Workfare versus Welfare: Incentive Arguments for Work Requirements in Poverty - Alleviation Programs American Economic Association BESLEYANDC OATE:W ORKFARVES . WELFARE THE AMERICAN ECONOMIC REVIEW

1. American Economic Association
Workfare versus Welfare: Incentive Arguments for Work Requirements in Poverty-
Alleviation Programs
Author(s): Timothy Besley and Stephen Coate
Source: The American Economic Review, Vol. 82, No. 1 (Mar., 1992), pp. 249-261
Published by: American Economic Association
Stable URL: http://www.jstor.org/stable/2117613
Accessed: 05/11/2010 14:34
Your use of the JSTOR archive indicates your acceptance of JSTOR's Terms and Conditions of Use, available at
http://www.jstor.org/page/info/about/policies/terms.jsp. JSTOR's Terms and Conditions of Use provides, in part, that unless
you have obtained prior permission, you may not download an entire issue of a journal or multiple copies of articles, and you
may use content in the JSTOR archive only for your personal, non-commercial use.
Please contact the publisher regarding any further use of this work. Publisher contact information may be obtained at
http://www.jstor.org/action/showPublisher?publisherCode=aea.
Each copy of any part of a JSTOR transmission must contain the same copyright notice that appears on the screen or printed
page of such transmission.
JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of
content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms
of scholarship. For more information about JSTOR, please contact support@jstor.org.
American Economic Association is collaborating with JSTOR to digitize, preserve and extend access to The
American Economic Review.
http://www.jstor.org

2. Workfareversus Welfare:IncentiveArgumentsforWork
Requirementsin Poverty-AlleviationPrograms
By TIMOTHY BESLEY AND STEPHEN COATE*
Manywould agree with SamuelJohnson
that "a decent provisionfor the poor is a
true test of civilization."Yet still there is
muchdebateaboutwhatformpovertyrelief
shouldtake.Particularlycontroversialis the
claim that recipients of relief should be
requiredto workin exchangeforbenefits.
The use of workrequirementsin poverty-
alleviationprogramsis of widespreadsig-
nificance. Perhaps the most notorious
historical example is the English system,
institutedbythe PoorLawof 1834,inwhich
poor relief was grantedthroughresidence
in a workhouse.Workfarewasalsocommon
in ancien regime France,where relief was
grantedin "charityworkshops."But work-
fare schemes are by no means just relics
fromthe past.Theyremainpopularin both
developedand less developedcountriesto-
day. In the United States, for example, a
numberof states now demandthat welfare
claimantsenrollin eithera trainingor work
programin orderto receivebenefits.1Simi-
larly,currentpracticein Indiarelies heavily
on public-worksprojectsas a tool for pro-
vidingpoorrelief.2
Thereare manyargumentsmadein favor
of workrequirementsin poverty-alleviation
programs.One consistenttheme, however,
is that workrequirementsserve to provide
the "appropriateincentives"for recipients
of poor relief. Thus, this paper providesa
detailed explorationof the incentive case
forworkfare.We analyzetwodistinctincen-
tive arguments:a screeningargumentthat
workrequirementsmayserveas a meansof
targetingtransfersanda deterrentargument
that they may serve as a device to encour-
age poverty-reducinginvestments.Familiar
notions from the economicsof incentives3
are shownto provideinsightsinto modern-
daysocialpolicyandhistoricaldebates.
The screeningargumentis motivatedbya
desire to direct poor support toward the
trulyneedy. In developingeconomies,it is
typicallytoo costlyforthe governmentto set
up a sophisticatedadministrativemachinery
to determinewhether a particularindivid-
ual is in need of poor support.Even infor-
mation about individuals'incomes is un-
likely to be availablein this way. In such
situations, it may be better to make the
relief system self-targetingby laying down
conditions for claiming support such that
only the trulyneedy present themselves.A
workrequirementis one suchtest. As Jean
P. Dreze (1990) makes clear, this logic lay
behind British administrators'reliance on
publicworksto relieve faminesin colonial
India. Moreover,as confirmedby Michael
B. Katz(1986),it was also a commontheme
in earlyargumentsforworkrequirementsin
the United States.
There may even be a screeningrole for
work requirementsin developedcountries,
where the administrative infrastructure
makesit easier to assess the circumstances
*Woodrow Wilson School, Princeton University,
Princeton,NJ 08544,and Departmentof Economics,
University of Pennsylvania, 3718 Locust Walk,
Philadelphia,PA 19104,respectively.We are grateful
to manypeopleforhelpfulcomments,discussions,and
suggestions.We particularlythankouranonymousref-
erees, DavidBradford,Anne Case,SandyKorenman,
JimMirrlees,Bill Rogerson,SharonTennyson,Martin
Weitzman,and RichardZeckhauser.They bear no
responsibilityfor the product.This workbeganwhile
we were attendingthe summerworkshopat Warwick
UniversityinJuly1988.Wearegratefulforthesupport
and hospitalityof the Departmentof Economics.We
alsogratefullyacknowledgethefinancialsupportof the
John M. Olin Programfor the study of Economic
OrganizationandPublicPolicyatPrincetonUniversity.
'For some recent discussionof such programssee
JudithM. Gueron(1990).
2See, for example,MartinRavallion(1991),which
considersthe screeningpotential of the Maharastra
EmplovmentGuaranteeschemein India.
3See, for example,Jean-JacquesLaffont(1990)for
an overviewof the main ideas on whichour analysis
draws.
249

3. 250 THEAMERICAN ECONOMIC REVIEW MARCH1992
of individualson a case-by-casebasis.While,
in contrastto LDC's,it maybe possiblefor
the governmentto attaina reliableestimate
of an individual'searnings,it is not always
possible to observe his earning opportuni-
ties. Thus,the governmentmaynot be able
to tell whether an individualhas deliber-
atelychosennot to workor, moregenerally,
reducedhis workhours in orderto qualify
forbenefits.We shallinvestigatethe screen-
ingargumentsforworkrequirementsinboth
the developing-anddeveloped-countrycon-
text, pointing out the differencesbetween
them.
The deterrentargumentforworkrequire-
mentsfocuseson the originsof poverty.Are
individualspoorjust because they have ex-
perienced bad luck or because of choices
made earlier in life? If the latter is true,
then publicassistancemaylead individuals
to makechoicesthatincreasethe likelihood
that theywill haveto drawon suchsupport
in future.This old idea has emergedagain
in recent discussionsof U.S. social policy.
Conservativeshave argued that increased
expenditureon social programssince the
1950'shas createda greaterdependencyon
state support.Morespecifically,it has been
suggested that welfare programshave re-
duced individuals'incentivesto acquirethe
human capital necessaryto avoid poverty
andmayeven haveled to irresponsiblepar-
entingdecisions.4
To avoidthis problem,it is argued,poor
relief must be made relativelyless attrac-
tive. One way of doing this is to impose a
workrequirement.This logic clearlyunder-
pinned the 1834Poor Law Commissioners'
proposalto place the poor in workhouses.
The idea was that the "conditionof the
able-bodied pauper be 'less-eligible'-de-
sirable,agreeable,favorable-than that of
the 'lowestclass'of independentlabourer"
(GertrudeHimmelfarb,1984 p. 163). The
Poor Law regardedthis as essential:"'Itis
only ... by making relief in all cases less
agreeablethanwages,that anythingdeserv-
ing the nameof improvementcan be hoped
for" (Himmelfarb,1984 p. 165). The same
argumentcan be found in both historical
and contemporarydiscussionsof U.S. wel-
farepolicy.5
The purposeof our analysisis to explore
these two argumentsfor workrequirements
in greaterdetail.We use a simplemodelto
illustratetheir logic and to bring out the
assumptionson which they depend. Along
the way, we are also able to draw some
conclusions about the optimal design of
workfare programs under various condi-
tions.
The idea that work requirementscan
servea screeningrole hasbeen notedprevi-
ously. Albert L. Nichols and Richard J.
Zeckhauser(1982), for example, mention
that the impositionof "ordeals"on welfare
claimantsmayimprovetargeting.GeorgeA.
Akerlof (1978) also points to the value of
manpower-trainingprograms in targeting
benefitsmore accurately.However,neither
of these papersprovidesanydetailedanaly-
sis of the screeningargumentfor workfare.
We characterizethe optimalworkfarepro-
gram for screening purposes and give a
sufficientconditionfor this to be less costly
thanwelfare.A furtherdifferenceis thatwe
concentrate on poverty alleviation rather
than welfare maximizationas an objective.
In lightof the focus of the policydebate in
this area, this seems entirely reasonable.
We do, however,discussthe differencebe-
tween these two approachesbelow. As far
as we know, the deterrent argumentfor
workfarehas not been exploredpreviously.
The structureof the paper is as follows.
The next section presents the basic model
which providesthe frameworkfor our dis-
cussion.SectionII providesa benchmarkby
describing the optimal poverty-alleviation
programwhen individuals'earningabilities
are both observableto the policymakerand
beyondindividuals'control.SectionsIII and
IV analyzethe screeningand deterrentar-
gumentsfor workfare.Section V discusses
4See,forexample,the controversialworkof Charles
Murray(1984).DavidEllwood(1988)andEllwoodand
LawrenceSummers(1986)providedissentingvoiceson
thisissue.
5See Katz (1986) for the historicaldiscussionand
MickeyKaus(1986)for a morerecentargumentalong
these lines.

4. VOL.82 NO.1 BESLEYANDCOATE:WORKFAREVS.WELFARE 251
extensions, and conclusionsare presented
in SectionVI.
I. TheModel
In our analysis,we follow John Stuart
Mill'scharacterizationof the poverty-allevi-
ation problemas "howto give the greatest
amountof needful help, with the smallest
encouragementto undue reliance on it"
(Mill, 1848p. 334).We capturethis by sup-
posingthat the governmentis concernedto
ensurethat each individualgets a minimum
income level, denoted by z, at minimum
fiscal cost.6 Thus, it wishes to assist only
those who would earn less than z without
intervention,while preservingincentivesfor
individualsto make choices that will put
themin a positionto earnmorethan z.
We considera populationconsistingof n
individuals,dividedintotwotypesaccording
to their income-generatingability, a E
{aL,aHI,whereaL <aH andwhereH stands
for highand L standsfor low.A fractiony
hasabilityaL.Inwhatfollows,we shalltake
aL and aH to be individuals'wage rates.7
Each individualhas identical quasi-linear
preferencesdefinedoverincomey andwork
1. Thus, utility is given by y - h(l) where
h(*)is increasingandstrictlyconvex.8
Throughoutthe analysis,we shallset aside
the revenue-raising implications of the
budget required to finance government
transfers, in order to focus directly on
poverty-alleviationissues.We havein mind,
therefore,a world in whichthe individuals
consideredhere form a target population,
expenditureson whomarefinancedbytaxa-
tion of the remainderof the populationor
by aid flows.
A poverty-alleviationprogram(PAP) is a
pair of benefit packages{b,Ci}i=L,H where
bi denotes a cash transferfor individualsof
abilitytype i and ci denotes a cost in terms
of a public-sectorworkrequirementneeded
to obtain this transfer.In order to focus
exclusivelyon incentiveargumentsforwork-
fare, we will assumethat the workdone in
the publicsectoris unproductive.9Thus,the
cost of the programto the governmentis
just n[ybL+(1- y)bH]. The government's
objectivewill be to minimizethis cost sub-
ject to the constraintthat each individual
obtainsan incomeof at least z.
Individuals of ability ai must choose
whetheror not to claimthe benefitpackage
{bi,cC}that is intended for them. Even if
they do so, they may continue to supply
some laborto the privatelabormarket.Let
l(b, c,ai) denote the private-sectorlabor
supplyof an individualwith wage rate ai
who accepts a package{b,c}. It is easy to
showthat
(1) l(b,c,ai)
I(ai) - c if c <l(ai)
0O otherwise
where l(ai) is the amount of labor that
would be suppliedto the privatesector in
the absenceof anyprogram[i.e., h'(l(ai))=
ai]. Thus, a workrequirementsmallerthan
l(ai) would cause an equal reduction in
private-sectorlabor supply, while a work
requirementin excess of (ai) would cause
the individualto cease private-sectorwork
altogether.Note from(1) that laborsupply
is independentof b (i.e., thereis no income
effect). This inessential, but analytically
convenient,simplificationallowsus to write
the individual'sprivate-sectorlabor supply
as l(c,ai).
6
A similarapproachis taken in Ronald Dye and
RichardAntle's (1986) analysisof in-kindtransfers,
althoughtheyspecifytheirobjectivein utilityterms.
7Here,we do notconsiderthe possibilitythatwages
changebecauseof the supplyeffectsof workfarepro-
grams.Thisis studiedin an LDCcontextbyRavallion
(1990).
8This simplifiesthe analysiswithout substantially
alteringthe characterof the results.
9Workfareprogramshaveboth costs and benefits.
The formerincludecostsof equippingandsupervising
workers,while the latter include direct provisionof
servicesand benefitsto claimantsin the formof en-
hancedproductivity.Forfurtherdiscussion,see Gueron
(1990).

5. 252 THEAMERICANECONOMIC REVIEW MARCH1992
Given this labor supply,the individual's
total private-sectorearningswillbe
(2) y(c,ai)
_ai(l'(ai)- c) if c <l(ai)
-
O otherwise
andhe will enjoya utilitylevel
(3) v(b, c, ai) b + y(c, ai)
- h(l(c,ai) + c).
The individualwill voluntarilytake up the
package{bi,ci}intendedfor himif andonly
if he is better off so doing, that is, if and
onlyif v(bi,ci,ai) 2 v(O,0,ad).10
To makethe probleminteresting,we shall
assume that only one group in the target
population(typeL) is poorwithoutgovern-
mentintervention.Hence,we assumethat
(4) y(O,aH) > z > y(O,aL).
We next examine the optimal PAP for a
benchmarkcase.
II. PovertyAlleviationwithObservableand
ExogenousAbilities
To establish a benchmarkcase for the
subsequentanalysis,we consider the opti-
malPAP when individuals'earningabilities
are observableto the policymakerand be-
yond individuals'control.In this case, it is
straightforwardto ensurethat each individ-
ual receives the benefit package designed
for him.However,the policymakerdoes not
have complete freedom in his choice of a
PAP. As policymakers in 19th century
Britainand Indiafoundto theircost, if the
PAP is made too arduous,then even the
verypoorestmaychoose not to participate.
Thus, the PAP must offer individualsat
leasttheirno-interventionutilitylevels:par-
ticipationmustbe voluntary."
The policymaker'sproblem is therefore
to choose a PAP thatminimizesthe costsof
poor relief and satisfies two constraints.
First,individualsmustbe willingto partici-
pate in the program: v(bi, ci, a ) > v(0, 0, a )
for i = L,H. Second, those with wage rate
aL mustescapepoverty:bL + Y(CL, aL) 2 z.
We canignorethe constraintthathigh-wage
individualsget at least z, since voluntary
participationimplies that they get an in-
come at least as great as they would have
hadwithoutintervention.
The solution to this problemis straight-
forward.High-abilityindividualsshouldbe
givenno governmenttransfer,sincetheyare
alreadyearningmorethanthe povertyline.
Low-abilityindividualsshould be given a
transferequalto the differencebetweenthe
povertyline and their private-sectorearn-
ings [whichis z - y(0,aL)],but shouldnot
be requiredto workfor this transfer.As (2)
makes clear, imposinga workrequirement
wouldonlyreducetheirprivate-sectorearn-
ings and increase the gap between their
income and the poverty line. This would
increasethe transfernecessaryto get them
out of povertyand increasethe cost of the
program.This argumentis summarizedin
the followingproposition.
PROPOSITION 1: If income-generating
abilities are observable and beyond individu-
als' control, the cost-minimizing PAP is a
welfare program (i.e., imposes no work re-
quirements). Low-ability individuals are of-
fered a cash transferthat is just high enough
to get them out of poverty, z - y(O,aL), and
high-ability individuals are offered no bene-
fits.
In the sequel, we will refer to the PAP
describedin Proposition1 as the benchmark
PAP.
'0We are assuminghere that individualsare only
able to choose the packageintended for their own
abilitytype.Thisis dealtwithin moredetailin Section
IIIof the paper.
1"Evenin economiesinwhichpolicymakersareable
to coerce individualsinto payingtaxes, it is hardto
force them into acceptingincometransfersif they do
not wishto. Hence, the requirementof voluntarypar-
ticipationseemsreasonablein mostcontexts.

6. VOL.82 NO.1 BESLEYANDCOATE:WORKFAREVS.WELFARE 253
III. The Screening Argument
Supposenow that,while the government
knows the distributionof abilities in the
population, it is unable to observe each
individual'sincome-generatingability.Thus,
it knows that a fraction y of the target
populationcan earn a wage of aL and that
a fraction1- y can earn a wageof aH,but
it cannottell whetheranyspecificindividual
is of highor low ability.It is clear that the
benchmarkPAP mightno longerbe imple-
mentable.High-abilityindividualsmay find
it worthwhileto claim to the policymaker
that they are of low ability in order to
obtainthe transferof z - y(O,aL).
For the PAP to be implementable,the
policymakermust also respect the require-
ment of incentivecompatibility,which is that
high-ability(low-ability)individualsprefer
theirown benefitpackageto masquerading
as low-ability(high-ability)individualsand
receivingthe other'sbenefitpackage.'2The
precise form of these constraintswill de-
pend on the government'sinformation.We
considertwo polarcases.The first,whichis
most relevant for developingcountries, is
that in which individuals' private-sector
earningscannotbe observedby the policy-
maker. In this case, it is possible for the
nonpoorto claim the benefitsintendedfor
the poor yet continue to work as much as
they would like in the privatesector. The
secondariseswhen the governmentcan ob-
serveindividuals'earnings.In this situation,
a high-abilityindividualwho masquerades
as poor must also reducehis private-sector
earningsinorderto passasbeingpoor.This
case is perhapsmorerelevantfor developed
countries,where the problemis often seen
as one of individualsoptingnot to workor,
more generally,reducingtheir work hours
inorderto collectwelfare.13
v(O,0,aH)
aHIl
(bL "CLaL
v(O,O,aL)
b L+a L('C L)
z L-
cL---
--
FIGURE 1. THE PACKAGE(b'L,c'L) SEPARATES
THE Two TYPES OF INDIVIDUALS
A. UnobservablePrivate-SectorEarnings
We beginwith the case in whichprivate-
sector earningsare unobservable.Since in-
dividualscan masqueradeand continue to
work as much as they like in the private
sector, the incentive-compatibility con-
straints are v(bL, CL, aL) ? v(bH, CH, aL)
and v(bH,cH,aH) 2v(bL,cL,aH). The pol-
icy problem is now to choose a PAP to
minimizecosts subjectto voluntarypartici-
pation, poverty alleviation, and incentive
compatibility.
If the policymakerhad no recourse to
work requirements, the incentive con-
straintswould necessarilyimplythat bL =
bH, so that the PAP could not discrimin-
ate between types. By imposinga workre-
quirementon those who claimto be of low
ability, however, the policymaker may
achieveself-selection.This is because high-
abilityindividualshavea higheropportunity
cost of supplyinghours of their time than
do low-abilityindividuals.Figure 1 illus-
trates the basic idea. In the absence of
benefits,low-abilityindividualsface a bud-
get line aLl and choose to work i(aL) hours.
12Thenotionof an "incentive-compatible"poverty-
alleviationprogramis analogousto the conceptof a
"self-categorizing"scheme discussed by Jonathan
Kesselman(1969p. 291).
13Itshouldbe noted, however,that even in devel-
oped countriesit is not uncommonto hearof welfare
recipientsworkinga good deal in the informalsector
(i.e., doing odd jobs for friends, casual labor, etc.)
withoutreportingtheirearnings.Indeed,recentwork
by ChristopherJencksand KathrynEdin (1990) sug-
gests that a sizable fractionof Aid to Familieswith
Dependent Children(AFDC) claimantsfall into this
category.

7. 254 THEAMERICANECONOMIC REVIEW MARCH1992
High-abilityindividualsface a more favor-
able trade-offbetween income and leisure
representedby the budget line aHl. Sup-
pose now that individualswho claim to be
of low ability are offered the package
(b'L,c'L). If a low-abilityindividualtook up
this package,he would face a budget line
with slope aL emanating from the point
(b'L,c'L). As illustrated, he would work
l(aL)- C'Lin the private sector and enjoy a
posttransferincome of z. By contrast,if a
high-abilityindividualwere to take up this
package,he would face a budget line with
slope aH emanatingfrom (b' ,c'L). This is
everywherebelowhisstatusquobudgetline,
and thus he has no incentive to mas-
querade. Thus, the PAP described by
{(b0LIC) (O,O)}is incentive-compatible,de-
spite offeringdifferenttransfersto the two
abilitytypes.
There is a cost to using work require-
ments to achieve self-selection, since the
poor will be working less in the private
sector as a consequence of the work re-
quirement and therefore will require a
highertransferto get them to the poverty
line. Thus,the key trade-offis betweenthe
cost savingsdue to lower transfersto the
nonpoor and the cost increases resulting
from higher transfersto the poor. To de-
scribe how this trade-off is optimally re-
solved,we need the conceptof a separating
workrequirement,denotedby C4.Thisis the
work requirementthat, if coupled with a
benefit sufficient to get the poor to the
povertyline, makes high-abilityindividuals
indifferentbetween claimingto be of low
abilityand receivingno benefit at all. For-
mally,it is definedby
v(O,O,aH) = V(Z - Y(CL, aL), CL,aH).
The readermayeasilyverifythat C4exists,
is unique,and is greaterthanzero.
PROPOSITION 2: If both income-gener-
atingabilitiesand incomesare unobservable,
one of thefollowingtwoPAP'sis cost-mini-
mizing:(i) (welfare)imposeno workrequire-
mentsandofferbothabilitygroupsa transfer
of z - y(O,aL); (ii) (workfare)offerself-cate-
gorizedhigh-abilityindividualsnobenefitsand
offerself-categorizedlow-abilityindividualsa
transferof z - y(c, aL) in exchangefor a
work requirementof C4. A sufficientcon-
ditionfor the workfaresolutionto be cost-
minimizingis that aL < (1- y)aH.
Thus,we can restrictattentionto one of
two possiblesolutions.14The firstpools the
two abilitygroupsandinvolvesno workfare.
All individualsaregivena transfersufficient
to get the poor to the povertyline, namely
z - y(O,aL).The second separatesthe two
groups.Those claimingto be of low ability
are offered a transferz - y(cL,aL) in ex-
change for a work requirement of C4.
High-abilityindividualsareofferedno bene-
fitsbut haveno incentiveto masqueradeby
definitionof the separatingwork require-
ment. The choice between these two solu-
tions trades off the savingsfrom givingno
transfersto the nonpooragainstthe cost of
reducingthe poor'sprivate-sectorearnings.
This boils down to a comparisonbetween
relativewage rates and the fractionof the
populationwho areof low ability.Workfare
is more likelyto be the optimalsolutionif
the trulypoor representa smallfractionof
the targetpopulationand if their earnings
potential is small relative to that of the
nonpoor. Both of these conditions mean
that the loss in the poor's private-sector
earningscausedby implementingthe work-
fareprogramwillbe small.
The twoPAP'sdescribedin Proposition2
are depictedin Figure2. The transferz -
y(O,a) is thatwhichis just sufficientto get
the poor to the povertyline in the absence
of work requirements.For the case illus-
trated,we find c4 fromthe intersectionof
the low-abilitybudget line startingat z -
y(O,a) andthe statusquo high-abilitybud-
get line."5It is clearthat thisworkrequire-
ment, when coupled with a transfer z -
14Formalproofsof this andthe remainingproposi-
tionscanbe foundin ourdiscussionpaper(Besleyand
Coate,1991a).
15Figure2 depictsa caseinwhichc' is to the leftof
l(aL). This need not be true. However,the reader
shouldfindit straightforwardto illustratetheworkfare
solution and verify the sufficientcondition for the
othercase usinga similarapproachto thatused here.

8. VOL..82NO.1 BESLEYANDCOATE:WORKFAREVS.WELFARE 255
v(O,O,aH) aaH
y
,-v(z-y(O,aL ),O,aL)
(O,aL)+a
z-y(cs,a L)
z-y(O,aL)
FIGURE2. ILLUSTRATIONOFTHESOLUTION
DESCRIBEDIN PROPOSITION2
y(cs,a), makes high-ability individuals in-
differentbetweenmasqueradingas low-abil-
ity types and receiving no benefits. If a
high-abilityindividualwere to masquerade,
his budget line would emanate from (z -
y(c L), csL), and he would end up on the
sameindifferencecurveas in the absenceof
anytransfer.The figurecan alsobe used to
verify that the cost comparisonhinges on
the conditionstated in Proposition2. Sim-
ple geometry reveals that z - y(cs, a)=
aHcL and that z - y(O,aH) = (aH- aL)cL.
Thus,the workfaresolutionwillbe cheaper
if Hyacs <(aH- aL)csL or, equivalently,if
aL < (1- y)aH.
B. ObservablePrivate-SectorEarnings
Suppose now that the policymakercan
observeindividuals'earnings.In contrastto
the case thatwe havejust discussed,it may
now be possible to implementthe bench-
mark PAP. This is because a high-ability
individualwho claimsto be poormusthave
earnings of y(CL, a), which requires labor
supply of Y(CL, aL)/aH. This suboptimal
labor-supplychoice reduces the value of
masquerading.The benchmarkPAP is im-
plementableif and only if v(O,0,aH) 2 z -
h(y(O, aL)/aH) [i.e., a high-abilityindivid-
ual prefersclaimingno benefitto reducing
his labor supplyto y(O,aL)/aH and con-
sumingz].
The problemof implementabilitywill re-
main, however,when this conditionis not
satisfied.The relevantincentive-compatibil-
ityconstraintnowtakesthefollowingform:16
v(bH,CH,aH) 2 bL+ y(cL,aL)
-h(y(cL, aL)/aH + CL).
The solution to the government'sproblem
hasthe samebasicstructureas in the previ-
ous case. We can againdefine a separating
workrequirement(denoted in this case by
AS
), which is now given by the equation
v(O,O,aH) = z - h(
A
+ Y(CL,aL)/aH) It
exists, is unique, and is positive provided
that v(0,0, aH) < z - h(y(O,aL)/aH). Since
the gains from masqueradingare smaller
when earningsare observable,CL will be
smallerthan the separatingwork require-
ment of the previouscase. Then, we state
the followingproposition.
PROPOSITION 3: If income-generating
abilities are unobservable,individuals'in-
comes are observable,and the benchmark
PAPis not implementable,oneof thefollow-
ing two programs is cost-minimizing:(i)
(welfare)imposeno workrequirementsand
offerself-categorizedhigh-abilityindividualsa
transferof z - h(y(O,aL)/aH)- v(O,O,aH)
andofferself-categorizedlow-abilityindividu-
als a transferof z - y(O,a); (ii) (workfare)
offerself-categorizedhigh-abilityindividuals
no benefitsandofferself-categorizedlow-abil-
ity individualsa transferof Z- y(CL, aL) in
exchangefor a workrequirementof A^SL.A
sufficientconditionfor the workfaresolu-
tion to be cost-minimizingis that RyaL<
(1- y)h'(y(O,aL)/aHXl - aL/aH).
Notice that, in contrastto the previous
case, the welfare solutiondoes not involve
payingout the same benefitto both ability
groups.Since the governmentcan observe
individuals' incomes, it can offer what
amountsto a benefit schedule.Those with
incomesy(0,a) are offereda transferz -
y(O,a), while those with incomes y(O,aH)
16Therewillalsobe ananalogousincentivecompati-
bilityconstraintfor low-abilityindividuals,but this is
effectivelyredundant.

9. 256 THEAMERICANECONOMICREVIEW MARCH1992
z-h(y(O,a L)/aH)
v(0,0,aH)
y aH1
z-y(O,aL)+aLi
C BI
z
Z-yQtspaL) ' -I
z-y(O,aL)aL A
CL (aL) t(aH) CL'i
FIGURE 3. ILLUSTRATION OF THE SOLUTION
DESCRIBED IN PROPOSITION 3
receive a (smaller)transfer,equal to z -
h(y(O,aL)/ aH)-v(O, O,aH).
Recallthatthe term h'(y(O,aL)/aH) rep-
resents the marginaldisutilityof labor at
the level suppliedby a nonpoorindividual
who is claimingto be poor. Since this is
lowerthantheywouldliketo supplyat their
wage rate, h'(y(O,aL)/aH) is less than aH.
The sufficientconditionfor workfarewhen
earningsare observableis thus more strin-
gent than the requirement that aL <
(1-Sy)aH in Proposition2. Indeed, it can
be shownthatwheneverworkfareis optimal
in the case of earningsbeing observable,it
is also optimal when they are unobser-
vable.17Thus,the demandforworkrequire-
mentswill be less acutewhen it is possible
to combatthe incentiveto masqueradeby
monitoringindividuals'incomes.
The twoPAP'sdescribedin Proposition3
are illustratedin Figure 3. We can deter-
mine A' byreferringto the dashedline AB.
If a "masquerading"high-abilityindividual
were to take up the package (z -
Y(A, aL),C he wouldhaveto end upwith
total income equal to z in order to be
consistentwithhis claimingto be of typeL.
Thus, he must consume at point B in the
figure.However,at B, a high-abilityindivid-
ual is on the same indifferencecurveas he
would be in the status quo. Thus, he is
indifferentbetween masqueradingand re-
ceivingno benefits.A similarprocedureal-
lows us to illustratethe transfergiven to
high-abilityindividualsin the welfare solu-
tion. If a high-abilityindividualtakes the
low-abilitycash transfer z - y(O,a) he must
consume at point C on indifferencecurve
z - h(y(O,aL)/aH). The transfer given to
high-abilityindividualsin Proposition3 is
just sufficient to deter them from mas-
querading[i.e., is equal to the verticaldis-
tance between the indifferencecurves z -
h(y(O,aL)/aH) and v(O,O,aH) at l(aH)].
Thiscompletesourmaindiscussionof the
screeningargumentfor workrequirements,
althoughwe shalldiscussthe robustnessof
the resultsto changesin the assumptionsin
SectionV.
IV. TheDeterrentArgument
In this section,we revertto the case of
observableabilities.However,to capturethe
idea that povertydependsnot only on luck
but also on choicesmadeearlierin life, we
allow each individual to make an ex ante
choice which influenceshis future earning
ability. Specifically,we suppose that the
probabilitythatan individualis of highabil-
ity is given by Tr(e), where e denotes effort.
The function7r(-)is assumedto be increas-
ing and strictlyconcave,so that higherlev-
els of effortincreasethe probabilityof being
high-ability,but at a diminishingrate.Effort
is measuredinunitsof (dis)utility.18
Individuals are assumed to know the
structureof the government'spoor-support
program when they make their effort
choices.Thus,giventhe existenceof a PAP
of the kind discussed above,"9individuals
17Theinterestedreaderis referredto ourdiscussion
paper(BesleyandCoate,1991a).
18To make this setup more concrete,the reader
mightimaginethat an individual'searningabilityde-
pends on graduatingfrom high school and that the
likelihoodof this dependsnot only on "luck"in the
formof geneticendowmentandqualityof teachers,but
alsoon howhardan individualworks.Alternatively,in
the AFDC context,one mightimaginethat a young
woman'sincome-generatingabilitydependsonwhether
or not she has a child. The likelihoodof this event
dependsnotonlyon luck,butalsoon preventiveeffort.
9Weareimplicitlyassumingthatindividuals'efforts
are unobservableto the policymaker.If thiswere not

10. VOL.82 NO.1 BESLEYANDCOATE:WORKFAREVS.WELFARE 257
willchoosetheireffortlevel,e, to maximize
(5) 7r(e)v(bH,CH, aH)
+ [1-7w(e)]v(bL,CL,aL) - e.
Givenourassumptions,thereexistsa unique
effortlevel, e*, which solves this problem.
Thiswillbe an increasingfunctionof the ex
post differencebetween the utilitylevels of
high-andlow-abilityindividuals.Thus,e* =
e*(,Y()), where
(6) X(bL,CL,bH,CH)
- v(bH,cH,aH) -v(bL9CL9aL).
The function X( ) representsthe ex post
utility difference between being a high-
ability individual and being a low-ability
individual.
The (expected) cost of a PAP is now
givenby n{[1--n.(e*(X))IbL + ,(e*(X))bH),
and the policyproblemis to minimizethis,
subjectto the constraintsof voluntarypar-
ticipationandpovertyalleviation.The main
differencebetweenthis and the case exam-
ined in Section II is in the endogeneityof
the expected numberof poor. The bench-
mark PAP is now problematicbecause it
reduces the ex post utility differencebe-
tween high- and low-ability types and
therebyreducesthe returnsto effort.Indi-
viduals will thus choose a lower level of
effort,andthe numberof poorwillincrease.
Thus,poorreliefmay,in the wordsof Mill,
create an "undue reliance"on state sup-
port.
Before we can state the solution to the
policyproblemin this instance,we need to
define the maximal work requirement, de-
noted cm. This is the work requirement
which, if coupledwith a transfersufficient
to get the poor to the povertyline, would
makethemjust indifferentbetweenthe sta-
tus quo andparticipatingin the program.It
is definedbythe equation
V(Z - y(c, aL),cL,aL) = v(O,O,aL).
Again, it can be verifiedthat cL existsand
is unique. Note also that cLmmust exceed
l(aL), since in the status quo, low-ability
individualsearn less than z. Thus, at the
maximalworkrequirement,low-abilityindi-
viduals do no work in the private sector,
andhence, y(c', aL) =0. We nowstate our
finalproposition.
PROPOSITION 4: If income-generating
abilitiesare observablebut dependpartlyon
choicesmadeearlierin life,thecost-minimiz-
ingPAPeitherimposesno workrequirements
andofferslow-abilityindividualsa transferof
z - y(O,aL), or imposes the maximal work
requirementCL'on low-abilityindividualsand
offersthema transferof z.
The logic behind this proposition is
straightforward.The benchmarkPAP de-
creases the ex post utility difference be-
tween high-abilityand low-abilityindividu-
als from v(0, 0, aH) - v(0, 0, aL) to
v(0,0,aH)- v(z - y(O,aL),0, a). This re-
duces the returnsto effort, increases the
number of poor, and raises the cost of
poverty alleviation.This problem can be
mitigatedby reducingthe poor'sgain from
the PAP, and introducinga work require-
ment maybe able to do this.20However,if
the work requirementis chosen to lie be-
tween 0 and l(aL), for example at c'L in
Figure4, it will haveno such effect.This is
because,to meet the poverty-alleviationob-
jective,it mustbe coupledwitha transferof
b'L.But this policywill not alterthe expost
utility level of poor individuals.They will
continueto receivean incomeof z and do
i(aL) units of work. The only difference is
that c'L units of theirwork is now done in
the public sector. Thus, imposinga small
workrequirementwillnot reducethe poor's
the case, then the governmentshouldoffereffort-con-
tingenttransfers.
20t mayalsobe mitigatedby increasingthe payoff
to becominghigh-ability.This may be achieved by
offeringtransfersto high-abilityindividuals.Note that
Proposition4 tellsus nothingdirectlyaboutthe roleof
this instrument.All we havebeen able to establishis
that,if thereis a benefitofferedto high-abilityindivid-
uals, then it will be higherin the welfarethan in the
workfaresolution.

11. 258 THEAMERICANECONOMICREVIEW MARCH1992
y v(z-y(O.aL),O,aL)
V(O,O,aL)
aLl
I~~~~~~
z-y(O,aL) I I
CL t(aL) CLCLL
FIGURE 4. ILLUSTRATION OF THE MAXIMAL
WORK REQUIREMENT, CfL
gainfromthe PAP.At the sametime,how-
ever, the government must increase the
transfer given to low-ability individuals,
since theirprivate-sectorearningswill have
fallen. Hence, the costs of povertyallevia-
tionareincreasingin CL in thisrange.
With work requirements in excess of
(aL), low-abilityindividualsdo no workat
all in the privatesectorandreceivea trans-
fer of z. Further increases in the work
requirementthus necessitateno furtherin-
creasesin the transfer.Moreover,since in-
dividuals no longer respond by reducing
theirprivate-sectorlaborsupply,raisingthe
workrequirementdoes reducetheir utility.
Thus, the poor'sgain fromthe PAP is de-
creasingin CL for valuesof the latterabove
(aL). The expected number of poor is
therefore falling and so are the costs of
povertyalleviation.Thereis,however,alimit
to the size of the workrequirementthatcan
be imposed, given by the maximalwork
requirementcl, whichis also illustratedin
Figure4. Clearly,if the work requirement
were set abovethislevel,the poorwouldbe
better off not participatingin the program.
Since costs are increasingin CL below
(aL) and decreasing thereafter, we obtain
the result reported in Proposition4. This
seems consonantwith the logic of the En-
glish Poor Law.By all accounts,those who
ended up in the workhouseworkedmuch
harderthan in a laissez faire equilibrium.
On the other hand,manyhave arguedthat
CL actuallyexceeded cm for many of the
poorandthe objectiveof povertyalleviation
was not met.
The choice between the two solutions
tradesoff the incentivebenefitsagainstthe
loss in the poor's private-sectorearnings.
The workfare solution will be preferable
when, in the absence of intervention,the
fraction of low-ability individualsin the
populationis smallandwhenthe low-ability
individuals'wage is low. Both of these con-
ditionsimplythat the loss in private-sector
earningsfromimplementingworkfarewould
be smaller.
V. Extensions
In the previous two sections, we have
shownthatworkrequirementscanplayboth
a screeninganda deterrentrole in poverty-
alleviationprograms.Our argumentswere,
however,developedunderveryspecificas-
sumptions.The purposeof this sectionis to
assesshowrobusttheyareto differentenvi-
ronments.
Our analysistook a "nonwelfarist"defi-
nitionof poverty.To be poorin the senseof
our model is to have insufficient income
rather than too little utility. This is conso-
nant, we believe, with the way in which
most governmentsview poverty.Whatever
its theoreticalmerits,they do not, for the
most part,attemptto measurethe value of
leisure when deciding who is poor.
Nonetheless, it is interesting to consider
howourargumentswouldbe changedif the
policyobjectivewereto providea minimum
level of utility.
Althoughthe detailsof the argumentare
changed slightly, the screening argument
goes throughlargelyunscathed.Whilewith
the income-basedviewof povertythe trade-
offis betweenthe screeninggainsfromwork
requirementsandahighergapbetweenpoor
individuals'private-sectorearningsand the
povertyline, in the welfaristcase it is be-
tween the screeninggainsand a highergap
betweenpoor individuals'actualutilityand
the minimumlevel.The deterrentargument
does not, however,makesenseundera wel-
farist objective. If the governmentcared
aboututility,thenworkrequirementswould
be self-defeating;they would necessitate a
compensatingincrease in income to pre-
serveindividuals'utilitylevels.

12. VOL.82 NO.1 BESLEYANDCOATE:WORKFAREVS.WELFARE 259
Ourmodelalsoassumedthatthe causeof
povertylayin a low income-generatingabil-
ity.Thismustbe contrastedwiththe case in
which poor individualssimplyhave a high
disutilityof labor.If thisweretrue,thenthe
screeningargumentwould be in jeopardy,
since,controllingfor earningability,a work
requirementwould discouragethose with a
high leisure preference most of all from
participatingin a PAP.A workrequirement
would therefore target benefits to exactly
the wronggroup!While it seems unreason-
ableto assumethatpovertyis primarilydue
to high leisure preferences,this point does
underscore a genuine difficultywith the
screeningargumentfor workrequirements.
Imposing mandatory work requirements
might cause those poor with high prefer-
ences for leisureto opt out of the program
at the cost of povertybeingalleviated.Even
so, somewoulddoubtlessarguethat society
has no obligationto help those able-bodied
individualswho are unwillingto work for
benefits.
Our model also assumed that the poor
were capableof work.Some fractionof the
poor may,however,be unemployables(i.e.,
thosewhomeithercannotworkdueto phys-
icalor socialdisability,or those whomsoci-
etydeemsshouldnot work,suchas mothers
withveryyoungchildren).If such individu-
alscanbe identifiedat reasonablecost,they
can be tagged in the way suggested by
Akerlof(1978)(i.e., offeredcategoricalben-
efits).However,absentthispossibility,work
requirementsmaybe flawedas a meansof
achievingmore accuratetargetingof bene-
fits. By imposingthem, one would risk de-
terringpart of the needy populationfrom
claimingpoorsupport.
How importantthis is as a practicalmat-
ter is moot. There clearlyare categoriesof
unemployable individualswho are easily
identified.These include motherswith in-
fants and individualswith obviousphysical
or mentaldisabilities.Nonetheless,there is
a significantgray area, encompassingindi-
vidualswith bad backs and certain mental
ailments. Following Nichols and Zeck-
hauser (1982) and Charles Blackorbyand
DavidDonaldson(1988),however,therere-
mains the possibilityof offering such in-
dividualsother in-kind transfers(such as
medical treatments)valued only by those
individualswith the relevantdisability.Em-
ployablepoor individualscould still be re-
quiredto workandwouldhaveno incentive
to pretendthat they had the unobservable
disability, since they would receive what
amountsto a transferof lowervalue.
Finally,our model assumedthat the gov-
ernment could offer only lump-sumcash
transfersin exchangeforworkrequirements
in the publicsector.Thisneglectsthe use of
earningssubsidies,which have had numer-
ous proponentsin the U.S. context.21The
availability of earnings subsidies would
weakenboth of the argumentsfor workfare
discussedabovebecausetheycouldbe used
to design a programthat involved lower
transfersto the poor,was less attractiveto
the nonpoor,andofferedthe poorless util-
ity. If, instead of giving the poor a cash
transfer of z - y(O,aL), the government
were to offerthem an earningssubsidyjust
sufficientto get them to the povertyline,
this would inducea higherlevel of private-
sector earnings and hence necessitate a
smaller total transfer.Moreover,it would
also make masqueradingless attractiveto
the nonpoor. A masqueraderwould still
have total income of z but would have to
workharderto get it. Finally,offeringsup-
portto the poor in the formof an earnings
subsidywouldalso reducethe poor'sutility
comparedwith a cash transfer.This would
reducethe relativeattractivenessof poverty,
therebylesseningthe adverseincentiveim-
pacton effortchoice.22
Our PAP also ruled out the use of pri-
vate-sectorworkrequirements.Ratherthan
demandingunproductiveworkin the public
sector, it might seem more sensible to re-
quirethatindividualsworka certainamount
in the privatesector in exchangefor bene-
fits. Unfortunately,however,there will be
21The meritsof thispolicyhavebeen stressed,inter
alia, by Ellwood(1988),Kesselman(1969),and Zeck-
hauser(1971).
22Fora morecompletetreatmentof this issue,see
BesleyandCoate(1991b).

13. 260 THEAMERICAN ECONOMIC REVIEW MARCH1992
substantialpracticaldifficultiesinvolvedin
implementingsuchworkrequirementsif, as
is likely,thereis uncertaintyaboutindividu-
als' private-sectoremploymentopportuni-
ties. Individualsmaythen be unableto ful-
fill their private-sectorwork requirements
throughno fault of their own, and policy-
makerswould be faced with the awkward
task of distinguishingthose who have tried
to fulfilltheirworkrequirementfromthose
who haveshirked.
VI. Conclusion
Thispaperhasexploredthe incentivecase
forworkfare.We haveanalyzedtwodistinct
arguments:a screeningargumentthatwork
requirementsserveas a meansof targeting
transfersanda deterrentargumentthatthey
may serve as a device to encourage
poverty-reducinginvestments.These argu-
ments have been advancedto supportthe
use of work requirementsboth historically
and in recentpolicydiscussionsin develop-
ing and developed countries. The main
pointsof the analysismaybe summarizedas
follows.
The cost of usingworkfarein our model
is thatpublic-sectorwork"crowdsout"pri-
vate-sectorwork,increasingthe size of the
povertygapandthe costsof povertyallevia-
tion. The screeningargumentfor workfare
relieson the benefitsdue to reducedtrans-
fers to the nonpoor exceeding the costs
stemmingfromreducedprivate-sectorearn-
ings. This is more likely to be true when the
governmenthas limited abilityto monitor
individuals'earnings,since masqueradingis
likely to be more widespread. Thus, the
screeningcase for workfareseems likelyto
be strongest in the context of developing
countries.It should also be noted that the
screening argumentrests on the dual as-
sumptionsthat the causesof povertyreside
in inadequateearningopportunities(rather
thana highdisutilityof leisure)andthatthe
poor are,in principle,capableof work.
For there to be a deterrentargumentfor
workfare,the benefitsfroma reducednum-
ber of poor must exceed the costs due to
lower private-sectorearnings.This will not
be the case if the governmentimposesonly
a small work requirement.Workfarecan
onlybe an effectivedeterrentif the amount
of workdemandedis considerablyin excess
of that whichpoor individualswould do in
the absence of intervention. Finally, it
should be noted that the deterrent argu-
ment rests critically on the government
adopting a nonwelfarist definition of
poverty.If povertyis definedas havingtoo
little utility, rather than too little income,
then this argumentis not coherent in our
framework.
REFERENCES
Akerlof,GeorgeA., "The Economics of 'Tag-
ging' as Applied to the OptimalIncome
Tax, Welfare Programs,and Manpower
Planning," American Economic Review,
March1978,68, 8-19.
Besley,TimothyJ. and Coate,Stephen, (1991a)
"WorkfareversusWelfare:IncentiveAr-
guments for Work Requirements in
PovertyAlleviationPrograms,"John M.
Olin Programfor the Studyof Economic
OrganizationandPublicPolicyat Prince-
ton UniversityDiscussionPaper No. 73,
1991.
and ,(1991b) "The Design of
Income MaintenancePrograms,"unpub-
lished manuscript,Princeton University,
1991.
Blackorby, Charles and Donaldson, David,
"Cash versus Kind, Self-Selection, and
Efficient Transfers," American Economic
Review,September1988,78, 691-700.
Dreze,JeanP.,"FaminePreventionin India,"
in Jean P. Dreze and AmartyaK. Sen,
eds., Hunger: Economics and Policy, Ox-
ford: OxfordUniversityPress, 1990, pp.
13-122.
Dye, Ronaldand Antle,Richard,"Cost Mini-
mizing Transfer Programs,"Journal of
Public Economics, April 1986, 30, 259-63.
Ellwood,David, Poor Support: Poverty in the
American Family, New York: Basic Books,
1988.
and Summers,Lawrence,"Povertyin
America:Is Welfare the Answer or the
Problem?"in Sheldon Danziger and D.
Weinberg, eds., Fighting Poverty: What

14. VOL. 82 NO. 1 BESLEYAND COATE: WORKFAREVS. WELFARE 261
Works and What Doesn't, Cambridge,
MA: Harvard University Press, 1986, pp.
78-105.
Gueron, Judith M., "Work and Welfare:
Lessons on Employment Programs,"
Journal of Economic Perspectives, Winter
1990, 4, 79-98.
Himmelfarb,Gertrude,The Idea of Poverty,
New York: Knopf, 1984.
Jencks, Christopherand Edin, Kathryn,"The
Real Welfare Problem," American
Prospect, Spring 1990, 1, 31-50.
Katz,MichaelB., In the Shadow of the Poor-
house, New York: Basic Books, 1986.
Kaus,Mickey,"The Work Ethic State," The
New Republic, 7 July 1986, 194, 22-33.
Kesselman,Jonathan R., "Labor-Supply Ef-
fects of Income, Income-Work, and Wage
Subsidies," Journal of Human Resources
Summer 1969, 4, 275-92.
Laffont,Jean-Jacques,The Economics of In-
formation and Uncertainty, Cambridge,
MA: MIT Press, 1990.
Mill,JohnStuart,Principlesof Political Econ-
omy, Books IV and V, Harmondsworth,
U.K.: Penguin, 1848.
Murray,Charles, Losing Ground: American
Social Policy 1950-1980, New York: Ba-
sic Books, 1984.
Nichols,AlbertL. and Zeckhauser,RichardJ.,
"Targeting Transfers through Restric-
tions on Recipients," American Economic
Review, May 1982 (Papers and Proceed-
ings), 72, 372-7.
Ravallion, Martin, "Market Responses to
Anti-Hunger Policies: Effects on Wages,
Prices and Employment," in Jean P.
Dreze and Amartya K. Sen, eds., Hunger:
Economics and Policy, Oxford: Oxford
University Press, 1990, pp. 241-78.
, "Reaching the Rural Poor through
Public Employment: Arguments, Evi-
dence and Lessons from South Asia," The
WorldBank Research Observer,July 1991,
6, 153-75.
Zeckhauser, Richard J., "Optimal Mecha-
nisms for Income Transfer," American
Economic Review, June 1971, 61, 324-34.