘0
,,0
..... ..... . . . . ‘1
...
izo
al
‘o
,
Wo
099
m
W*
Z99
86S
as
Ws
;i_:!__oi_920
.i...io.................i9
... ...............
.................il ....
/
0s1
WV ;
-";:o--------"`"-"""""-"-""l"---"----";;""-""-""i;-"---"-]"-"""""ii---ml...s..!
..............,Zo
........‘No
...........,.
........flL---
....l--------
‘Zo
f-”-:”-~~--;;;i--:”-----”--”---
S19
fLP
iLP
---i------------
BY;...: ....... ..;;;-‘s
6t I
88
-
..... ..-- 61-~1
.......... .. ...’
w,.... ..-w’1
~.
I co
n5
~n$
MC
01
LI”O
Zcs
ffs
I
~.... ........--!!!--..l_..z
'"-"t""----"-"-"""--------"--"-"-""""-1"--"-""-"--"""''"""--"-""--"""""-"`--""
““””---”--------”
-------------ZSf
]
UPS
WI
. .. .... ...
I
;6--"-!6----"":-:-"----":--::--";"-"""-"":""---"-"-"::"-.:fv---"ii------;------ii---.::f...-::----.:--:----:--~g~----:-;--::------------
LL
OLP
....... ..______ ..-. ——
__ ..... ....... ............... ........................................
S6
o
~
zzo
,,0
MO
i
,0s
69
L,,
'-"";;;--":--"";;--"-"t---------;;-Ky---;;-l--;;;;-""----i;
I
L86
I
SLP
30V
anaa
averages, md the fourth and the fifth columns list results for the tw,o range mewures.
statistics outline essentially
the same picture
as that portrayed
These
by the result for v,eekly
earnings prmented in Tables 3.3. This is true for the pattern of both the average and the
ranges of hourly eartings
3.5
across age-education
of Employment
Characteristics
groups, as we~. as the profles
over time.
Activities
There are a wide mriety of variables that one can infer from the WCAIY work histories
of the YNLS to d-tribe
the employment
acti~.ities of youths during calendar
above discussion of earnings already introduced
formulat=
a few additiond
Tables 3.5-hf
qumtities
and 3.5-W
of a year that individuals
several of these wiables,
The
md this analysis
to provide a richer chmacterization
report summary statistics for qumtities
years.
of work aperience.
capturing the fraction
work and the number of jobs that they hold during the year. The
first column presents results for the quantity
D.4EAfP
= dummy vtiable
for whether m indi~,idud is employed
time during the yem
This mriable,
of course,
Tables 3.5 reports
characterizes
(DAEAfP
at any
= 1 signifies employment)
annual employment
rates.
The second
statistics” for the number of weeks worked per year, (AM-W),
column
md
of
the
tkird column lists estimates for the variable
AEfifPS
=
number of employers over the year.
does not capture the extent to which indi%.iduds hold multiple jobs at the same
AEJIPS
time because
To protide
a person can work for diffment employers
mewures
of the atent
of simultmeous
at distinct times during the year.
job holding,
the fourth, fifth =d
sixth
columns of these tables prment statistics for the nriables:
ADAfJ
=
dummy -ables
signifying whether an inditidud
mtitiple jobs in any week dting
holds
the year (ADJfJ
= 1 indicates
simultaneous jobs);
dWMJ
AH’hfJlA14’i$’
=
number of weeks associated with multiple jobs; and
=
fraction of total weeks worked in whi~
20
mdtiple
jobs were hold.
VARMLE
EDuc. AGE
8+
18-19
DMP
i
89.0
I
I
I
I
36
22
51
I
91.1
I
25-24
I
I
I
29
I
34
39
91.3
I
12
18.19
[
I
D-24
I
9’.7
23.27
13+
20-22
25.2’
I
I
I
99.9
I
I&
23.24
I
97.8
25.21
I
I
52
2.0 I
“o
I
AWUAW
I
f
14’3
19
I
2.0
I
1.’
2.0
I .6
1.0
‘1
2.0 I
1.5
52
34
I 12
58
I
30
14
‘3
1,
‘7
19
44
1 10
1
33
I 11
42
1 40
3’
7s
I
I .0
46
45
5.9
10
‘9
I
14.9
1,
22
13.3
40113
‘7
[
22
~7
43
40
I 14
43
‘7
41
‘!9
4’
‘9
45
I 24
I
!3.8
2.0 I
1.0
58
29
I
18
24
I
41
4’
I.4
52
t
1.0
Il.’
2.0 I
46
52
I
15.4
1.’
2.0
1.0
I
,
I
I
2’
13
55
88
I
16
19
33
‘,2
15
22
49
113
38
i 18
37
6,
I
1
t5.5
49
48
52
;
1.0
‘“’
2.0 [
52
]
1.0
1“5
2.01
47
I
46
I
50
14.0
49
52
I
1.6
52
50
52
I
\
!.0
1“4
I
44
92
m
6
42
‘3
I6.6
19
2.0 I
1.0
I
98.7
~
9.3
1
I
I
1
I
1
1
I
I .6
1.6
52
43
1.46
99.9
I
/ !.0
99.0
I
23.24
I
I
2.0 t
52
I
97.0
I
I
1.0 “6
I
45
I
I
N
0
.
38
AWJ
(
I
97.7
i
2,0 1
I
52
43
I
#
8.9
I
40
32
96.4
I
m22
I
I
1.0
ADMJ
8.1
52
42
91.9
I
23.2’
I
and Jobs Per Ymr
1
1.9
!
!
2m22
TAJJLE 35-M
ror Wwks Workd
AEMPs
Summary SWILstb
AW
2.0 I
17.5
24
441,7
47
84
n
42118
44
80
I
19
VARIABLH
TABLE 3.5-W
Summary St#(islics for Weks Workd
AWW
AEMPS
DM~
EDUC.
AGE
8+
18.19
I
I
64.6
I
20.22
I
I
28
I
I
13
[
15
3’
50
1
I
i
18
34
52
i
I
15
50
; 1.0
45
62.4
I
and Jobs Per YearADMJ
1.0
‘“
2.0
AWJIAWW
I
I
7.6
14’*,51,23fl
I
I
I .5
1.0
AWMJ
43
I
7.5
2.0
[
I
[
10.2
2,0
I
3.9
14
‘3
,9i
829
[6
‘6
231[3
44
i
23.24
54.9
1.0
‘“
39
f,
3’
4,
40
65
4’
65
50
a3
I
25.27
31
52.4
I
I .3
I
12
I
90.3
I
/
20-22
a4.2
I
I
I
t
80.5
I
I
N
25.27
0
u
20.22
I
4’
52
11.0
52
I 1.0
41
33
41
52
34
“7
2,o
1
2.0
I
I .5
15
I .4
] l.”
[
35
42
52
I
1.3
I .0
I
91.9
42
Ill
I 8.2
52
I
I
1.0
‘“6
2.o
aa.1
[
98.0
25.21
I
42
44
52
; ,,0
52
; 1,0
44
42
‘“4
2.0
:
2.0
I
1
47
45
52
92.6
I
43
46
52
I
ao
a2
84
!
i ::
0.0
I
-1.!
,.2
.0.9
1,9
24
42124
7
27
I
I
41
I
I
61
13
I 1.0
“6
[ 1,0
2.0
2,”
.“
29
59
1 la
I
17.5
1
‘7.26!’2
23
I :;;
, 1,9
33
45
19
-!,0
4.0
,.0
I
50
78
- .—
9
42
.2
.0
I
.5
.1
.;
I “:
.;
I
I
.;;
0.0
40
a
16
“4
94
19
I
19.6
I
1.[:
10.0
‘46~17
I
-;
.;
I
48
19
13.4
I .4
,
I
M*
79
al
a3
,.
f
I
/
13
12.9
I
I
23-24
‘9
19
1
I
91.5
25-21
i
!
45
I
2a
I
a.6
2.0
I
19
17
I
I
I&
152959
2,1,,
6
I
I 0.9
2.0
‘5
I
11.5
I
t
77.2
I
23.24
30
13.4
33
I
I
13+
3
I
18.19
23.24
14
I
1Jr
[ :
,,
The results for the variable A“DJfJ
data aIld indi~iduds,
only those persons for which .4DAfJ
using all available
and AI1-J/All-!I”
m-hereas the results for the ~-~iables A1l”fifJ
to samples incorporating
The
reported in Tables 3.5 are computed
refer
= 1.
findings in Tables 3.5 generally. convey a picture of mten!jye
annual emplOymel~t
participation
and only a modest amount of simultaneous job, holding.
labor-market
in>.olvement is greater for men than for women; and it generally rises with age
and education
in the case of men, but follows a nonmonot ofic
Asone
would predict,
relationship
in” the case of
women.
Tables 3.6-M
across the vaious
and 3.6-W’
age-education
describe
the mriation
categories and over time.
dl the variables appearing ‘in these tables.
columns
of X_
in annual and weekly hours of work
Comparing
The previous
discussion defines
the findings in the first. and secOnd
reveals that wage data are missing for only 6V0 of annual hours (i e.
is roughly 0.94 times that of N).
However, as noted in Section
A. 5. of. Appendix
A, these missing wages affect over a quarter of the sample in any one yem.
subsequent work described
the procedures
in this report, missing wages are imputed,
outlined in Appendix
A more comprehensive
=mination
k
all of the
where possible, using
A
of the full complement
of r=ults
reveals a fairly wide dispersion in annual hours across the population,
calculated only over those weeks in v.hich an individud
and AR(\VH)
are
works. The memures of relative and
of person-specfic
that he or she works per week during times of employment
21
in these tables
but a relatively narrow
dispersion ii average hours per week. The variables AI “E(IVH), RR(IVH)
absolute ranges suggest a large =ount
the a~.crag?
vmiation
in the number of hours
in a year.
.—,
Summary
VAR3ABLE
Et3UC, AGE
8+
In-lq
AH_
AH
I
1015
3n
la
[
I
23.24
25.27
I
I
I
1~
I152
~z
Im
1
2n
I
m
‘w
Ia-lq
20.22
I
I
11$1
I
2012
I
Im
987
I
IW
36
I
37
Zms
, 35
I
I 32
1
36
2w8
14q5
I
mu
I 922
m8
I
q19
I
I
23.24
Iw
I
I
1043
I
Im
25.27
I3+
2U22
IS58
t
I
I
23.24
I*
23.24
I
I
mltm
m8
t422
1646
I
1717
IW4
lm
N,
Y-, Etbd,
~o
81
82
83
84
I
1434
‘w
f
36
2m$
I
\
‘q
8
0.35~ o
II
7
0.26
0.36, 0
10
w
[ O.m
0’33
0.47\ o
‘o
15
@
/ O.m
‘“24
0,29; o
8
,0
0.25
I
0.29I O
1
028
‘
! ‘“m
! 0.00
!
“24
10
0.28
41 I 0.00
m
/ O.m
!
o
10
!
Om
38
35
I ID4
I
I
Ztaz [ 14W
I’lq
“22
I
9
o
I
I
0.22, 0
1811
I
2~8
; ,~
’745 N8
41 ! O,m
1 3S
I
I
0.23
38
2W3
II
7
9
7
14
0.34I o
40
38
\ 37
39
44 ~ O,w
“3°
0.4, ;
o
‘o
‘“9
I
027 I o
6
45 ; Om
!5
10
1
,
I
13
I
mg
~
‘
I ‘“m
40
I
la
I IU8
I
Z-21
0.36\ o
w
I
1
1233
I
m
m133
w
I
1743
I
I
Z-27
1759
1
]6
36
1465
m8.
1432
1427
0’2’
10
o
0.26
33
I
33
I
[
0.41 I
40 ~ O.w
1073
1423
maim
1044
36
ARm)
0.31
40 I O.m
40 ‘ O.m
28
I
I
22
1sq4
I
1755 I 288
W5
35
35
1768 I 30
Im
12
\ ‘q
1066
I 380
I
1065
I
RR(WH)
913
t
1234 I 363
Hours d WorW
AVE(WH)
I
20.21
TABLE 3.6-W
for Annual and Wekiy
S@tislti
13,4
1.1
13,6 i
49
I
18.3
37.1 I -295
9.1
10.1
I
10”
-’H
I
o
41.6 ~ 0
m I 0.04
a
0.01
.1 I J.m
4.01 r ;
olo~
0.01,
.0
I
1
4. UI EligiNIIity
.Amajor
source ofcontroversy
amount ofrmemployment
towkich
and Use Among
inthe
experienced
Young
Workers
literature about theinfluence
of UIpoliciesorl
by youths in the U.S. turns on the issue of the extent
youths participate in the UI system. Where=
many empiriml studies suggest tkat
both the level of weekly U1 benefits and the dur=tion of these benefits ~ert
on the extent of rmempIoyment,4
role in youth unemployment
the
a significant effect
other studies argue that UI programs play only a minor
because most young people areineligible
for..compensation
from
these programs.5
This section
utilization
kistory
pr=ents
designed to describe
among young workers during the years 1979-84.
data described
combines
an mray of m=sures
in the previous
the information
to cdculatetke
section
with data “pie\-ided by the YNLS on UI collection
measures developed
below.
The discussion constructs
lengths
unemployment
m the pertinent base for dculating
on the v,ork
and
over the year
measures considering
as the rele%.ant time frame and vieu.ing both nonernployment
and
eligibility md usage of UI.
Afeasures of Eligibility
Considering
of an individual
a period covering one yea,
for UI compensation.
for UIifheorshe
In particular,
Such a classifi=tion
(4.1)
E/N=
where the quantity
there are se~:eral ways of measuring the eligibility
one can designate i person as eligible
is not working sometime during theyemad
to collect UI benefits.
“# ehgible”
lified for UI compemation
t~s]ndi~idudis
---
# digible
# nonemployed
who are deemed qua-
at sometime during the year, and the quantity
(Of course, # ehgible is necesstily
qualified
scheme suggests the measure
designates the number ofindividuds
represents the number of individuals
-d
The analysis relia
and UI
to impute UI e~igibility and benefits
several period
4.1
UI coverage
“# nonemployed”
who are not employed
during some put
of the year.
a subset of # nonemployed;
so E/N rmges
between zero
one. )
4 Examplesofsuch studiesinclude Feldstein(1978), Hammermesh(1977), Topel and Welch (1980), Ehrenberg and O=aca (1976), Newton and Rosen (1979), Moffitt and Nicholson (1982), and C1arkand Suml!lers
(1982a),
5 See, for example, Feldsteinand Ellwood (1982) and Clark and Summers (1979).
22
Suppose that one wishes to calibrate this meuure
fraction of nonemployment
such a calibration is
we/wn
# weeks efigible
=
WE/WN
= A.e
weeks efigible”
—
# weeks nonemployed
(4.2)
“#
{we/w.}
gives the number of weeks during a year in which a
person is eligible for UI benefits, the quantity “# weeks nonemployed”
designates the number
of weeks that this person spent not working during the year, and the notation Ave {.}
the average of the vtiable
Instead of considering
in brackets computed
norremployment
over individuals
perspective
for assessing the extent of eligibihty.
adjustment
yields a third measure given by
(4.3)
E/u
as unemployed
“#
making up a sample.
Modifying
status m the proper reference
the memure E/N
to reflect this
# eEgible
=
# unemployed
represents the number of individuals who are damified
at sometime during the year. (Note that “# eligible” need not be a subset of
unemployed”,
modification
“# unemployed”
denotes
as the relevant frame of reference as is presumed
by the above measures, suppose that one }.iews unemployment
where the quantity
to the
time during the year that they were qualified for UI compensation,
A second measure incorporating
where the quantity
to v,eight indi}-iduals according
so E/U
of WE/WN
can in principle
yields a fourth me-nre
we/wu
=
(4.4)
WE/~
where the quantity
WE/WN
Similarly, an analogous
given by
# weeks eligible
# weeks unemployed
= Ave
“# weeks unemployed”
reports w being unemployed
The wriables
go above one in value.)
{we/w.}
denotes the number of weeks that an individual
during the year.
and WE/W
correspond
to average point-in-time
me=ures
of
ehgib~lity, while the variables E/N and E/U reflect notions of ehgibihty over a period lasting
a year. Assuming
a random sample of individuals
(.4.4) give proties
for the kind of efigibifity
concerting
weeMy status: U’E/~’N
and a stationary
environment,
measures derived from CPS-tYpe
corresponds
23
(4.2) and
information
to the ratio of the number of persons digible
for UI in a survey week ova
the number not working; and }i-E/Ii~
measures the ratio of.
the. number of ebgible persons in a survey week over the number reported
The >-ariables E/N
observation
4.2
and E/V
analogous
have
interpretations
as unemployed.
if one adjusts the period
of
from a survey week to a surve~- year.
Afeasures of Utilization
Considering
a period covering one year, there are several quatities
to which individuals
translation
eligible for UI compensation
of the concept
introduced
(4.5)
describing the extent
draw on thtir avtilable benefits.
A direct
above yields the following two meastirest
# recipients
R/E =
# eligible
and
# weeks receipt
wT/we =
# weeks dlgible
(4.6)
bl”R/14’E
where the quantity
“# recipients”
=
Ave
{wr/we}
designates the number of individuals
efits during the year, and the quantity
“#
weeks receipt”
during the year in which UI recipients
draw benefits.
who collect UI ben-
represents the number of weeks
(Both
and tVR/WE
R/E
must lie
between zero and one because a person cannot colIect benefits rrldess he or she is eligible for
compensation).
‘UI programs,
the variable R/E corresponds
Where=
the vmiable WR/lT’E
to m mnual
reflects a point-in-time
utilization
measure of use.,
The most popular statistic cited to describe the degree of UI utilization
insured unemployment
unemployed
over total unemployment,
as the relewt
measure of
which implicitly
frame of reference for calcdating
is the ratio of
takes d] those who are
usage.
A measure based on
this statistic is
WT[WU =
(4.7)
WR/WV=
This quantity represents=
A measure
comparable
spent unemployed
# weeks wemployed
Ave{wr/wu},
anrmdlzed
to WR/WV
as the appropriate
# weeks receipt
average of apoint-in-time
tales
measure of usage,
time spent nonemployed
mther
than time
reference for guaging the extent of UI utilization,
24
Tkis
.
quantity is
wT/wr
=
(4.8)
ll:R/li’N
# u,eeks receipt
# weeks no,lemployed
= Ave {UVT/WU],
which provides another mewrrre of an annualized average of point-in. time usage.
A fifth concept
titlements
capturing
of utilization
that are actually
summarizes the fraction of the dollar amounts of UI en.
collected
during the year by eligible indi~.idud.s.
A measure
this concept is given by
ar[ae
=
(4.9)
where the quantity
“$ amomt
$ amount received
$ amount eligible
recei~,ed” denotes the number of dollars an individud
collects
iIl UI benefits during the year, and the quantity “$ amount efigible” designates the maximum
dollar amount of UI compensation
that the indi~.id”d
could have couected
had he or she
dran.n benefits during dl weeks of eligibility in the year.
4.3
A Data
Set Integrating
UI Eligibility
and Utilization
To calculate the various measures discussed above, the follov,ing analysis uses a szmple
created
by more stringent
selection
critmia than are invoked
to carry out the empirical
study of Section 3. The sample considered here consists of 3028 individuals
nationally-representative
conditions:
component
of 6,111 youths in the YNLS who met the following five
(1) interviewed in each of the first 7 years; (2) worked at least once since January
1979; (3) have valid beginning
jobs and in the miktary;
ad
drau,n from the
(5) have a re~onably
and ending dates for time periods spent employed,
betwmn
(4) left school md did not return prior to the 1985 interview date;
accurate
and complete
time series of we~y
with Jmruary 1978 or the last date of school attendance.
earnings beginning
As noted previously, the Yh’LS does
not provide wage data for secondary jobs of short duration or which involve mdy part-time
hours of work, For jobs fdhng
the aalysis
into ttis category detertied
assigns wages using a procedure
delete obserntions
from the saple.
described
to be in covered employmeIlt,
in Appendix
The resulting data set includes
25
A to avoid having to
1409 men and 1619
women
who experience
period 1979–85.
For” ead
4031 and 4250 episodes
of nonemployment
r~pectivdy
over the
Section 6.1 presents summary statistim describing this data set in dettil.
calendar year, the YNLS provides not ody
histories as described in” Section 3, but dso r&able
comprehensive
iriftiirnatitin on work
data on the total number of weeks that
a youth receives UI payments during the year along with the average weekly benefit mount
over this period.
Combiti.ng
m iridividud’s
weeuy
with data on his or her State of, residency
one cm infer this pmson’s ~
relevant ye=,
nonemployment
to cdcrdate
unemployment.
of UI co~ection
the UI benefit ties
&gibihty
These constructed
memures of UI efi~bifity.
and -ounts
cmy
ad
ad
earnings hist6ry in covered employment
Integrating
duting the cdendm
,of that State in the
and avtilable benefits during times of
data protide the sample used below
these data and the information
ymr cr-te
the s-pie
exploited
on time
below to
out the analysis on UI utilization.
Appendix
B outhn=
our procedure for inferring each individud’s
periods of nonemployment.
benefits yidds
remarkble
As discussed in this appendix,
accurate predictions
UI entitlements during
ow imputation
of the average wee~y
of atilable
benefit mounts
are seH-reported by UI recipients in the YNLS. The differences b our imputed dues
UI
that
and the
VAUM reported in the YNLS averages about $2, with lower and upper qumtfies of – $11 and
$20. Our -s&sment
also appe=
and of total benefits atilable
to match weU with data protided
youth receives ~
payments over the ye=
For detertiting
due to qtits
of UI &gibihty
UI ehgibfity,
as a disqurdification
benefit receipt me voluntmy
work md wemploymat
labor stoppage.
Unfortunately,
UI ben~ts.
While su~
do not disqu~lfy
A l=ge
number of stat=
~ow
Major reuons
tithout
resdting
the curent
for dsqitification
good cause, dis~mge
Eteratme has inte~reted
me often =biguously
“quitters”
phrmed,
inditidud
re-
in m orgtized
the provision for vol.
are ine~gible to rective
the majority
who quit for reasons related to the employment
=
horn UI
for mis-conduct,
from direct involvement
good cause to mean that d
provisions
individuals
md the months in which benefits were co~ected.
for benefits.
sepmations
leaving work tithout
in the YNLS on the totrd number of weeks a
the YNLS offers two options for defining job sepmations
fusal of suitible
unttily
horn UI compensation
of states
rdationsbip.
to coUect benefits if he or she quit to accept
26
“better”
work or join the armed forms.
only those individuals
Thus, in practice this provision usually disqualifies
v-ho quit for personal reasons,6
to alternative interpretations
of ~,oluntary separation
To examine the sensitivity of results
provisions in deciding an individual’s
eligibility for UI, the following analysis considers t~.o definitions of eligibility:
a narrow con-
cept that presumes dl persons w,ho quit their pre~.ious job ,are ineligible for compensation,
and a broader definition that disqualifies individuals
4.4
Pattew
of UI Eligibility
There are two dimensioris of interest for cdcdating
and of UI use: the first involves a comparison
It do=
using regression
so by decomposing
fraework
(3.1)
the various measures of UI eligibihty
across different education
the second focuses on the time path of these me~ures.
dimensions.
only if they quit for personal reasons.
This analysis considers both of these
each measure into age-education
introduced
in Section
Yitdenoting an observationassociated ~.ith an efigibiIitr
individual
considered
and time effects
3.1, w,ith the dependent
Or utilization
measure fOr tile ith
group
experienced
k over
the period
1 to T, and the @j’s represent the common
by dl groups in ye= j (j = 1, . . . . T).
below are the s=e
The age-education
ages calculated
using estimates of the regression coefficients
averages for the tious
the coefficients
graduates,
of equation
by “M” provide results for men, v.bile those marked by “W”
women. Tables 4.1-M
and 4. 1–W present estimates of the coefficiats
age-education
groups.
Tables 4.2-M
chmges
with the a~,er-
(3.1).
The tables
report findings for
7k, which characterize
and 4.2-W
@j + yk for j = 1979, . . . . 1984 with k designating
which describes
categories
u those analyzed in Section 3, as are the calendar years.
Tables 4.1 md 4.2 present values for the four measures of UI digibility,
designated
\.ariable
in year t,7 As previously, the coefficient ~k represents the average of y associated
with age-education
deviation
and age groups; and
fist estimates of
25 year-old
over time using the 25-27 age mtegory
Klgh schOol
of high sckool
6 A c=ud survey of the data on benefit
determination
c=es suggests
thatonly 15-20 percent ofnew
insuredunemploymentSPWS come to a determination
overseparation
from work issues
and only 30-40
percentofthe-es thatcome todetermination
aredefiedbecauseofvoluntary
se~rationfromwork.
variable
thatt&u thevd”e ofone when
7 For the me=rrres E/N, E/U ad R/E, v;I isan indicator
individudiisa member ofthegroupsmakingup thenumeratorand thedenominator
-d takesthevduc
oizeroifthisintitidud
isa member ofonlythedenominator
gmnp, In thecme oftheothermem”r~, such
wefwn istheobservation
forindividudiinyeart.
M WE/WN, y,, = wefwn whcm thevariable
27
TABLE 4.1-M
Measures of Unemployment Insurance Eligibility by Age-Education Categories
Broad Definition of Eligibility
Measure
Cataqorv
$ducation
8-11
12
16
Measure
Aqe
EIN
WE/NN
E/U
WE/WU
E/N
WE/WN
E/u
WE/WU
18-19
0.413
0.325
0.446
0.642
0.256
0.175
0.307
0.335
20-22
0.532
0.421
0.594
0.743
0.394
0.284
0.447
0.449
23-24
0.656
0.526
0.732
0.753
0.541
0.405
0.625
0.577
25-21
0.533
0.427
0.602
0.651
0.397
0.279
0.489
0.433
18-19
0.486
0.419
0.601
0.910
0.287
0.219
0.399
0.416
20-22
0.596
0.495
0.694
0.787
0.475
0.369
0.600
0.608
23-24
0.659
0.553
0.785
0.819
0.541
0.436
0.663
0.631
25-27
0.689
0.583
0.785
0.832
0.547
0.440
0.671
0.632
20-22
0.480
0.413
0.581
1.213
0.301
0.243
0.411
0.622
23-24
0.520
0.468
0.663
1.229
0.313
0.256
0.433
0.635
25-27
0.607
0.559
0.748
0.880
0.416
0.338.
0.’585
0.532
23-24
0.492
0.472
0.750
1.242
0.325
0.281
0.490
0.559
25-21
0.685
0.609
0.870
0.081
0.343
0.287
0.350
0.299
E
13-15
Narrow Definition of Eligibility
TABLE 4.1-W
Measures of Unemployment Insurance Eligibility by Age-Education Categories
Broad Definition of Eliqiblity
Narrow Definition of Eligibility
Measure
Measure
CatagOry
E/N
WE/WN
18-19
0.243
0.179
0.303
20-22
0.311
0.205
23-24
0.283
25-27
Education~
8-11
12
13-15
16
E/N
WE/NN
E/U
0.835
0.118
0.016
0.157
0.321
0.449
0.823
0.158
0.105
0.253
0.376
0.223
0.379
0.777
0.147
0.113
0.236
0.380
0.291
0.199
0.323
0,633
0.130
0.089
0.207
0.278
18-19
0.397
0.329
0.457
0.904
0.195
0.149
0.263
0.410
20-22
0.360
0.275
0.484
1.184
0.193
0.137
0.285
0.439
23-24
0.311
0.231
0.481
0.887
0.154
0.114
0.289
0.431
25-21
0.344
0.243
0.560
1.102
0.197
0.132
0.387
0.694
20-22
0.319
0.274
0.421
1.132
0.120
0.103
0.196
0.259
23-24
0.364
0.299
0.492
0.894
0.126
0.103
0.206
0.253
25-27
0.292
‘0.210
0.465
0.763
0.112
0.082
0.215
0.310
23-24
0.435
0.340
0.514
0.724
0.218
0.165
0.213
0.365
25-27
0.459
0.407
0.655
2.655
0.361
0.321
0.511
2.236
WE/W
TABLE 4 .2-M
Measures of Unemployment Insucance Eligibility by Year
(Mea9ured
as deviations
from
the
mean
Broad Definition of Eligibility
for
a 25
year old with 12 years Of education)
Narrow Definition of Eligibility
Measure
Measure
Year
E/N
1979
0.803
0.722
0.919
0.956
0.603
0.500
0.776
0.696
1980
0.829
0.724
0.947
1.082
0.645
0.518
0.790
0.744
1981
0.807
0.682
0.904
1.096
0.591
0.477
0.702
0.695
1982
0.151
0.631
0.835
0.866
0.597
0.478
0.709
0.659
1983
0.569
0.430
0.652
0.599
0.487
0.376
0.590
0.572
1984
0.375
0.302
0.453
—
0.3s,9
0.292
0.459
0.425
Averiqe
0.689
0.583
0.785
0.547
0.440
0.671
0.632
— 0394
.——
0.832
TABLE 4.2-W
Measures of Unemployment Insurance Eligibility by Year
(Measured a9 deviations from the mean for a 25 year old with 12 years of education)
Broad Definition of Eligibility
,
Narrow Definition of Eliqibilitv
Measure
M
t
u
n
t
Measure
E/N
WE/WN
E/U
1979
0.361
0.291
0.558
1.657
0.151
0.110
0.320
0.952
1980
0.401
0.287
0.674
1.263
0.212
0.142
0.427
0.681
1981
0.451
0.318
0.698
1.247
0.221
0.140
0.418
0.715
1982
0.430
0.316
0.666
1.283
0.253
0.175
0.459
0.712
1983
0.260
0.150
0.437
0.736
0.193
0.121
0.375
0.587
1984
0.160
0.097
0&6
0.42~
O.~52
0.094
0.324’
0.517
Average
0.344
0.243
0.560
1.102
0.197
0.132
0.307
0.694
graduates
as z reference group. Each table reports two sets of results to examine the impli-
cations of adopting
the two different definitions of eligibihty
of four columI~s fist estimates ~suming
ployed indiv”iduds
qualifications
described “above. .The first set
the broader definition,
vrhich iriterprets dl nollem-
v,ho did not quit their jobs for personal reasons and v,ho meet earnings
as eligible for UI benefits. The second set of four columns presents results pre-
suming applii”ability of the narrower definition of ehgibility, whid
assumes that all quitters
(for personal reasons or not) me ineligible.
According
ass=sment
to these findings,
the definition
of &gibiLty
matters with respect to role’s
of the extent to which youths are ehgible for UI benefits.
The broader definition,
which does not exclude dl quitters, typically implim 50 to 100 percent gre”ater eligibility over
the narrow- definition.
There ii””rio systematic relationship
point-in- time m-sures
of e~gibility.
Certtin pitterni”emerge
regardless of the definition or mewure used to quantify eligibil-
ity. In the case of men, eligibility increas=
for women u,ith respect to education,
is more =tensive
that eligibility generally
4.5
Patterns
with both age and education.
eligibility
For men, the results for time effects indimte
declined over the period 1979-1984,
dramatically
so for the broad
The time trends are either less prominent or nonexistent
for women.
of UI Use
Tables 4.3 md
Tabl=
4.3-M
4.+M
and 4.4-W
4.4 report analogous
and 4.3-W
present dues
to the definitions
estimates for the five measures of UI utilization.
associated
with age-education
list estimates for the time effects, with 25 yem-old
serving u the reference group.
Once agtin ead
used to detemine
ehgibility.
columns of each table are not dependent
categories.
The measures considered
according
in the first tw,o
on the definition of ~gibility.
to the fraction of insured unemployment.
for that measure of unitization correkspection
of these resdts
this rate rises with age in the cme of men, but does not necessarily increue
more education.
Tables
high school graduates
table provides two sets of resdts
The first column of each table presents fidings
sponding
The same is true
but not with regard to age. As expected,
for men than for w,omen.
defiIlition of eligibihty.
between annual and comparable
This same pattern holds in the cue of women =cept
28
reveals that
m men acquire
for the lowest educa-
i
TABLE 4.3-M
Measures of Unemployment Insurance Utilization by Age-Education Categories
CataqOrv
Education
8-11
12
13-15
16
Broad Definition of EligiblitV
Narrow Definition of Eligibility
Measure
Measure
Neasure
Aqe
WR/WU
WR/WN
R/E
WR/WE
AR/AE
R/E
WR/WE
ARIAE
18-19
O.O76
0.050
0.181
0.137
0.129
0.276
0.216
0.204
20-22
0.202
0.119
0.246
0.193
0.188
0.294
0.239
0.233
23-24
0.353
0.255
0.430
0.348
0.339
0.481
0.397
0.389
25-21
0.311
0.202
0.553
0.452
0.447
0.579
0.483
0.478
18-19
0.281
0.132
0.306
0.228
0.220
0.467
0.374
0.361
20-22
0.359
0.230
0.477
0.388
0.363
0.541
0.448
0.423
23-24
0.456
0.322
0.594
0.482
0.470
0.652
0.547
0.534
25-27
0.874
0.379
0.654
0.561
0.536
0.726
0.635
0.607
20-22
0.146
0.090
0.198
0.165
0.144
0.294
0.250
0.218
23-24
0.249
0.165
0.319
0.280
0.262
0.454
0.431
0.402
25-27
0.286
0.195
0.323
0.267
0.241
0.459
0!408
0.314
23-24
0.169
0.090
0.182
0.187
0.190
0.262
0.271
0.214
25-27
0.694
0.247
0.346
0.359
0.352
0.408
0.428
0.433
graduates as a reference, group. Each table reports two sets of results to examine the impli.
mtions of adopting
the two different definitions of eligibility described
above.
The first set
of four columl~s fist estimates assuming the broader definition, which interprets all nollemployed individuals
qualifications
who did not quit their jobs for personal reasons and who meet earnings
= efigible for UI. benefits.
suming applicability
The second set of fovr columns presents results pre-
of the narrower definition of eligibility, which assumes that all quitters
(for personal reasons or not.) are i:digible.
According
wsmsment
to these findings,
the definition
of eligibility
matters with respect to one’s
of the extent to which youths are ehgible for UI benefits.
The broader definition,
which do= not exclude U quitters, typically implies 50 to 100 percent greater eligibility over
the narrow- definition.
point-in-time
There is no systematic
of the definition or measure used to quantify eligibil-
the case of men, eligibility increases with both age md education.
for women with respect to education,
is more extensive
that ehgibility
4.4-M
ad
eligibility
for time effects indicate
dramatically
The time trends are either less prominent
so for the broad
or nonexistent
for women,
of UI Use
Tables 4.3 and 4.4 report analogous
Tables 4.3-M
For men, the resdts
generally decfined over the period 1979-1984,
Patterns
The same is true
but not with regard to age., As wpected,
for men than for u,omen.
definition of ehgibility.
4.5
between krinud and comparable
measures of ebgibility.
Certain patterns emerge regtidless
ity. h
relationship
and 4.3-W
4.&W
estimates for the five mewrrres of UI utilization.
present values associated
with age-education
categories,
Tables
hst estimates for the time effects, with 2j year-old high school graduates
serving u the reference group. Once agtin each table provides two sets of results according
to the definitions
used to determine
ehgibility.
columns of each table are not dependent
The measures considered
on the definition of dgibility,
The first column of each table presents fidings
sponding
to the fraction of insured unaployment.
for that measme
kspection
.
of utilization
of these resdts
this rate rises with age in the case of men, but does not necessady
more education.
in the first two
corre-
reveals that
increase as men acquire
This same pattern holds in the case of women except for the Iowest educa28
1
TABLE 4.3-M
Measures of Unemployment Insurance Utilization by Age-Education Categorie9
Broad Definition of EliqiblitV
CataqOrV
lducation
8-11
12
13-15
16
Measure
Narrow Definition of Eliqibilitv
Measure
Measure
Age
MR/WU
WR/WW
R/E
WR/WE
ARIAE
R/E
WR/WE
ARfAE
18-19
0.076
0.050
0.181
0.137
0.129
0.276
0.216
o.2oq
20-22
0.202
0.119
0.246
0.193
0.188
0.294
0.239
0.233
23-24
0.353
0.255
o.q30
0.348
0.339
0.q81
0.397
0.389
25-27
0.311
0.202
0.553
O.qsz
0.447
0.579
o.q03
0.478
18-19
0.281
0.132
0.306
0.228
0.220
0.467
0.374
0.361
20-22
0.359
0.230
o.q77
0.388
0.363
0.541
0.448
0.423
23-24
0.456
0.322
0.594
0.482
0.470
0.652
0.547
0.534
25-21
0.874
0.379
0.654
0.561
0.536
0.726
0.635
0.607
20-22
0.146
0.090
0.198
0.165
0.144
0.294
0.250
0.218
23-24
0.249
0.165
0.319
0.280
0.262
o.q54
o.q31
O.qoz
25-27
0.288
0.195
0.323
0.267
0.241
0.459
0’.400
o.37q
23-2q
0.169
0.090
0.182
0.187
0.190
0.262
0.271
0.274
25-27
0.694
o.2q7
0.3q6
0.359
0.352
0.408
0.428
o.q33
.
TABLE 4.3-W
Measures of Unemployment In5urance
Cataqorv
Education
8-11
12
Utilization
by Age-Education
Categories
Broad Definition of Eliqiblitv
Narrow Definition of Eligibility
Measure
Measure
Measure
Aqe
WR/WU
WR/WW
R/E
WR/WE
AR/AE
R/E
WR/WE
ARIAE
18-19
0.103
0.043
0.247
0.177
0.175
0.399
0.307
0.298
20-22
0.233
0.055
0.239
0.184
0.184
0.386
0.301
0.302
23-24
0.209
0.079
0.421
0.370
0.347
0.642
0.593
0.558
25-21
0.104
0.023
0.095
0.010
0.008
0.225
0.099
0.089
18-19
0.163
0.073
0.249
0.182
0.166
0.400
0.302
0.275
20-22
0.334
0.106
0.328
0.262
0.243
0.527
0.440
0.412
23-24
0.364
0.103
0.403
6.324
0.308
0.603
0.501
0.491
25-27
0.750
0.114
0.476
0.361
0.346
0.686
0.531
0.510
20-22
0.077
0.030
0.149
0.118
0.103
0.264
0.228
0.220
23-24
0.317
0.066
0.197
0.162
0.149
0.481
0.402
0.375
25-27
0.463
0.063
0.113
0.100
0,090
0.356
0,316
0.293
23-24
0.200
0.055
0.098
0.064
0.068
0.044
0.050
0.055
25-27
0.257
0.065
0.137
0.060
0.074
0.192
0.116
0.137
N
m
T
E
13-15
16
TAU[,Z 4. 4-M
Measures
(Measured
as deviations
of
Unemployment
from
Broad
Mea9ure
the
mean
Definition
Insurance
for
of
Utilization
a 25 year
Eliqiblitv
old
by Year
with
WR/WN
WR/WU
R/E
1979
0.366
0.855
1980
0.448
1981
of
education)
Narrow Definition of Eligibility
Measure
Year
12 years
Measure
WR/WE
AR/AE
R/E
WRIWE
AR/AE
0.595
0.509
0.492
0.677
0.597
0.579
0.974
0.686
0.598
0.572
0.777
0.695
0.666
0.3?6
0.875
0.609
0.525
0.500
0.677
0.600
0.572
:?~~
0.412
0.091
0.694
0.600
0.577
0.707
0.690
0.663
1983
0.366
0.835
0.675
0.591
0.562
0.138
0.649
0.615
1984
0.306
0.005
0.665
0.544
...——
_O-.,?
06
0...!..-..,,._
513
0.580
0.547
Averaqe
0.379
0.874
0.654
0.561
0.536
0.727
0.635
0.607
TABI,E4. 4-W
Measures
of
Unemployment Insurance
Utilization by Year
(Measured as deviations from the mean for a 25 year old with 12 years Of ed”catio”)
Broad Definition of Eliqiblity
WarcOw Definition of Eligibilit~
Measure
Measure
Meaeure
N
m
a
Year
WR/WW
WR/W
1979
0.094
1980
R/E
WR/WE
AR/AE
R/E
WR/WE
AR/AE
0.711
0.372
0.268
0.266
0.633
0.488
0.483
0.117
0.775
0.430
0.326
0.315
0.659
0.501
0.492
1981
0.109
0.744
0.432
0.316
0.302
0.683
0.512
0.492
1982
0.126
0.766
0.473
0.349
0.332
0.142
0.564
0.536
1983
0.124
0.806
0.526
0.428
0.399
0.698
0.581
0.546
1984
0.113
0.698
0.615
0.478
0.461
0.361
0.346
Averiqe
0.114
0.750
0.476
—-—
tion group v.here there is no apparent relationship
between the insmed rate and age. Not
surprisingly, women ha~,e lower rates than men.
Examining
the other measures of utilization does not change the story for men in terms
of the in~uence of demographic
For these other quantities,
exception
rather tha
According
the positive relationship
between age ““red utilization
is nou. the
the rule.
to the findings in Tables 4.4-M
&aracterizing
and 4.4-W,
the nature of the time trends
utibzatimr depend on the measure choosen as a ~ide.
of insured unemployment
accounting
%.ariables on UI use, but it do= cloud the picture for w-omen,
follows a downward path stating
For men, the fraction
in 1980, but the other mewures
for eligibility based on either definition show essentially no trend. For womexl, on
tke other hand, there is no apparant time pattern conveyed b~- either the fraction of insured
unemployment
ehgibility.
or the other measures of UI utilization
of
How-ever, the quantities based on the broad definition indicate a strong upward
trend in UI utilization
4.6
based on the narrow definition
Comparison
among women over the period covered.
with Findings
in the Literature
Several recent studies provide
a duable
context
for interpreting
results presented abo~,e. The studies of Burt less (1983),
and evduati~lg
the
Burtless and Saks ( 1984), Corson
and Nicholson (1988) and Blank and Card (1988) examine trends in insured, ehgible and total
unemployment
covering the 1980’s. M’bile data limitations
of the measur=
of important
of eligibility md utilization analyzed here, these studies document
patterns that serve u useful guides for identifying
youths are representative
The most frequently
the proportion
did not permit examination
of the average unemployed
cited me=ure
of the unemployed
beginting
whether the experiences
the utihzation
regular UI benefits.n
measure over the last 50 yems shows a general downw=d
with a shmp downturn
a number
of
worker.
used to aamine
receiting
of dl
in 1981. For exmple,
of the UI system is
An ~miuation
of this
trend beginning in the emly 1960’s
dufing the 1940’s =d
1950’s the
6 This memnre does not count individlltis receiving=tcnded benefitsor workerscovered underspecial UI
programs m U! recipients. The special programsare the Unaployment Compensation for Federd Employe= (UCFE), UnemploymentCompensation for Ex-servicemember(UCX) and the Rtilroad Unemployment
insurance program. While individuals who file dtims under the UCFE or UCX progrm must satifiy the
same set of qudifietion requirementsu other claimants, they are not eligible for regular UI.
29
ratio was approximately
0.50, it declined to 0.44 in the 1960’s, fell to 0.40 over the 1970’s, and
during the 1980’s it ranged from 0.44 in 1980 to 0.30 in 19S4. }Vtile one can attribute
gradud
dechne throughout
composition
the 60’s. md 70’s to \-ariations “in the demographic
of the unemployed,
somewhat unexplained.
1980’s to unobserved
the
and industrial
the recent drop in the fraction of insured unemployment
is
All the studies just cited associate at,least hdf of the decline in the
danges
or in State administrative
either in the propensity
of individuals
to co~ect entitlements
practices.
Another measure of UI utilization exmined
by Corson and Nichols6n (1988) and BlaIlk
and Card (1988) is the ratio of the average weekly number of UI recipients under dl programs
to the total number of unemployed.
This quantity
displays a similar time pattern to the
more restrictive measure above, except a slightly steeper define
Specifically,
Nicholson
this ratio dropped
from a *-alue of 0.51 in 1980 to 0.34 in 1984 (Corson
(1988) p. 9). Almost dl of this drop-off
policies relating to extended
is observed iI1 the 1980 ‘s.
a,ld
appears to wise fmm changes in Federal
benefit programs and the reduction in the receipt of regular UI
benefits.e
Blad
and Card (1988) further exmine
and total or insured unemployment.
proportion
of the unemployed
ngs information
individual’s
the rdationship:
between eligible unemployment
One measure that Blank and Card invatigate
eligible to receive UI benefits under au programs.
from a previous year a~-ailable in the Mach
ehgibility to receive
is the
Using earni-
CPS to infer an unemployed
UI, they find that the &action of ehgible unemployment
re-
mains roughly constant over the 1980’s ranging from 0.50 in 1980 to 0.40 in 19”84. Coricernillg
the relationship
between inswed
pattern of the tale-up
biting
and efigible unemployment,
rate defied
as the ratio of insured to efigible unemployment.
CPS data with administrative
that there was a si@ficmt
decre=e
Blank and Card aplore
data on the nmnber of UI recipients,
the
Com-
they conclude
in the take-up rate over the 1980’s. 10
for extendedbenefits,
eliminated
the
s In 1981 the U.S. Congress tightened
the eligibdity
standards
national
insured
unemploymentrate trigger foI extendd benefits md incre-ed the State trigger rate by I
percent. In addition, the FederdSupplematd Benefit
programthatw= enacted in responseto the recession
in 1982 w= not x generous= the Federd SupplementalCompensationprogram en=ted during the 1974-75
recession.
III co,,verse~y, ~ sample Of “ntmPioYmentsP,lls from thePanel Sttldy of Incom= Dynamics anal~~edb?
B1-k and Card show incre=hg take-up rates from 1980 through 1982.
30
Witk
regard to relating our findings on the patterns of UI utilization
literature,
there is considerable
agreemeIlt.
Comparing
the aggregate
to those in the
trends presented
in
Corson aIld Nicholson (1988) md Blank and Card (1988) with our results reported in Tables
4.4 indicates
that the experiences
of the population
data.
at large. This is especially true for the measures b~e’d on administrative
For example,
comparing
of insured unemployment
similarities.
of young workers in the 1980’s are ftirly representative
the measure ltrR/ltV
under d
While the magnitude
it describes the behavior
are almost identicd.
UI progrms
from Table 4.4-NI with the fraction
in Corson and Nicholson
reveals striking
of our utilization measure is tigtificantly
higher because
of what is essentially a young prime-age
rode,
the time patterns
From 1980 to 1984 both the aggregate measure and our value declined
by 17 bwis points i.e., 0.51 to 0.34 and 0.974 to 0.805 respectively.
On the other hand, our findings b~ed
on imputed
different patterns than those put forw-ard in other work.
decreasing
ployment,
tale-up
Specifically,
Blank md
conx.ey
Card find
rates and relatively constant measures of the fraction of efigibIe unem-
w,hile the WR/U-E
oppOsite patterns.
measures of UI &gibility
and li’E/liV
measures in Tables 4.4 and 4.2 exhibit just the
Further researck is needed to reconcile these disparities.
5. A General
Estimatio-n Approach
This sectionconsidersseveralissuesrelevantin designingan empiricalmodel thatenables
one to measure the influence of U1 policies on the duration of unemployment.
analysis
not only provides
framework for integrating
5.1
Altematiue
a bwis for the subsequent
empirical work, it dso offers a useful
and evdrrating other empirical
Estimation
The follo~~.ing
findings
in the }~t~r~~~re
Schemes
To characterize the *-arious procedures for analyzing the amount of unemployment experiencedby individualsover some time horizon,the discussionbelow relieson the following
definitio]ts:
u=
weeks of unemployment;
R=
(B, T) =
B=
rules of a U1 program that define individuals’
T=
rules of a UI program that determine the taxation scheme used
UI policy regime;
UI entitlements;
to finance the program;
(5.1)
E=
~
H=
w,ork history;
indicator of UI receipt;
6=
z=
If
demographic
=
PA =
f(wlx)
entitlement variabla;
=
characteristics;
macroeconomic
population
vmiables;
attributes
density or probability
condjtiofing
consisting of elements in H, Z
function
a subscript “i” to a mriable d=ignates
serwtion
refers to a mriety of potential occurrences
or some bed
time interd.
&Ild
X.
.thObser=tion of this variable.
the z
U: to measure the number of weeks of unemployment
or the total number of weeks of reemployment
flf;
of the variables w
on the variable
Attacting
and
of unemployment.
An Ob-
Thus, one may specify
that occurs in a SPCHof unemployment,
occuring within either a nonemployment
spell
The variable Ri designates the rules of the UI program applicable
32
when Ui occurs,
The T component
of R encompasses such features as the experience-ratings
rele~.ant for firms in making”their contributions
R determinm
tile values of E wsigned
into the UI system, md the B component
to indi~,idtials according
of
to their u,ork experiences
H.
The vmiables Ei include the weekly benefit amount and the number of weeks of U] efigibiiity
that m individud
dimensions
qualifies for w,ho experiences
of the individud’s
6~ equals 1 if the individud
variables
Zi provide
pwt work aperience
experiencing
information
time of Ui, md Jfi incorporates
of unemployment
Knowledge
The attributes
Hi summarize
various
when Ui occurs.
The indicator
v=iable
Ui.
Ui collects UI benefits and equals O otherwise.
on the demographic
characteristics
variables capturing exogenous
of the person at the
m~roecmromic
determinants
durations.
of the distribution
choice of the conditioning
~ (U IR, PA) for a judicious
ables PA provides the principal information
needed to ~sess
aspects of UI programs on the extent and the composition
vmiables
the consequences
of unemployment.
Z among the covariantes PA specifies a distribution
group; inclusion of H in PA determines a distribution
rating Af in PA adtits
adjustments
Inclusion of the
for a particular demographic
for ~orrs
for macroeconomic
characterized
vari-
of changil~g
worker types; and incorpo-
conditions.
Fitting the distribution
~ (UIR, PA) determines how U varies w one alters policy instruments incorporated
a population
The
by the attributes making up the covmiates
in R for
PA.
Estimating ~ (U \R, PA) is not an ~sy task for tw,o reasons. First, there is no simple u-ay
to quantify R. Programs differ quite substantially
weekly benefits
and we&s
set of explanatory
mriables
of ehgibility,
arises
from
constmt
par-eters
accounting
of unemployment
the due
of R.
differences in UI pohcies across states.
states’ populations
poficy
= one mies
dso
differ, one typically
R occur
simultaneously
for these composition
R. As a consequence
chages
of the difficdties
irrdi%,iduals’
these rules are not readily summarized
that wry along some continuous
the effects of R on the distribution
composition
md
in theh rules for determining
The primary
Recognizing
dmges
resdts in indd
involved in obttining
33
Second, estimating
requires one to hold a population’s
encouters
tith
spectrum.
by a
source of mriation
in R
that the characteristics
a situation
of
in which shifts in tke
in popdation
composition.
Not
inferences about the influence of
a direct estimate of f (Ul~
PA),
empirical research on the effects of UI on unemployment
Specifically,
utilizes other estimation approaches.
this research tends to focus on estimating some variant of the distributions:
f (U16 = 1, E, PA),
and f (U/E,
f (U16, PA),
P.4),
Studies adyzing
program
data from
State UI offices (e.g. hfoffitt (1985), Meyer (1988), Katz and Meyer (1988a, 1988 b)) estimate
a specification
duration
of ~ (U16 = 1, E, PA),
of unemployment
corr=ponds
and an environment
U me~ures
characterized
number of weeks of UI receipt.
such sources as the CPS, PSID or NLS ~timate
Empirid
to, the distribution
describing
for a UI recipient who qualifies for UI entitlements
comes from a population
such analysm
whid
analyses comparing
the huard
rat=
the
E and v,ho
by the attributes
PA — in
Studies using survey data from
variants of f (U16, PA)
md
f (UIE,
PA).
of UI recipients md non-UI recipients (e.g.
Katz (1986) and Katz and Meyer (1988 b)) in msence explore differences in the distribution}
f (U]6, PA) when 6 = I and 6 = O. Other work concerned
entitlements
on unemployment
rely on some specification
5.2
Assess?ng
witk predicting
(e.g., Clink and Summers (1982a)
and Topel (1983, 1985))
of ~ (LTIE, PA) as the basis for their estimation,
the influence
of U1 on Unemployment
A key question that has gone unanswered in the hterature cmrce~
about the distribution
a shift in th~e
is typically
w-hat can be learned
PA) from estimated variants of the densities j (U16 = 1, E, PA),
~ (UIR,
f (U16, PA) and J (UIE,
the effects of VI
PA).
k
particular, if one finds that higher UI entitlements E impl}
of U indicating
distributions
greater unemployment,
can one conclu de,, as
done in the literature, that a more generous UI policy regime will lead to more
unemployment?
The answer to this question is no unless one incorporates
measures of work history mtiables
H to serve as controls among the comriates
To ensure that estimated effects =sociated
pretationtypidy
the appropriate
PA.
with UI entitlement mriables have the inter-
given to them in the hteratwe, the tiation in ~ admitted in empirical
specifications
must reflectpurely differences
in pohcy re~mes.
qtite substantially
in te-s
While ~-1programs differ
of the roles they apply to detertine benefits,
dl programs de-
finebenefitsusing informationon only a few =pects of a p=son’s recentwork history.As
outlinedin Section 2.1,these =pects include.such items as base period earfirrgs (BPE),
high quarteI
emnings
(HQE),
average wee~y
34
arnings
(AWE),
the circumstances
under
which employment
terminated
covered by the UI system.
(e.g. quit, fire, etc. ), and whether pre~.ious employment
Incorporating
three essentially exists a functional
W-as
these item> among the work history *-ariab]es H,
relationship
linking H and a UI policy regime R to VI
entitlemexlts E. k particular, one can specify functions of the form
(5.2)
E=
which show. how a person’s
UI weeMy benefit amount and weeks of eligibility are assigned
@(H,
Jf,
R) = @(H,
given this individud’s
past v.ork experiences
of the macroeconomic
vtiables
some program featues
11, B),
and the rules of the UI program.
Jf ~ arguments of the function
that only the B component
the od}
source of variation in E of interest for drau-ing inferences about
the influence of UI poficies operates tkougk
the regime \ariables R or B. Inspectioxl of tke
@ highlight the point that E vari= across observations
shifts in R, but dso as a consequence
possibly due to changes in dues
~-ariables PA in the distributions
one ~n
of chages
The importance
interpret estimated effects associated
E.
However,
of UI POECY. These policy
Sur\,eys of this topic (e.g.
(1981))
in the empirical lit~ature
Welch (1977),
discuss a nriety
of not capturing the appropriate
while virtually
the covariates to obttin
effects has long been recognized
Danziger, Haverman and Plottick
present as a consequence
with
of considering individuals covered b“y different state programs
of includzng work history mriables -ong
unemployment.
of
in UI poficy over time.
able estimates of entitlement
UI ad
f (U16, E, P.4) or
to differences in the generosity
entitlements= reflecting
responsesto varying the daracteristics
or as the rmdt
in H and Jf appearing
variables included
in E may be attributed
Under such circumstance=,
shifts afise as the consequence
and
of 14 either across states or over time. If one incorporates
in (5.2) as elements of the conditioning
f (U lEj P-4), then dl variation
in a sample not only due to
of differences in the work histories of individuals
the group of work history md macroeconomic
UI systems.
of
the entitlements of a system.
In intimation,
functions
@ accounts for the fact that
such as extended benefits depend on the levels of state unemployment
rates. The second expression for @ given h (5.2) recognizes
R determines
The inclusion
Haruermesh (1977)
of possible Klases that tight
relion
md
be
source of variation in the nriables
d] empirical studies account for some measure of H in their
35
analyses, none of w,hich we are aware includes the specification
of cent rols needed ill theory
to purge E of variwbifity other than that due to shifts in ~.
One finds a variety of w,ork history variables incorporated
.
allalyies of tll
effects. The most popular choice for H mnsists of a single measure of an iidividud’s
weekly earnings (A W7E). Researchers typically
.
in empirical
a ‘wage
capture
replacement
the notion
ratio”,
introduce
of ordy AM’E
to detefine
is incomplete.
even more setious sourc=
is introduced
cost more accmately.
history variables in ad”&tion “to-AU’E
H consisting
enter AMTE through some representation
and quite often Alt7E
of opportunity
in an after-tax
Consequently,
quantities.
alldysis
as base period or high quarter eartings
go into the determination
complement
progras.
of nriables
~rithout
is likely to
It is not uncommon
many work history tiabies
such qumtities
of
because UI benefits are based on
find empirical studies incorporating
- including
form to
a specification
The use of an aftm-tax form of AWE
of bias in estimation
of
All UI system”s use many work
benefits.
vdrres of A M’E rather than on titer-t=
the before-t=
a~,erage
other than AI!’E
to
in their
wkich actually
of entitlements - but we know, of no attempt to account for the full
and interactions
accounting
needed to characterize
the benefit structure of UI
for this structure, UI entitlement and rectipt variables E and
6 in part perform the task of identifying worker types, with higher values of E and 6 signifying
those types v,ho experience
incorrect
more unemployment.
Such an occurrence
inferti”ces about the role of these variables as determinants
in principle leads to
of unemployment.
Of
course, the inclusion of only a subset of the relevant work history variables may be sufficient
to avoid any serious bi=es
5.3
Measuting
Predicting
mulation
qumtity
f (UIR,
the Impact
of Shifi
comprehensive
of the distribution
in UI Policy
effects of UI policies
f ( UI R, PA).
PA)
consists of combiting
which
on unemployment
As noted above,
involves a number of compficatimrs.
~ (51E, PA),
tribution
in estimation.
non-UI recipient popdations
approach
on estimated tiants
represent the types of distributions
~ (U 16, E, PA) indicates the extmt
the direct estimation
A more attractive
information
according to levels of. UI mtitlements.
of this
for constructing
of j (U16, E, PA),
anrdyzed ti the Eterature.
to which rmaployment
36
requires some for-
and
The dis-
differs across UI aud
The divergence between
f (UIJ = 1, E, PA) and j(U[6
tributable
to participation
the probability
= O, ~, PA) offer, a memure of the shift in unemployme,lt
in LTIprograms.
The second distribution
that indi$.iduals characterized
facing vdrres of entitlements
by attributes
f (61E, PA) determiIles
P.4 become
UI and non-UI recipients is small, a modification
in a UI progrn
if it results in a blg adjustment
of UI participation.
Developing
distributions
the relationship
~(UIR,
The summation
continuous
that enables one to construct
requires several steps. According
(5.3)
PA)=
~
f(U]J,
6,E
to ftiliar
E, R, PA)
COUId have a largeefect
~ (UIR, PA) from these other
j(61E,
R, PA)
f(E]R,
PA).
are discrete -“” if tbe~ w-ere
an integral sign would be used instead — with the summations
represents the joint distribution
all the *.=iables
5 and E. The right-hand-side
of formula (5.3) merely
of the variables Lr, 6 md E condition
simplification
occurs
tke Enkage relating entitlements,
functions (5.2 ). According
H, the functional
in tkis representation
on R ~d
of f ( U[R, PA)
work history and policy regim~
to these functions,
PA, wi tk
if one fully
conveyed
by the
as long as one includes sufficient information
in
relationship linking ~, H, 1{ md B given by (5.2) means that. know-ledge
of E and H eliminates the ned
distributions,
carried out over
other than U integrated out.
A substantial
avoid estimating
between
rmilts in statistics,
sign used here assumes that rdl distributions
the admissible range of the mriables
uploits
UI recipients when
equal to E. Even if the difference in unemployment
in the probability
at.
the distributions
to know B. This obsermtion
appearing in formtia
(5.3),
a~ow,s one to simpIify or to
Mrith respect to the first two
one obttins
~(U16, E, R, PA)=
f (U16, E, T, PA)
(5.4)
f (61E,
One -
eliminate B as conditioning
dl the essential information
substantidy
R, PA)=
f (51E,
mriables
T, PA).
because E, H md M implicitly
in B. The ability to i~ore
B in specifying
sum~ize
these distributions
reduces the problem of estimating them because one need not tackle the difficdt
task of quantifying
B.
Concerning
the third distribution,
(5.3), there is not even a need to estimate this quantity.
37
~ (E IR, PA),
Knowledge
appe~ing
in formula
of H md R determines
.
E exactly.
Alternatively,
one can express this result as
(5.5)
f(EIR,
where the function
Combining
P44) = Q(H,
Jf
R)
Q is given by (5,2).
these results, provides the relationship that permits construction
horn distributions
off
that are more readily analyzed in empirical work. Substituting
(UIR, PA)
(5.4)-(5.5)
into (5.3) yields
(5.6)
f(UIR,
PA)=
~
f(U16,
6,E
O,,e can estimate specifications
using micro data.
eration.
f(ulR,
5.4
of the distributions
The functions
Formda
E; T, PA) \(&]E,
T, PA)
11, R).
f (U16, E, T, PA) and f (61E, T, PA)
$ are knov-n depending
(5.6) shows how to combine
@(H,
thae
on the UI policy
under consid-
quantities to compi”te in estimate of
PA).
An Alternative
Formulation
When the variable U me~ures
the accumulative
number of week
of uxlemploymexlt that
occur over some period of time - which is the type of measure analyzed in the subsequertt
empirical work - it is not convenient for estimation
etrization
of the distribution
that a standmd
market activities
f (U16, E, T, PA) appearing in formrda (5.6).
duration model describes spells of unemployment
as well, then the implied specification
weeks of unemployment
over a time hotizon)
To avoid such complexities in devdoping
tributions for these separate components.
as unemployment.
md ~ (p It, 6, E, T, PA),
(5.7)
~ (U16, E, T, PA)=
U into two components
~
~ufl
f(plt,
38
and specifying the dis-
U = PI where t = the length of
k particular, defie
infer the distribution
~
!=1
of U (i.e., of total
a specification for the distribution f (U16, E, T, R4),
From the two condition
one cm
and spdls in other labor
for distribution
time horizon over which total nmremployment
of t classified
If one presumes
is quite complex,
u attractive alternative involvm decomposing
the rdemnt
purposes to work directly with a param-
is meuured,
and p = the fiactioll
distributions
~ (Z[6, E, T, PA)
associated with U via the formula
6, E, T, PA)
f (L16, E>-~,
~A),
The quantity f (116, E, T, PA) represents a conventional
the spell length t; and we refer to ~ (p/f, J, E, T, PA)
because it characterizes
duration distribution
that describes
as a time-proportion
the portion of a duration [ spent in a particular status.
39
distribution