Contours of employment protection reform

f

L!BRARI£6|

Digitized

by

the

Internet

Archive

in

2011

with

funding

from

Boston

Library

Consortium

Member

Libraries

1B31

.M415

3^

Massachusetts

Institute

of

Technology

Department

of

Economics

Working

Paper

Series

Contours

of

Employment

Protection

Reform

Olivier

Blanchard

Jean

Tirole

Working

Paper

03-35

November],

2003

Room

E52-251

50

Memorial

Drive

Cambridge,

MA

02142

This

paper can be

downloaded

without

charge

from

the

Social

Science

Research

Networl<

Paper

Collection

at

MASSACHUSETTS

INSTITUTE

OF

TECHNOLOGY

NOV

2 C

2UU3

Contours

of

employment

protection

reform

Olivier

Blanchard

and Jean

Tirole

*

November

1,

2003

Abstract

Startingwith a simplebenchmark, we firstderivethecharacteristics ofoptimal

employmentprotection. Inthebenchmark,employmentprotection takestheform

of layoff taxes,usedtofinanceunemploymentbenefits.

We

then consideranumber

ofextensions,andshowhowthisprinciplemust bemodifiedandrefined,but not

abandoned.

We

then turn to the employment protection system in place in Prance today,

andshowthat itdiffersfrom thisprinciple in twomaindimensions. First,

con-tributionsbyfirms totheunemploymentinsurancefundtaketheformof payroll

taxes rather than layoff taxes. Second, the layoff process is subject to heavy

administrativeandjudicialcontrol.

This leads ustomaketwo mainrecommendationsforreform: The introduction

ofa layofftax, witha corresponding decrease inthepayroll tax; and a reduced

roleofthejudicialsystem in thelayoffprocess.

* This paper is an english adaptation of a report written forthe French Conseild'Analyse

Economique [BlanchardandTirole 2003b].

We

thank DaronAcemoglu, David Autor, Martin

Baily, Olympia Bover, PierreCahuc, Daniel Cohen, Mathias Dewatripont, FVancis Kramarz,

FiorellaPadoa-Schioppa, GillesSaint-Paul, RobertSolow, NicolasVeron,Robert Wagner, and

especially Denis Fougere, Jacques Freyssinet for discussions and comments.

We

thank Jean

Employment

ProtectionReform

Introduction

There

may

be no labor market institution more controversial than

employment

protection regulation

—

the set oflaws and procedures regulating separations

be-tweenfirms andworkers.

• Firms complainnotonlyaboutthedirectcost, butalsoaboutthe

complex-ityandthe uncertainty introducedby suchregulation.

They

argue that it

makesit difficultfor

them

to adjust tochanges in technologyand product demand, andthatthisinturn decreasesefficiency,increases costand, inso doing, detersjob creation.

• Workers,ontheotherhand,focusonthepainofunemployment, andargue that such pain should be taken into account byfirms

when

they consider

closing a plant, or laying off a worker. That workers protected by em-ployment protection wouldfavoritisnogreat surprise. Butevidence from surveysshowsthatsupportforemploymentprotectionismore broadbased.

•

Many

economistsandinternationaleconomicorganizations,fromthe

OECD

to the

IMF,

tend to side with firms. Thereis, they argue, a trade-off

be-tween insurance and efficiency.

The

current system impedes reallocation,

and, by implication, reduces efficiency. It leads to higher costs, and thus loweremployment. Ata

minimum,

it couldandshould be

made

more

effi-cient.

More

likely,overall employment protectionshould be reduced.

• Faced with conflicting advice and demands, the governments ofWestern Europe have been prudent (or timid, depending on one's point of view.)

They

havelearned,oftenthehardway,thatworkers coveredbyemployment protectionarenot eagerto seeitreduced, andthattheseworkersrepresent

the majority of the labor force, and a large part of the electorate. So,

most ifnot all ofthe recent employmentprotection reforms have worked

at the margin, through the introduction and extension ofthe scope for

fixed duration contracts

—

contracts subject to

more

limited employment protectionandsimpleradministrativerules.For themostpart,employment protection forregularcontracts hasremained unchanged.

The

evidence so

far is that this dualsystem has led toan increasingly dual labor market, withmixed efficiencyand distributionaleffects.

Employment

ProtectionReform

Despite the heat and the rhetoric, we are struck at

how

little work has gone

into the question of

how

"good employment protection regulation" should look.

Starting from the status quo, firms and international organizations have argued

for less protection. Workers and unions have fought to keep the protection they

had. Governments have looked for politically feasible incremental reforms. But the ultimate goal, the shape of optimal employment protection, has been left

undefined.

Consider forexample the following questions:

• Shouldtherebeanystate

mandated

employmentprotection, orshould "em-ployment at will" remain the principle, leaving any additional protection

tovoluntary agreements betweenworkersand firms?

• Ifthereisanargumentforstate

mandated

employmentprotection,should

thisprotectionsimply take the formofa schedule ofpayments by firmsin

case oflayoffs,with thelayoffdecisionthenlefttothefirm?

Or

shouldthere be, inplaceorinadditiontosuch aschedule,othernon-pricerestrictions?

In thatcontext, whatshould betheroleofthe judicial process,ifany?

•

How

largeshould payments by firms either toworkersor to thestate be? Should firms pay workers directly, or should they pay the state? Should the paymentscover,inexpectedvalue orinrealization, theunemployment

benefitsand otherpaymentsreceivedbylaid-offworkers?Shouldthe pay-ments be

made

byfirms at the timethe layoffstake place, orshould they be paid overtime, asis the case inexperience rated systems?

Thisreport'sprimary purposeistoanswerthese questions,andtousethe answers

todrawthe contoursof institutional reformin the Frenchcontext.

We

organizeour discussionbystarting fromasimple benchmark. Inthat bench-mark, firms and the state are risk neutral. Workers are risk averse and cannot

fully self-insure against unemployment. In that context, characterizing optimal

employmentprotectionisstraightforward. Firmsmustbe

made

to internalizethe cost ofunemployment.So,ifforexample,unemploymentbenefitsareadministered by an unemployment agency, firms must pay contributions to the agency equal

tothe presentvalue of

unemployment

benefitspaid by the agency to the worker

they layoff. Put another way, the contribution rate of firms, defined as the ratio

Employment

Protection Reform

equal toone. In that sense,

unemployment

insurance and employmentprotection

are both integral parts oftheoptimal setoflabormarketinstitutions.

This benchmark, like all benchmarks, is both useful and too simple.

The

labor marketsuffers from

many

imperfections,and mostof

them

impingeonthe nature

ofoptimal labor market institutions. Theserange from theneed to give the un-employed incentives tosearch forajob, to thescope for ex-post renegotiation of

wages, to liquidity constraints faced byfirms.

Each

ofthese imperfectionsaffects

theoptimal contributionrate, andthus theoptimal degreeofemployment

protec-tion. But the general principle remains, that of

making

firms internalize, to the

extent possible,the socialcost ofunemployment.

Turning to actual institutions, it isclear that this principle isat odds with the

Frenchsystem ofemploymentprotectioninat leasttwo

main

dimensions:

First,intheFrenchsystem,contributionsbyfirmsto the

unemployment

insurance

fundtake theformof payroll taxes:

A

firmwitha higherlayoffratedoes notpay higher contributions. Severance payments, as set in the law, are low. In other words, the contribution rateis(close to) zero.

Second, thelayoffprocess issubject toheavyadministrativeandjudicial control.

Firms have to proveeitherfault bytheworkerinthecaseofanindividuallayoff,

oreconomic needinthe caseof collectivelayoffs. Judgescanandoftendodisagree

with thefirms'decisions,leadingtosubstantiallossoftimeandfinancialpenalties

for firms.

This diagnosis naturally leads toourtwomain recommendations forreform:

• First, to increase the contribution rateof firms (that is introduce a layoff

tax, and decrease the corresponding payroll tax) so firms internalize the cost ofunemployment.

• Second, tolimit the role ofthe judicial system.

To

the extent that firms are willing to incur the financial costs associated with laying offworkers (and

we

arearguingthat these costsshouldbe higher atthe margin than they are today), judges should not be allowed to second guess thefirms' decisions.

Employment

ProtectionReform

Section1characterizesoptimalemploymentprotectioninthebenchmark economy.

The

next three sections explorethree

main

deviations from the benchmark, and

theirimplicationsforemploymentprotection.Section2introduceslimitson

unem-ployment insurance,coming fromsearchor shirking incentive constraints. Section 3explorestheimplications of liquidity constraintsonfirms.Section4exploresthe

implications of alternative forms ofwage setting. Building on these extensions,

Section 5discusses quits versus layoffs,and the roleofthe judicial process.

Section6 attemptstoputalltheseelementstogether,and drawsconclusionsabout the contours ofoptimal employment protection (the impatient reader can

jump

directlytothat section...)

Section 7 then returns to the employment protection system in place in France

today.

Our

purpose here is not to give an exhaustivedescription ofthe system, but rather to examine it in the light ofour earlier analysis of

how

an optimal system mightlook, andtopoint out the majordifferences.

Having doneso,

we

sketchinSection8 thecontoursof

what

employmentprotection reformmightlook likeinthe case ofFrance.

Employment

ProtectionReform

1

Designing

employment

protection.

A

benchmark

Inthinkingabouttheissues,itisusefulto startfrom asimplebenchmark,which showsmostclearlytherelationbetween

unemployment

insuranceand

employment

protection.

1.1

The

benchmark

economy

Think

ofthe followingeconomy.^

• Firmshire workers.

• Aftera worker hasbeenhired byafirm, thefirm learnsabout the

produc-tivity ofthe worker (this productivity

may

depend on the quality ofthe

match

between the worker and thefirm, oron the

demand

for the firm's

product,and so on).

•

The

firm

may

thenwanttokeep theworkerandproduce,or laytheworker

off. Ifthe workerislaidoff, hebecomes unemployed.

• Absent anyadditionalincome, theutilityofthe worker

when

unemployed,

islow:Put another way, and the terminologywill beuseful below, absent

additionalincome, the "wageequivalent ofbeing unemployed" is low. • Workersare riskaverse. Firmsareriskneutral.

• There are no information problems, so everything is observable and

con-tractable.

Under

these conditions, firmswill offerthe followingcontract toworkers:

•

They

will fullyinsure workers.

They

will doso bypaying a constant

wage

totheworkers theykeep, andaseverance paymenttotheworkers theylay

off.

The

severance payment will be such that the severance pay is equal to

the wage, minusthe

wage

equivalent ofbeing unemployed. Workers will

thereforehave the

same

levelofutility, whether theyare employed or un-employed.

•

They

will layworkersoff

when

productivityislowerthanthe wage

equiv-alentofbeingunemployed.

1.

A

formalmodel underlying the arguments(ortobe honest, most ofthearguments, aswe

havesometimes taken educatedguessesbeyondwhatourmodelcouldanswer...) in thisandthe

Employment

ProtectionReform

Thisisclearlythesociallyefficientrulefor layoffs:

From

anefficiencypoint

of view,workers shouldbekeptononlyiftheirproductivityishigherthan

what

thewageequivalent ofbeingunemployed.

And

firmsdonotneedto

commit

todoingso,because theyfullyinternalize

the cost ofalayoff forworkers. Giventhewage and theseverancepay,

ex-postprofitmaximizationleads

them

to laya workeroffonlyifproductivity

is lessthan the netlabor cost, that is less than thewage minus severance

pay, equivalently ifproductivity is less thanthe wage equivalent ofbeing unemployed. Butthisisexactlythe

same

astheefficiencycondition above. Inshort: Severancepaywillbe usedtofullyinsure workers.

And

itspresence

will lead firmstotakeefficient separation decisions.

•

Knowing

that theywillreceiveseverancepayments iftheyare laid off will

leadworkerstoaccept a lowerwagein exchange.

And

becauseworkersare

risk averse, the provision ofinsurance by firms will decrease their overall

expected laborcost.Thus, firmswillbe eager toofferseverancepayments:

thisincreases theirexpected profit.

In that economy, there will be substantial employment protection. It will take

the form of severance payments by firms to laid offworkers, sufficient to insure

them

against the lossofutilityif unemployed. But, in that

economy

also, there

willbe noneedforthegovernmentto

mandate

employmentprotection:Firmswill

provideit willingly, andin the rightamount.

1.2

Introducing

an

unemployment

agency

To

fully insure workers, firms must beable to assess the utihtyloss from

unem-ployment. Thisis noteasyfor

them

to do:

•

As

ofthetimeofthelayoff,this lossis a

random

variable:

The

outcomeof

searchis uncertain,and the workerdoes not

know

how

long heis goingto

beunemployed. Ifthefirmwereto

make

aone-timeseverance

payment

to

offsetthat loss,thisone-timepayment would doapoorjobofinsuringthe

laid offworker.

• Ifthefirm decides instead topaythelaid offworker overtime,contingenton

Employment

Protection

Reform

to actuallytrackthe worker, and determinewhether heisstillunemployed

orhasfound anotherjob; thedifficulty inmonitoringhis searcheffortand

making

sure that heisindeed lookingfor anotherjob.

• Rather obviously, individual firms cannot monitor laid off workers well

enough toprovide

them

with adequate insurance.

The

role ofmonitoring

unemployment

status and searchintensitymust be therefore delegatedto

an agency, privateor public.

The

state, given its existing administrative structure, is likely to be in

the best position to do the monitoring, and to administer the payment

of

unemployment

benefits, either alone or inconjunction with the private sector.^

So, returntoourbenchmark,but

now

supposethatan agencyisputin chargeof

monitoring and distributing

unemployment

benefits totheunemployed. Suppose

further that the agency can perfectly monitor and thus provide

unemployment

insuranceatnocost interms ofsearchintensity oftheunemployed.

How

will this

changetheoutcomerelative to thebenchmark?

The

answeris: Not much.

• Firms,

when

theylayoffa worker,will

make unemployment

contributions to

the

unemployment

agency

—

payments equaltothe expected present value

ofthe

unemployment

benefitstobepaidtothelaid offworker, orpayments overtime correspondingtothe

unemployment

benefitsactuallypaidtothe

laid offworker.

•

The unemployment

agency in turn will monitor and give

unemployment

benefits tothe laid offworkers for aslong as theyare unemployed.

• There will be a sharper institutional distinction between

unemployment

contributions (paid by firms to the agency) and

unemployment

benefits (paidbythe agencyto workers). But, inthisbenchmark,thetwowillstill

be equal.

The

contribution rate

—

defined as the ratio ofthe contribution paidbythefirmtothevalue of the unemploymentbenefits receivedbythe

laid offworker

—

will be equalto one. (As

we

shall see, thiswill no longer

bethe case

when we

introduce other labormarket distortions).

2. Thisis indeed theFrench solution, with thecombinationofthe state run

ANPE

andthe

Employment

Protection Reform

•

The

allocation will be the

same

as before. Workers will be fully insured.

Firms, because ofthe

unemployment

contributions they haveto

make

to

the agency in case of layoff, will fully internalize the social cost of un-employment and choose an efficient levelof separations. There will be no

trade-offbetweeninsurance andefficiency.

1.3

Unemployment

contributions or severetnce

payments?

We

havesofarinterpreted unemployment contributions asa formofemployment

protection.

A

system in which all payments are

made

from firms to the state ratherthan toworkers indeed providesemploymentprotectionto theworkers: It

makesit

more

expensive forfirms to layworkersoff,andthus reduceslayoffs. But

it

may

not lookand feel like employment protection to the workers,

who

do not

seethe

unemployment

contributionspaidbyfirmstothestate,and donotreceive

payments directlyfromthe firm.

It is therefore worth askingwhether

some

paymentscould orshould be

made

by

firms toworkersdirectlyat thetimeoflayoff.

To

doso, itisuseful todistinguish

betweenthecosts of"becoming unemployed" andthecostsof"beingunemployed."

•

The

cost ofbecoming unemployedis the cost associatedwithlosingajob,

not with

unemployment

perse. It is a psychic cost, and while it is often

ignored by economists, it plays animportant role in public discussions of

employment protection'', and its relevance has been well documented by

social psychologists*.

The

loss of along held job can and often does lead to a loss ofa network ofworkplace friends, healthdeterioration, a loss of

selfesteem.

•

The

cost ofbeing unemployed is the financial and psychic cost from

re-maining unemployed until one has found anotherjob.

For our purposes, the main difference between these two costs isthat the firstis

incurredatthetimeofthe separation,andthuscanbeoffset(intermsofutility)by

3. Twoquotes from judges at the Prud'hommes, the French labor courts (translated from

French) : "Employers havea hardtimeunderstanding thatthe main issueis not thefinancial

loss. Psychologically, alayoffisvery tough. For thefamily. Foryourhealth. Itputsyour whole

life intoquestion." and "People put their livesin theirjobs. And, atonce, everythingistaken

awayfrom them". (Liberation December2,2002).

Employment

Protection

Reform

10

aone-time

payment

fromthefirmtothe worker.

The

second, instead, is

random

at the time ofthe layoff. This suggests a natural division of tasks: Severance

payments fromfirms toworkers,atthetimeofthelayoff,tocompensate

them

for

thecost of becoming unemployed.

And

unemployment

benefits from the agency

to workers, paid over time, and financed by payments from firms to the agency,

to compensateworkers for the costofbeingunemployed.

In that light,

what

should the schedule of severance payments look like?

The

psychiclossappearstobeprimarilyafunctionoftimeinthejob,ofseniority.It is

likelytobe lowforworkerswith lowseniority,andto

become

highonly with high

seniority.^Thissuggestsanincreasingand convexscheduleofseverance payments

as a functionofseniority.®

Havingestablishedaframework,

we

considerinthenextthree sections threemajor deviations fromthe benchmark, and discuss, in eachcase,

how

theymodify our

conclusions.

2

Limited

unemployment

insurance

Inour benchmark, the

unemployment

agencyfully insured laid offworkers. But,

to the extent that the agency cannot fully monitor the search behavior of the unemployed, it can only offer limited insurance: OflPering anything close to full

insurancewould lead the unemployed tostop searchingand remain unemployed. This issue has been studied at length in the theoretical and empirical literature

on

unemployment

insurance.^

And

a

number

ofrecent reforms ofthe

unemploy-ment

system in Europe, such as the

PARE

in France, have indeed had as their goal to combine

more

generous and longer lasting unemployment benefits with

5.

A

factconsistentwiththis hypothesis (but alsowith anumberof others): InFrance, there

are fourtimesasmanyquits as layoffs forworkerswith2 to9years ofseniority,butfour times

as many layoffsthan quits for workerswith 10 or more years ofseniority [Goux and Maurin

2000,TableAl],

6. For usual incentive constraint reasons, the schedulecannothowever betoosteep;otherwiseit

wouldgiveincentivesforfirms tolayoffworkersatmid-career,i.e.whenthe severancepayments

associatedwithlayingthemoffarestillrelativelylow.

Employment

Protection Reform 11

strongerincentivesforthe unemployedtoaccept jobsifofferedbythe

unemploy-ment

agency. These reforms clearly go in the right direction.

They

potentially

offerbetter tailored insurance: Iftrulynojobs are available,thentheunemployed

continueto receive

unemployment

benefits.

And

they remove, at least inprinciple,

some

oftheproblems associatedwith theopen ended unemployment benefitsof the past. But realistically, even the best designedsystems cannot fully eliminate monitoring problems, andso, lessthanfull insurance isoptimal: There hastobe

some

utility cost to

unemployment

tomotivatesearch.

In such a context, the optimal employment protection/unemployment benefits

system is

more

complex to characterize.

The

general architecture remains the same, but the details aredifferent:

•

The unemployment

agency pays

unemployment

benefits to workers,

pro-viding as

much

partialinsurance asisconsistentwith searchincentives.

•

The

lower the feeisiblelevel ofinsurance, the higher theutilitycosts that

layoffs impose on laid-off workers. So, to lead firms to take these costs

into account,

unemployment

contributions by firms to the agency must

now

exceed the

unemployment

benefitspaidbytheagencyto workers;

The

optimal contribution rate is

now

greater than one.

And

the layoffrate is

smallerthanin the benchmark.

• So, the more stringent the constraints on the

amount

of insurance the agencycanprovide, thehigher the contributionraterelativetothe bench-mark,the lower the layoff rate, and, in this sense,the higher the optimal degree ofemployment protection.

Three remarksbeforemovingon.

•

Under

this deviationfrom the benchmark,

unemployment

insurance and

employment

protection are substitutes, notcomplements.

The

poorer the

insurance,the higher the optimal degree ofemployment protection. While

the result is normative, this negative relation appears to be present in

the data across Continental European countries.®

The

countrieswith the

highestdegreeofemploymentprotection (usingthe

OECD

index) are also

Employment

Protection

Reform

12

thecountries where

unemployment

insurance coverage has been relatively limited.

Here, thepolitical

economy

explanation

may

actually followthe

same

logic

as our normative argument.

To

the extent that

unemployment

insurance washistoricallylimited,

employment

protectionprobablyservedas a partial substitute.

It ishowever a potentiallypoorsubstitute, leadingto toofew layoffsfrom the pointofviewof allocation.Thus, reforms ofthe

unemployment

system whichintroducebettermonitoringandthusallowforbetterinsurancehave the added advantageof potentiallyallowing for a decrease in

employment

protection towards the benchmark, and thus a smaller cost in terms of reallocation.

•

The

results above bear a close relation to the results obtained in the

"implicit contract literature" ofthe 1970s and early 1980s (in particular [Baily 1974], [Azariadis 1975], [Akerlofand Miyazaki1980].)

That

literature

looked at the optimal contract between risk neutral firms and risk averse

workers.

Under

theassumptionthattherewereneitherseverancepayments nor

unemployment

benefits,oneofthe conclusionswasthat therewould be overemployment, that firms would layofftoo little relative tothe efficient

outcome.

One

ofthecriticisms addressedtothosepaperswas thequestion

of

why

firmsdidnot offer

unemployment

insuranceorseverancepayments. In thediscussion here,thelimits

come

frommonitoring problems, andthe

solutiontakes the form ofalayofftax rate imposed bythe state. But the

logicisvery

much

the same.

• Returningtothediscussion of

unemployment

benefits versusseverance pay-mentsdiscussedinthe previoussection: Ithassometimes been argued that severance payments are preferable because theydonot lead tothe search

incentive problems discussed here. This is correct, but fixed payments in

the face of

random unemployment

duration deliver very poor

unemploy-ment

insurance.

Even

if search considerations imply declining

unemploy-ment

benefitsovertime, it isunlikelythatthe optimal scheduleconsistsof

a

lump

sum

payment

at thestart, andnothingthereafter.

Only if the administrative costs of setting up an

unemployment

agency appearprohibitive,does asystembasedonseverancepayments

make

sense.

Employment

ProtectionReform 13

This

may

bethecasefor

some

lowandmiddleincomecountries;itissurely

not the caseforFrance.

3

Risk

aversion

and

shallow

pockets

A

secondassumptionofourbenchmark wasthat firmswereriskneutral,had deep

pockets, and could therefore fully insureworkers (withthe help ofan agency to

run the

unemployment

insurance system, and subject to the discussion

we

just

had aboutincentives to search

when

unemployed).

This assumption isalso clearly too strong. It is based on the ideathat, iffirms arewidely held, most ofthe risk faced bya firmis diversifiable. Butwhile most

ofthe variations faced by firms are idiosyncratic,

some

are not.

And

most small

firms are not widely held.

Many

are privately held, and theirowners' wealth is

not

much

diversified. So,the assumptionofriskneutrality is, especiallyforsmall

firms,toostrong.

Even

if

we

wereto assumethat firms are risk neutral, the assumptionthat they have deep pockets, and thus can pay workers in bad states, is also too strong.

Clearly a firmthat hasgone bankrupt

may

notbeable to payits

unemployment

contributions or

make

severancepayments. But short ofthisextremecase,

corpo-rate financesuggests that the shadow price of internal funds tofirms is likelyto

be a decreasingfunction ofthestate:

The

shadowprice ofseverancepaymentsto

workersorpaymentstothe Statein badstates, eveniffeasible,

may

be high; the funds could bebetterusedforother purposes.

Now,

to state the obvious: Layoffs are more likely to take place in bad states,

when

the shadow price of internal funds is high, than in good states.

And

so, a higher layoff tax

may

potentially

make

things worse for firms, imposing a high

utility coston the small entrepreneur, or preventing thelarger firmfrom taking

othermeasuresrequiredtoimproveitssituation.

One

may

hopethat,in response

toanincrease in layofftaxes, financialmarkets willpartly adjust to alleviatethe problem, providingmorefundstothefirmsinbadtimestoallow

them

topaythe

now

higherlayofftaxes. But, morelikely thannot,theadjustmentislikely tobe

Employment

ProtectionReform 14

What

should thestatethendo?

• Separate the timing oflayoffs from thetimingof

unemployment

contribu-tions. Ideally,thestateshouldcollectlayofftaxesfromfirms ingoodstates

ratherthaninbad states.

And

taxpaymentstotheagencyshoulddepend onthe probabilityoflayoffs: Firmswhich, for

some

reason (adifferent

dis-tribution of productivity or

demand

shocks for example), have a higher probabilityof layingworkers offshould

make

highercontributions.

The

problemis

how

to designsuch asystem, oran approximationtosuch a system, in practice.

One

possibility, and that adopted for example in

the United States, is to introduce a bonus-malus, or an experience rated system.

We

returntoit below.

•

Even

the bestdesigned experience—ratedsystemsare unlikelytofully elim-inatethe additionalliquidityproblems createdbylayofftaxes. Ifso,it

may

then be optimalforthe governmenttochoose a lowerlayofftax rate, and thus a contributionratelower than one. This decreases the tax burden on

firmsin

bad

times. Obviously, itdoesso atthecost of raisinganother

dis-tortion:

A

contribution rate belowoneleads firms not to fully internalize

the costsoflayoffs, andthus leadstotoo high alayoffrate.

In the rest of this section,

we

take up two related issues. First, the design of

experience-rated systems; second, the issues raised by limited liability and the possibility ofbankruptcy, issues

we

haveleft aside upto now.

3.1

Bonus

malus,

and

experience

rated

systems

As

we

have just seen, an ideal collection system for layoff taxes is onein which the state (tothe extent that it has deeper pocketsthan thefirms) collects layoff

taxesin goodstatesratherthaninbadstates, and wherethe taxrate tobebased onthe firm specificprobability oflayoffs.

Two

obvious problems inpractice are that boththe state faced bythefirm and the probability of layoffbythe firmarelikelytobeunobservable bytheagency.

A

naturalsolutionisthentobase thepaymentsoffirmsontheirpast behavior (asin

abonus-malussystem),and toallow

them

topaythetaxes overtime. Thisisthe

Employment

Protection Reform 15

usedinparticularintheUnitedStates(Ausefuldescription ofthe U.S. experience

isgivenin [Fougere and Margolis 2000], )

The

systemsvary across U.S. states. It is usefulto describe the most

commonly

used system, called the "reserve ratio" system of

unemployment

contributions.

Leaving asidethe

many

complicated details,itsprinciple issimple:

Each firmhas a running balance with the state

unemployment

agency,with con-tributions by the firm to the fund on one side, and benefits paid bythe agency

tothe workers laid offbythe firmonthe other.

Once

ayear, thestate computes thenet outstandingbalance, andrequiresthe firm topay

some

proportionofthis

outstanding balance over thefollowing year.

The

factorofproportionalitydepends both on the net balance ofthe firm, and the net balance ofthe state fund as a whole. Thissystem hastwoimplications:

• Ignoringdiscounting,and assuming that firmsdonotgobankrupt and do not hit the various ceilings that limit contributions (all these considera-tions being veryrelevant in practice),firms eventuallypaythefull cost of

unemployment

benefitsforthe workerstheylayoff

—

thecontribution rate

isequal to one.

•

The

factorofproportionality determines

how

the timing ofpayments

de-pendsoncurrentandpastlayoffs.Ifthe factor of proportionalityisequalto one, so firms areaskedto returntozero balanceeachyear,thenpayments

are closely related to current (or more precisely last year's) layoffs.

The

lower thefactorof proportionality,the

more

contributionsdepend onpast

layoffs.

How

should one then think about the choice ofthe factor ofproportionality? If

firms are operating in a stable, ergodic, environment, going sometimes through goodtimes,sometimes through badtimes,thenlettingthefactor of

proportional-itybesmallwill

make

thefirm'scontributionsdepend onits

mean

observedlayoff

rateinthepast, whichisalsoequaltothe probability ofalayoff inthefuture. If,

however, asis more likely, the underlying probability changes over time, then a higherfactorof proportionality, givingmoreweighttorecentlayoffs, willbecloser

tothe underlying current probability. But itwill imposehigherliquiditycostson

Employment

ProtectionReform 16

3.2

Bankruptcy,

and

the role of firm

guarantees

The

possibility for firms to pay layofftaxes over time rather than at the time

the layoffs take place raises an issue that

we

have avoided so far. This is the

possibility forfirms to evade taxesbygoingbankrupt. Absentfirm guarantees,or

other commitments, a firm that lays offits workers at the

same

timeit declares

bankruptcy

may

be able to avoid paying most ifnot all layofftaxes.

And

the problem is likely to be worse under an experience rated system.

The

longer the

lag betweenlayoffs and tax payments,thelarger theproportion of layofftaxesa firmwill beable to avoidthroughbankruptcy.^

Thisislikelyto lead

some

firms toreorganizeso as to

make

iteasiertoavoidpaying

taxes.

Ways

ofdoingsoincludeisolatinghighriskdivisionsandtransforming

them

intoseparatelegalunitswithlittlecollateral,so that,incase ofbankruptcy, there

are fewornoassetslefttothe agencyoranyothercreditor torecover (leavingthe

unit "judgment proof"). Such a behavior has beenwell documented inthe case ofenvironmentalprotection (seeforexample [Ringleand Wiggins1990].) And,as recentexamplessuch asthe bankruptcy ofMetalEurop show,

some

French firms arealready

moving

systematicallyinthat direction.^"

What

should thestate then do?

• It should, as is already the case for other legal obligations, have senior creditor status.Butthere

may

notbeenoughassetsleftevenforthesenior creditors tocollect.

• It can extendresponsability for payments ofthesetaxesto third parties.

This is for example the approach taken by the law on environmental

lia-bilitypassedintheUnitedStates in1980 (alawcalled theComprehensive

EnvironmentalResponse, Compensation, and LiabilityAct,or

CERCLA).

Under

that law, if a bankrupt firm cannot pay for decontamination ofa

9. Ourunderstanding from Table 4 inFougereand Margolis [1999] isthat the proportion of

contributionsdue but not paidbecause ofbankruptcyis under 10% in most U.S.states, with

someexceptions(forexample, Californiawith 13%.)

10. Interestingly, the correlationbetweenthe stated contribution rateandthebankruptcyrate

across U.S. states appearstobesmall[Fougere andMargolis 1999].) Ifinterpreted as a causal

relation from the contribution rate to the probability of bankruptcy, this would imply that

increasingthe contribution ratemaynothavemuchimpact on bankruptcyrates. Butthere are

reasons tobeskeptical ofthiscausal interpretation. Forexample, firmswithhighriskaremore

Employment

Protection Reform 17

site, the state cangoafter any "potentiallyresponsibleparty".

What

this

means

is unclear andevolving; it

may

include suppliers, investors, or

new

site owners.

• Itcan ask formoreguarantees, intheformofcollateral, physical or

finan-cial,or ofbank guaranteestocover potentiallayofftaxes.

Guarantees, collateral, and extension ofliability to third parties all have costs.

Collateral

may

be better used for other activities. Third parties

may

prefernot

todealwith afirm altogether ifthisexposes

them

to potentialtaxhabilities.

We

havenosetview, andtheevidence fromenvironmentalprotection isstill unclear. Nevertheless,anyproposaltoincreasethe contributionrate of firmsmustconfront

theissue.

4

Insurance,

employment

protection,

and

wage

determination

A

thirdassumptionofour

benchmark

wasthatwagesweresetatthetimeofhiring.

So,totheextentthatfirms(ortheoverallsystemofunemploymentinsurance

cum

employment

protection) offeredinsuranceincase oflayoff,riskaverseworkerswere

willing to accept a lowerwage on thejob, and willingto accept lower expected incomeoverall.

This

may

not be the rightview ofwagesetting. True, initialwagesare set atthe

timeofhire.

But

theseareonlyset forashortperiodoftime,atwhichpointthey

may

be renegotiated. Atthat point, wageswill reflect the bargaining position of

eachside. This hasimportantimplications.

4.1

Ex-post

wage

setting,

and bonding

Considertwo firms:

•

One

offers severance payments to its workers, makes unemployment

con-tributions to the state, and the workers it lays offreceive unemployment

benefits.

The

otherdoes not offerseverance,does not

make

unemployment

contributions, and the workers it lays off do not receive

unemployment

Employment

Protection

Reform

18

Ifwagesaresetatthetimeofhiring,thefirstfirmwillbeabletoofferlower wages,andindeed,becauseworkersareriskaverseandvalue the insurance thefirm provides, it will have lower expected labor coststhan the second.

•

Now

supposethat workerscanrenegotiatewagesafterhiring. Then,

work-ers in the first firm will be in a

much

stronger bargaining position than

in the second. If they find themselves unemployed, they will receive un-employmentbenefits. And, ifthe firmwants to lay

them

off, thefirmwill

haveto pay severance andcontributionsto thestate. Thus, thefirm that provides insurancewill

now

have higherwages andbyimplicationexpected labor coststhanthe second.

• Giventhechoice, firmswillthereforenotbeeagertoofferinsurance. And,if

thestateputs inplaceascheduleofseverancepayments,of

unemployment

contributions, andof

unemployment

benefits, along thelines

we

described

inthebencfunark,allthreecomponentswilllead tohigherwages,andthus

tohigherexpected costs forfirms.

This view ofwage setting

may

itselfbe too extreme.

The

central issue here is

known

inlaboreconomics as "bonding"

.

Supposefirmscouldextract "bonds" fromworkers

—

thatispayments fromworkers

at thetimeof hiring to compensatefortheincrease inwagesthey

know

willtake place after hiring. Firms could then eliminate the effects ofex-post bargaining on cost.^-' Ifbonding was prevalent,

we

would be

much

closer tothe

benchmark

model, orthe extensions

we

sawearlier: Firms would bewilling to payseverance

totheworkersor

make

paymentsto the state asunder ourbenchmark.

Whatever

increaseincost this implied,theycouldrecoupthroughthereceiptofasufficiently

large

bond

atthetimeof hiring.

The

obviousremark at thispointisthat

we

justdonot observe "naked" bonding: Workers do not pay firms at the timeofhiring. Bonding however exists in more

disguised forms:

Some

workers accepttobepaid alowinitialwage,in effectpaying a

bond

earlyintheirjobtenure,to partlycompensatethefirmforthe higherwages

laterin theirjobtenure. Yet, in practice,the

room

forbondingislimited, andso

11. For anearlydiscussion oftherole andthescope ofbondinginthecontext ofemployment

Employment

ProtectionReform 19

the conclusion must bethatinsurance

cum

employment protectionis morelikely

toincreasethantodecrease labor costs.^^

What

should thestatethendo? It clearly facesatrade-off:

• Choosinga contributionrateequaltounity,i.e. asystemthat

makes

firms

payforthefull cost ofanadditionallayoff,will lead firms totake theright decisionat the destruction margin; Layoffs will take place only

when

the productivityofajobislessthanthewageequivalent ofbeingunemployed. Butthis high contribution rate will alsoincrease the bargainingpowerof

workers, and thus increase the wage. Thiswill increase the overall cost of labor, both directly and indirectly, and will adversely affectjob creation.

How

much

will dependonthe

amount

ofeffective bonding.

• Choosing a contribution rate less thanone will lead firms, in contrast, to

destroytoo

many

jobs,andlead totoo

many

layoffs.Itwillhoweverlead to

a smallerincrease intheoverall cost of labor, both directlyandindirectly

(through the effect on wages), and thus have a smaller adverse effect on employment creation.^''

Parallel argumentsapplytothe direct severancepayments part (but for one

dif-ference: In the case of the contribution rate,

we

were looking at the effects of

varying firm contributions, keeping unemployment benefits the same. Here, by the very nature of direct severance payments, we are changing both the firm's

contributions, and the benefits received by workers).

The

closer these payments

are to fully compensating for the psychic costs to workers ofbeing laid off, the

less distorted the destruction decision. But the higher is the cost of labor, both

directlyandindirectly, and sothemore distortedis thecreationmargin.

12. Thinkforexampleof publicemployees.Giventhe high degree ofemploymentprotectionand

the typicallygenerous retirementbenefits,manyarewilling tobecomepublicemployeesevenif

wagesarelowerthaninthe privatesector.But,becausetheycannotbelaidoff,exceptatgreat

cost, publicemployeesareina very strong bargainingposition,and sometimesuseittoextract

higherwagesorotheradvantages fromthestate.

13. Thereisasetoftaxesandtransferswhich canachievebothefficientdestructionandefficient

creation.

A

contribution rate of one, so there is no distortion at the destruction margin.

A

subsidy to newjobs to eliminatethe adverseeffects ofthe increasein costonjob creation (see

forexample [Mortensen andPissarides 2001]) Butthis raises inturn theissueofhowthesejob

subsidiesthemselves are financed (they may havetobevery large). Soour discussion here is

Employment

ProtectionReform 20

Inshort, the

more

firms are

made

topay forthe expectedcost of

unemployment

benefits,the smaller thedistortionwillbeatthe destruction margin, but thelarger

the distortionwillbeatthe creationmargin.^^ _{Becauseofthese}distortions,there

is

now

a trade offbetween insurance arid efficiency.

Even

ifit were feasible (if

there was no problem in monitoring the search behavior oftheunemployed), it

will no longer be optimal to provide full insurance tolaid off worlcers.

And

the optimal contributionratewill belessthan one. It willbe closer to one,

•

The

higher thescopeforbonding,andso,the smaller the adverseeffecton

layofftaxeson creation.

•

The

lowerthebargainingpowerofworkers,orthehigher the

commitment

abilityof firms and workers.

•

The more

elasticjob destructiontothelayofftax.

•

The more

inelasticjob creation. ^^

4.2

Heterogeneity

of firms

and

workers

Not

allfirms and allworkers arealike.

Some

firms operatein morevolatilegoods markets, andsoare

more

likelytohave a highlayoffrate.

Some

workers,because

of their characteristics, are

more

uncertainand more likely tobe laid off. Iflaid

off,

some

workers are likely to find ajob

more

easilythan others.

What

does a

positive contributionrateimplyfortheir respective fortunes?

To

thinkaboutthisquestion,firstgobacktoour

benchmark

case, inwhich wages

areset at thetime of hiring.

• In that benchmark, firms must offer the

same

level of utility to a given worker,otherwise theworkerwill not accept the joboffer.Thus, firmsthat

14. Thisiswhythelineofargumentoftenusedinthecontext of experience rating to argue that

the contribution rate should beequal tooneis misleading.Such a rate removesdistortionsat

the destructionmargin,but can have a large adverseeffectoncreation.

15.

A

case oftenanalyzedinthe researchonlabormarketequilibriumisthe case ofno bonding

andafullyelasticsupplyofcapital (see, forexample[Pissarides 2000]). Inthatcase,astrong

—

anddepressing

—

result emerges. The "pain ofunemployment", morespecificallythedifference

inthe value ofbeingemployedoverthe value ofbeingunemployed, remainsconstant: Whatthe

unemployedgain relativetotheemployedthrough,for example,higherunemploymentbenefits

whenunemployed, theymustlose inequilibriumthroughhigherunemployment duration.

(Oth-erwise, wageswould be too high, profitstoolow, and firmswould not createjobs).Theresult

is extreme, but an important warningnevertheless thatgeneral equilibriumeffectscanlead to

Employment

ProtectionReform 21

face

more

volatile demand, and thus higherlayoffrates, willhave to

make

higher overall

unemployment

contributions and will not be able to pass those costs on to workers through lower wages.

They

will therefore have higher costs. This is indeed as should be, given that they impose larger social costs.

• In thatbenchmark, ifworkersaresubstitutesin production, aworker with higher expected unemployment durationiflaid off willbe hired by afirm

only ifthe total costhe imposesonthe firm is the

same

asthat for other workers. Thus, workers with worse labor market prospects, will have to

accept lower wages. Atthose lower wages, firms will be willing to employ them.

Now

turn to the case where, instead ofbeing set ex-ante, wages are set ex-post through bargaining, and the contributionrate ispositive.Then:

•

As

wages are

now

likelyto increaserather than decrease in responseto a positive contribution rate, all firms will face highercosts. But, to a first

approximation, the increase in thewage willbe the

same

acrossfirms, so

the increasein costs (relativeto the benchmark) is the same at all firms.

Thereisthereforenoobvious reason

why

thecontributionrateshould thus be modulatedacrossfirms, forexample,

why

itshouldbe smallerfor firms

with a high turnoverrate.

•

An

issue arises, however, with respect to firms operating in isolated labor

markets.Takeforexamplethecase ofafirmoperatinginadepressedregion.

Ifthefirmistheonlyonearound andclosesitsplant,it

may

be verydifficult for workers to findotherjobs.

The

layoffswill have highsocial costs.This suggests imposing larger contributionson thefirm that islayingoff. But, with such large contributions, which firm will ever want toopen another plantinthat labormarket?To the extentthatthe statewantstomaintain

employment

in the region, the solution is probably not to modulate the

contribution rate,but rather to usejob creation subsidies.

•

The

situation is

now

different forworkers. Workers

who

are perceived by

firms to be more risky, inthe senseof eithera higher probability that the

workerwillhavetobelaidoff,or ahigherexpectedunemploymentduration

Employment

ProtectionReform 22

hired, they will be able torenegotiate the wage, and thus increase firm's

costs.

The

increasein costs will belarger, the higher the probabilitythat

theworker

may

belaidoff,orthelongerhisorherexpected

unemployment

duration.

Knowing

thishowever,firms will notwant to hiretheseworkers

inthefirst place.

Thus, a positivecontributionrate (in general any employment protection)

will lead toincreased discrimination by firms inthe labor market.

Work-ers with a short labor market history,workers with poorskills, and older

workers,

may

have a hard timefinding jobs.

What

should thestatethen do?

To

reduce the problem of different ex-ante probabilities oflayoffs for different

workers, a natural, if partial, solutionis to give time toboth partiesbefore the usual rules of

employment

protection and

unemployment

insurance apply. This

may

take two,not mutuallyexclusive, forms:

•

A

trial period, during which any ofthe two parties can separate at no

cost. This period must be long enough to allow the firm to learn about

the worker, but short enoughto

make

it unattractive forfirms to fill jobs through rotations oftrialperiod workers.

•

A

transitionperiod during which,incase of separation,boththepayments bythe firmtotheagency, andthe

unemployment

benefits receivedbythe

laid offworkerarelessthanunderstandardrules,andconvergetostandard

levelsas seniority increases.

To

the extent that thisisnot enough,one

may

thinkofothertypes of solutions,

forexample:

• Targeted hiring subsidies, but this is likely to add yet another layer of

complexity and arbitrariness in the

employment

protection system.

The

French experience isnotparticularlyencouragingin this respect.

• Or/ and a contribution rateby firmswhich depends onthe

number

of

lay-offs, rather on than the expected or actual total

unemployment

benefits

paidbytheagencytotheworkers laid offbythefirm.This secondsolution

does not eliminate theproblemraised bydifferent ex-anteprobabilities of

layoffs fordifferentworkers, butiteliminatesthe problemraisedby

Employment

Protection

Reform

23

5

Quits

and

layoffs

We

have focusedso far on adverseshocks to productivity, shockswhich lead the firmto lay aworkeroffeveniftheworkerdoes not have another job opportunity.

The

reasonwasthatthese areobviously the shockswhere unemploymentinsurance and employment protection

may

have aroleto play. Butthese are not the only shockstriggering separations. Workersleaveforotherreasons,oftenbecausethey havea more attractivejob opportunityelsewhere. In Francetoday, leaving aside

the separations that take place at the end of fixed duration contracts

(CDD),

layoffs account only for about one third of separations, quits for the remaining

twothirds ([Goux andMaurin 2000]).

The

presenceofboth layoffsandquitsintroduces a

number

of issuesinthedesign

ofemploymentprotection, and these aretheissues

we

discussin thissection.

5.1

Introducing

quits euid layoffs

Go

back toour

benchmark

model.

Assume now

that there are two shocks that takeplace aftera worker hasbeenhired. First,asbefore,productivityis realized.

Second (and simultaneously), with positive probability, the worker receives an outside job offer. Suppose, for simplicity, that ifthe outside job offer comes, it

dominates anyofferthe firm can

make

to theworker.

Thereare

now

tworeasons

why

there

may

aseparation. Productivity

may

below,

andtheworker becomes unemployed. Let'scall thisalayoff; it isinitiatedbythe

firm. Or, the worker

may

receiveanoutsidejoboffer, inwhichcase hewill leave.

Let'scall thisaquit.

Ifthelevelof productivity,andthe existence ofa joboffer,arebothobservable,and

ifrequired,verifiable in court, the conclusions

we

reachedearlier extend

straight-forwardlytothis case.Ifa separationcomesfrom lowproductivity,andistherefore

alayoff,firms

make

contributions to thestate,and payseverancepaymentstothe

worker. If a separation comes from an outside job offer and istherefore a quit,

it triggers neither severance payments, nor unemployment contributions by the

firm, norunemployment benefitsto the worker.

The

problems arise

when

the reason behind the separation is unobservable, or

Employment

Protection

Reform

24

two types ofpotential games, first between firms/workers and the state, second betweenfirmsand workers themselves.

5.2

Games

between

the

firms/workers

and

the state.

Focusfirst on the payments from the firmtothe state to finance

unemployment

benefits.

And

assume, forreasons

we

discussedintheprevioussection, that firms

support only a proportion ofthese costs: the contribution rate is less than one.

This opensthe possibility that, for the firm andthe workertakentogether, each

layoff

may

be associated with a net subsidy from the state (the firm pays less

to the state than the

payment

ofthe state tothe laid offworker). Thus, to the extent that firms and workerscollude, they

may

have an incentive to call every separation alayoff.

Isthislikelytobeaserious issue in practice?Probablynot:

The

twoparties

may

haveneitherthe abilitynor the incentive to collude:

• It

may

not be easyforthe firmandtheworkerto collude. Collusionimplies

a payment from theworker tothe firm, so as to offset the

payment

from thefirm to the state.

To

the extent that the

payment

comes from future

unemployment

benefitsor future wagesreceivedbythe worker, theability ofthe firm to

make

sure that such payments actually take place

may

be

limited.

•

The

exact nature of contributions by firms to the agency matters here.

If contributions by firms depend on actual

unemployment

benefits paid

tothe workers

who

were laidoff, then indeed firms and workers together

benefit from calling a quit a layoff. Suppose instead that contributions depend (for the reasons discussed at the endofthe previous section) only on the

number

oflayoffs, or, equivalently, onthe

number

oflayoffs times the averagedurationof

unemployment

benefits.Inthis case,itis

much

less

obvious that the firm andthe worker togetherwill benefit from declaring

aquitto be alayoff:

Workers

who

quitarelikelytohave ashorter

unemployment

durationthan

average, andthusreceivesmaller

unemployment

benefits,than average. In particular,

many

ofthequits are directly toanotherjob, inwhichcasethe

Employment

Protection Reform 25

gain totheworker, and alosstothe firmindeclaring it alayoff. Thereis

noincentivefor

them

tocall thequit alayoff.

5.3

Games

between

workers

eind the firm

Assume now

that the contribution rate is one, so that

we

can ignore the

previ-ous

game

between firms/workers and the state. There is another

game we

have

to consider: Otherthings equal, firms would ratherhave a separationbe called a

quitand saveonseverance payments and unemployment contributions.

Symmet-rically,workerswouldratherhave a separationbecalledalayoff,andreceiveboth

severance payments and

unemployment

benefits.

Iftheworker could not affect the productivity ofthe match, and the firmcould

not affect the relative attractiveness ofthe outside option ofthe worker, there

wouldstillbe noproblem.^® Firmswith a low productivity shock could not force

the worker toquit. Workerswith an outsidejob offercould not force the firm to lay

them

off.

But,in fact, workers canaffect the productivityofamatch, andfirms canaffect

therelativeattractiveness ofthe outsideoptionofthe worker:

A

worker

who

wants to quit but also wants to receive severance payments and

unemployment

benefits, can shirk and decrea.se the productivity of the match,

leavingnochoice tothe firmotherthanto lay

him

off.

A

firm thatwantsto lay a workeroffbutwouldratherhave

him

quitso as tosaveonseverancepayments and

unemployment

contributions, can harass the worker into quitting.

The

stronger

the stakes,that is the higher the contribution rateand the higherthe

unemploy-ment

benefitsorthe severancepayments, the higher theincentives toharassor to shirk.i^

As

in the

game

betweenfirms/workers andthe state

we

discussedearlier,there is

arelevant difference between severance payments and

unemployment

benefits. If

16. This statementmaybetoo strong,astheremightstillbesomeroomforgaming.Iffor

exam-ple,theworkerreceivesanoutsideofferandthefirmsimultaneouslyreceivesabadproductivity

shock,both have anincentiveforhavingthe other side takethe decision to separate.

17. [AndersonandMeyer1998]showthat the1985 increaseinthe contribution rateinthe state ofWashingtonledtoa substantial increaseinthenumberofdenial ofbenefitcasesbroughtby

Employment

ProtectionReform 26

a workerhas anoutsidejoboffer, itmakessensetoshirk,so as tobe laidoff, and

receiveseverancepay. But, iftheworker intends totake the other job right away,

shirkingsoastobelaid offandreceiving

unemployment

benefitsisofnovalue to

that worker:hewillnot be unemployed.

This hastwoimplications:

Unemployment

benefitsare, inthatrespect,lesslikely

tolead to

gaming

than areseverance payments. Shirking byworkers

may

beless

ofan issuethan harassmentbyfirms.

Until now, in our argument, there wa.s no reasonto have courts involved inthe process of separation (except for the usual reasons:

Making

sure that existing

rules

—

payment

of severance, advance notice, no discrimination on the basis of

sex,age, physicalappearance,nolayoffbecauseofunion activity,andsoon

—

that

arein place arenot violated.) Buttheissues

we

justdiscussed

now

createsuch a

role. Let'sturn tothis.

5.4

The

role of courts

Under

thelogicofour arguments,what courtshavetodoisconceptuallyclear (if

not necessarilyeasytodoin practice):

• Ifaseparationhasbeenreportedasalayoff,look, ifrequestedbythefirm, atevidenceof shirking bythe worker. (Thiscan take differentforms,with

differentwaysofallocatingtheburdenofproof

A

firm that does notwant

topay

unemployment

contributionsandseverancepayments

may

statethat the separation istheresult ofmisbehaviorbytheworker, and,ifchallenged bythe worker, has toproveit incourt).

• Ifa separationhas been reportedas a quit,look, ifrequested byworkers,

at evidenceofharassment bythefirm.

An

importantremarkat thispoint,towhich

we

shallreturnafterhaving described therole ofcourts inthe current French employmentprotection system:

The

role

ofcourtsdescribedabove isverydifferent fromtheirrole in Francetoday. In

par-ticular,inourframework,ifafirmiswillingtocall aseparation alayoffand

make

the associated paymentstothe state and tothe worker, there isnojustification

forthe courttosecond guess thedecision ofthefirm,nojustificationforthecourt

Employment

Protection Reform 27

6

The

contours

of

optimal

employment

protection

The

purposeofthissectionis simplytosummarizethe

main

conclusionsreached

inthe previousfivesections.

•

Employment

protection is a natural counterpart to

unemployment

insur-ance.

A

fulldiscussion ofunemploymentinsurance fallsoutsidethescopeofthis

chapter. Nevertheless, letus

make

a fewremarks here.

Individual selfinsurance isnotsufficienttoinsureworkersagainst therisk

ofjob loss and unemployment. Perhaps

more

could and should be done here (for example along thelines ofthe

unemployment

accounts proposal presented in [Feldstein and

Altman

1998]; see also [Kugler 2002], for an

analysisofseverancepaymentsavings accounts in Colombia). Inany case,

we

takeas given in this chapter that suchprivate accounts cannot simply

replace traditionalunemploymentsystems.

Thisimpliestheneedforanagencytoadminister

unemployment

insurance.

This agencycan be eitherapublicagency,ora public-privatepartnership:

Onlythestatehas therequiredadministrativeinfrastructure, to followthe unemployed,to distributebenefits,andtocollect contributionsfromfirms.

A

publicagency

may

howevernot haveall theright incentives.

We

seethis in

some

ofthe problems emerging inthe implementationofthe

PARE

in

France.

The

PARE

represents an attempt to provide more generous (in

time)

unemployment

benefits, in exchange for stronger incentives forthe unemployed toacceptjobsifsuch jobsare available.

Agency

employeesdo not howeverhave strong incentives to force the unemployed totake jobs,

andthe preliminary evidencesuggestsanincrease inbenefitshasnot

come

with

much

strongerinducementsfortheunemployedtotakejobs.

A

public-privateagencywouldhavestronger incentives to placetheunemployedinto jobs. ^*

•

The

general principle should be that firms

make

payments to the

unem-ployment agency equal to the expected or actual

unemployment

benefits

paidtothe laid offworkers. Inother words, the contribution rate offirms.

18. For a similar discussion ina differentcontext (Who should runprisons?), see [Hartet al

Employment

Protection Reform 28

defined astheratio ofcontributionspaid by the firm to the (expected or actual)

unemployment

benefitspaid tothe worker,should be equal to one.

Sucharate leads firms tofullyinternalizethesocialcost oflayoffs andtake anappropriate layoffdecision.

•

The

principle is important. But a

number

of other imperfections in the laborand othermarketsrequire however a

number

of qualifications:

To

the extent that

unemployment

insuranceis necessarily incomplete (for

exampleto maintain incentives to search), it is then optimal to choose a contribution rate larger than one, therefore decreasing layoffs below the

efficient level, but in doingso, providing

more

insurance toworkers.

To

the extent that firmshaveliquidityproblems, ahigh contributionrate

andthe

payment

of

unemployment

contributions

may

create serious prob-lems for firmsalreadyinfinancialtrouble. Inthis case,it isbetter to sepa-rate thetiming of

unemployment

contributionsbyfirms and the

payment

of

unemployment

benefits, according, forexample, toabonus mainsoran experience rated system. It

may

also be optimal to choose a contribution

ratelessthanone, so as todecreasetheburden onfirmsinfinancialtrouble,

again at

some

cost in efficiency.

To

the extent that wages do not fully reflect the provision of insurance,

a contribution rate equal to one will avoid distortions at the destruction margin, but it will also increase labor costs, decrease profits, and thus

create distortions at the creationmargin. Inthis case, it is again optimal

to balance the two distortions by choosing a contribution rate less than

one.

On

net, given ourstate ofknowledge

—

theoretical and empirical

—

no one canstatewith

much

confidence whatthe optimal contributionrateshould

be.

Our

guess,butit ishardly

more

than aguess,isthat thelasttwofactors

dominatethe first,andthe contribution rateshouldbe positive, butbelow

one.

•

A

related questionis whether the contribution rateshould be modulated

Employment

Protection

Reform

29

Some

sectors and

some

firms have a

much

higher turnover than others.

This turnover will decrease as the contribution rate is increased.

But

it

is likelythat

some

sectorswill continue to have higher turnover andthus

higherlayofftaxcosts.Thisishoweverasitshouldbe:Thesesectorsimpose highercostsonsociety,andthisshouldbereflectedinhighercostsforfirms

inthose sectors.

The

contribution rate

may

howeverhave tobe modulated across workers.

Some

workersare

more

uncertain andthus morelikely tobe laid offthan

others;

some

workers have higherexpected

unemployment

duration than

others;this

may

bebecauseof age, ofskill,orothercharacteristics.Ifthese workers accepted sufficiently lower wages, firms would be willing to hire

them. But, in the presence ofwage floors, orex post wage setting, wages

are unlikely to adjustenough andthese workersare likelyto cost

more

to

employ. This in turnwill lead firms to discriminate against workers

who

are,or are perceivedas,

more

likelytobe laidoff, ormorelikelytoremain unemployedfor alongtime.

Partial solutionsare atrial period during whichseparationcanhappen at

nocostto either party,andatransitionperiodduringwhich

unemployment

benefitsandemploymentcontributions arelowerthanunderstandardrules,

and increasewithseniority.

Other partial solutions include targeted hiring subsidies (but experience suggests that there are

many

pitfalls with such targeted subsidies), and

unemployment

contributions by firms that depend not on actual or

ex-pected

unemployment

benefits paidtolaid offworkers, butonthe

number

oflayoffs. In this case, firms have no incentives to discriminate against

workers with longerexpected duration ofunemployment.

• Inthe eventofbankruptcy, unpaidoutstanding layofftax balancesshould be counted as a liability of the firm, and the state should be a senior creditor.

As

the experience with environmental liabilities has shown, this

may

not beenough: Firms

may

systematically reorganize and spin off risky units

so as to leave

empty

shells in the event of bankruptcy. In this case, it

may

be desirable tohave theoutstanding liabilitiesto the

unemployment

Employment

Protection

Reform

30

the monitoringofthebalancesheet of firms tobanks orother creditors.

•

The

previous points have concentrated on contributions by firms to the

state. But there isalso a potential role forseverance payments, payments

made

directlytoworkers.

Their role should not be to help workers finance unemployment. This is

betterdone through

unemployment

benefits. Their roleshould be to

com-pensate, at least in part, for the costs ofbecoming (as opposed tobeing) unemployed.Thesepaymentsshouldbea(non hnear) functionofseniority,

with lowpaymentsuntil highseniorityhasbeen achieved.

Thus,on the financialside, employmentprotection could take two forms:

Unemployment

contributionstothestate; whilethese arenot directly vis-ible to workers, they protect employment in the sense of

making

layoffs

more

expensive for firms.

And

severance paymentsdirectlytoworkers.

•

The

roleofthejudicial systemshould thenbe,inadditionto

making

sure that administrative steps are followed, to assess whether declared layoffs

are indeedlayoffs, and declared quits are quits.

To

avoid paying

unemployment

contributions and severance payments,

firms

may

harassworkers into quitting. Inorderto qualify for

unemploy-ment

benefits andreceive severance payments, workers

may

shirk so as to

belaid off.

The

roleofthejudicialsystemshould then betwofold.Ifasked byworkers,

tolookforevidenceofharassmentofworkersifa separation hasbeencalled

aquit. Ifasked byfirms, tolook forevidenceofshirking byworkersifthe

separation hasbeen calledalayoff.

The

role ofthejudicial systemshould nothowever betosecond guess the

layoff decisions of firms. If a firm is willing to call a separation a layoff,

follow the relevant administrative steps, and pay the associated financial

costs, thisdecision should not be subject to judicial challenge (except on