Tilburg University
Risk sharing under incentive constraints
Wagner, W.B.
Publication date:
2002
Document Version
Publisher's PDF, also known as Version of record
Link to publication in Tilburg University Research Portal
Citation for published version (APA):
Wagner, W. B. (2002). Risk sharing under incentive constraints. CentER, Center for Economic Research.
General rights
Copyright and moral rights for the publications made accessible in the public portal are retained by the authors and/or other copyright owners and it is a condition of accessing publications that users recognise and abide by the legal requirements associated with these rights. • Users may download and print one copy of any publication from the public portal for the purpose of private study or research. • You may not further distribute the material or use it for any profit-making activity or commercial gain
• You may freely distribute the URL identifying the publication in the public portal Take down policy
If you believe that this document breaches copyright please contact us providing details, and we will remove access to the work immediately and investigate your claim.
CBM
,//
K.U.B.Bibllotheek Tllburg
Risk Sharing under Incentive
Constraints
PROEFSCHRIFT
ter verkrijging vandegraadvandoctor aandeKatholieke
Univer-siteit Brabant, op gezag van derector
magnifcus, prof. dr. F.A.
van der Duyn Schouten, in het openbaar te verdedigen ten
over-staan van een door het college voor promoties aangewezen
com-missie in de aula van de Universiteit op vrijdag 14 juni 2002 om
10.15 uur door
WOLF BERNHARD WAGNER
Preface
Thepresentthesis isthe resultofthree and ahalfyearsofresearch atthe Center for Economic
Research at Tilburg and at the University of Bonn. It was aproductive time, and I enjoyed
writing this thesis. Several people have
contributed to that, and I
am happy to express my appreciation.My foremost thanks go to my supervisor Sylvester Eijffinger. Sylvester's enthusiasm and
skillful advice formed theoptimal researchenvironment for me. Inhissupervision, heshowed
the skill of providingthe large amount offreedom I neededto pursuemyideas and interests,
without forgetting the fact, however, that I eventually had to write a thesis. The relaxed
discussions in hisofficeshaped my research to aconsiderable extent -even though thesmoke
from two cigarssometimes madeit difficult to spot him on the other side of the desk. He also
showedanadmirable patienceasregardsmypassionforchangingresearchinterests; apassion
which,
starting with
my Master thesis, took me on ajourney from money demand functionsover European monetary integration, status-seeking, devolution and central banking to risk
sharing. Furthermore, Sylvester is the co-author of two papersonwhich chapters ofthisthesis
are based, and has in that way essentiallycontributed tothisthesis.
I alsofeel indebted to Oved Yosha from Tel
Aviv
University. The discussions with Ovedhavebeenessential for thisthesis. Hisinterest in my work wasagreat encouragement for me.
I am therefore very pleased that heiswilling to act asamember of mythesiscommittee.
Most of
this thesis waswritten
in Tilburg. The staff ofthe Department of Economicshas provided me in the last two and a half years with the efficient and smooth environment
necessarytofocusexclusivelyon research,which I appreciate very much. I would also likethank
my colleagues and friends in the department, in particular Riccardo Calcagno and Michael
Krause, andall others who have, in one way or another,contributedtomakingitpleasant and
effective to work inTilburg.
In Bonn I acquired the technical tools
that
proved necessary in this thesis, and I wouldlike express my gratitude for that. Special thanks go tothe members of the 'Herrenzimmer',
Mathias Drehman and Gerd Muehlheusser. I owe them manydiscussions about the meaning
of
economicscience.Part of thisthesiswaswritten duringafour-month stay at Tel
Aviv
University. Theenvi-ronmentthere wasverystimulatingand challenging. I want tothank theeconomicsdepartment
fortheirhospitality. Iwould especially like tothank ItayGoldstein andJules Leichter for the
regulardiscussionsabout risksharing.
This thesis has alsobenefited fromthe comments
of
seminarparticipantsat variousconfer-ences. Thefinancial support by CentER, CEPR and theDeutsche Forschungsgemeinschaft is
gratefully acknowledged.
In
particular, discussionswithin
the RIN-network on International Capital Flows were very helpful.I am honored that Lans Bovenberg, Harry Huizinga, Theo vandeKlundertand Casper de
Vriesshowinterest in my work andarewillingtoparticipate in myPh.D. committee.
envi-appreciation for that. My particular thanks go tomy parents, whoare ultimatelyresponsible
Contents
1 Introduction 1
Part I International Risk
Sharing 11
2 The Welfare Gains from Incentive Compatible International Risk Sharing 13
2.1 Introduction . . . 13
2.2 Incentive Compatible Risk Sharing . . . 14
2.3 Measuring Welfare Gains . . . 19
2.4 Results . . . 21
2.5 Summary
. . . .
. . . 253
International Diversification and Governmental Moral Hazard
27 3.1 Introduction . . . 273.2 Related Literature . . . 28
3.3 TheModel 30 3.4 Discussion andExtensions . . . 33
3.4.1 Foreign Direct Investment and Centralized Risk
Sharing . . . 35
3.4.2 Optimal Degreeof Diversification .. . ... .. .. ... ... 37
3.4.3 Constraintsto OpportunisticGovernment
Policies . . . 40
3.5 Summary . . . 41
Appendixto
Chapter 3. . . 42
)A Consumption under FDI . . . . . 42
3B Optimal Home Bias . . . 42
4
Capital Taxation under International Diversification: Tax Exportation
ver-susThx Competition
47 4.1 Introduction 47 4.2 PublicGood Provisionwith
ForeignOwnership . . . 49
4.3 Capital TaxationunderFull Financial Integration . . . 51
4.4 Empirical Evidence on theTaxExportation
Effect . . . 58
4.4.1 Data . . . 59
Appendixto
Chapter 4. . . 66
5 Summary
and
Conclusions to Part I 69
Part II
Entrepreneurial Risk Sharing 73
6 Divestment, Entrepreneurial Incentives and the Decision to Go Public 75 6.1 Introduction
. . . , , , . . , , , , , , , , , , 75
6.2 The StockMarketand Entrepreneurial Incentives . . . 78
6.3 The Dynamic Divestment Problem . . . 80
6.4 The Decision toGoPublic and theLife Cycle of the Firm ... .. . .. . 82
Chapter 1
Introduction
This thesis deals with risk sharing among individuals and countries, and in particular the
restrictions opportunism ofeconomic agents can impose on risk sharing. The study of risk
sharing and the understanding of its potential limitations is ofconsiderableinterest because
ofthe importance of risksharing for economicwelfare. Inthe firstplace,risk sharing creates
potentially large benefits for individuals by reducing consumption volatility. Moreover, risk
sharingmakesindividual consumptionlessproneto idiosyncratic risk and in thatwayreduces
theriskpremiumsof individualincomes. Thisgives riseto variousadditionalbeneficialeffects
due to amore efficient allocationofeconomic resources. For example, diversification of
firm-specific risksreduces the required return on
capital. This will
enhanceinvestment and alsospecialization, and will have consequences for productivity and growth. Equally, areduction
inindividual labor income risk due to risksharingincreasesthe incentives to invest in human capital andtospecialize.
A
deviation from full risksharing may, in view ofitsbenefits, indicateimportantinefficiencies inthe economy and shouldberelevant forpolicy-makers. Asamatter of fact, the empirical evidence conclusively rejects theimplications of fullrisk sharing; the aim
ofthis thesis is to add tothe understanding of this lack ofrisk sharing.
Most of thisthesis
will
focus onthe internationaldimension ofrisksharing. Inthisrespect,the presenceofcountry-specific risk (suchaspolitical,economicornatural risk)shouldmotivate
individualstosharerisksacrosscountries. In particular,standarddiversificationconsiderations
wouldsuggestthat individualscompletely poolcountry-risk.
In
contrast,some evidencepointsto a significant lack ofsuchinternational risk sharing. First, for most countries the
volatility
of consumption is higher than the
volatility
of output.If
there was significant risk sharingamong countries, country-specificshocksto outputwould not be borne bythe country alone,
which would make consumption less volatile than output. Second, under full risk sharing
consumptionshould grow at the same rateacross countries (this is because shocks are borne
by allcountries and countries
will
furthermoresmooth consumption). However, consumptionpaths differ widely acrosscountries. For example, (real) consumption in the U.S. rosesince
the 1950s by roughly 150%, while Japan's consumption increased by more than 450% (see
also Figure 2.1 on page 24). Moreover, studies have suggestedthat there is asmall degree of
insurance among countries. For example, S0rensenand Yosha (1998) found that only 5% of
the shocks to national income are insured across OECD countries. Further evidence against
full
risk sharing comes from individual portfolios. Full diversification of country-specific riskwould imply that individuals hold internationally balanced portfolios, but there is a strong
home biasinportfolios. In most industrialcountries,individuals hold more than 80% of their
portfolio indomestic assets. 1
What canexplain this lack of international risk sharing? Thefirst (and obvious) guess is
that
the welfare gains fromrisksharing beyond current levels arejust small, andhencealreadyminor frictions can makea largerextent of risk sharingnot worthwhile. The absence of full
internationalrisk sharing should then not bea cause for concern.
A
large number ofpapershave addressed this issue and haveestimatedthe welfare gainsfrom moving from the current
degree ofrisk sharing to full risksharing. These paperstypically find thedirect welfaregains
(that arise from the reduction in consumption
variability) to be in
the range of 0.5 to 5% of permanent consumption (some papers find even much larger welfare gains). The lack ofinternational risk sharing cantherefore not beexplained by small welfaregains, especially if taking into account the additional effects on investment and specialization (which caneasily
exceed 100% of the wealth
of
individuals). The literature haslabelledthese apparently highunexploitedwelfare gainsfrominternational risksharing'theinternational risk sharing puzzle'.
A lackof international risksharing may also stem fromadditionalcostsofinvesting abroad.
If, for
some reason,foreign investors are atadisadvantage comparedtodomestic investors, it is plausiblethat individuals willbiastheir portfoliostowards domestic assets. Highertransactioncosts faced by foreigners may form such a disadvantage. However, at least among industrial
countries, transaction costs arenot considered large enough toexceed the substantial welfare
gainsfrom international risksharing. Thereis, moreover, indirectevidence againsttransaction
costs as an explanation of the home bias: studiesof equity markets in severalcountries have
revealed that the turnover rateof equity for foreign investors is significantly higher than for
domestic investors, which is exactly opposite to what one would expect in the presence of
highertransactioncostsfor foreigners. Asymmetric informationbetween domestic and foreign
investors may constitute a further barrier to the acquisition
of
foreign assets. Domesticin-vestors may, for example, be better informed about the prospects ofdomestic firms. While
this may beavalidexplanation for the home biasindirect investment, market efficiency would
rule out
any advantage of domestic over foreign investors for portfolio investment. This isbecause unlessinvestorsbehavestrategically, anyprivateinformation willberevealedthrough
prices. Moreover, existing financial instruments (such as index-certificates orindex-tracking
funds) allow investors to buy the whole market. This overcomes potential lemon problems
forforeign investors sinceinvestorscan secure themselvesthe averagemarket performance by
investing inthe market index. Therefore, asymmetric information does not offeraconvincing
solution tothe (portfolio) homebiaspuzzle. Further impediments to foreign investment arise
l In general, there is not necessarily a one-to-one relationship between the home biasin portfolios and the lack of international risk sharing: the formerconcernscapital income risk, while internationalrisksharing relates to consumption risk (which ultimatelymatters for welfare). The analysis inthis thesisconcentrates
1. Introduction 3
from government interference. Despite the fact that sincethe collapse of the Bretton-Woods
system capital canmoveinternationally withoutrestrictions, foreign and domestic incomes are
still
treated differently. Withholding taxes incombinationwith
imperfect tax credits forfor-eign incomeincreasetherequiredpre-taxreturn onforeigninvestment and, in thisway, favor
investment at home. Foreigners may also face the riskof expropriation. While the impact of
these costs isgenerallydifficult toassess, it isquestionable whether theycanoutweigh the high
welfare gainsfrom internationalrisk sharing. Even if they do, theycannot directly explain the
lack of international risk sharing; instead
it
raises the question why benevolent governments imposerestrictions thathinder risk sharing by individuals.A potentialcause for the lackof international risk sharingin portfolios isthatequity may
provide a hedgeagainst other risks, such as fluctuations in good prices and human wealth.
For example,
if
domesticlaborincomeisnegatively correlatedwith
domesticcapitalincome, itmaybeoptimal to biastheportfoliotowards domesticassets. While itiscontroversialwhether such additional riskscanrationalize the home bias in portfolioinvestment, ahedging motive saysnothing about the lack ofrisk sharinginconsumption. Summarizing theabovearguments and other contributions in the literature, the profession has not yet settled on a conclusive
story for the lack of international risk sharing. In particular, the 'international risk sharing
puzzle' remains basically unresolved (e.g., Lewis, 1999, and Athanasoulis and Van Wincoop,
2000).
This thesis considers incentive constraints on the side of the government as the reason
for the absenceof international risk sharing. Suchincentive constraints basically arise from
two sources: the non-enforceability of international contracts and governmental moral
haz-ard. They can limit international risk sharing as follows. In the absence ofa supranational
organization that canenforceinternational contracts,international risk sharing has to be
self-enforceable. However, acountry that has a favorable endowment realization, and, therefore,
expects netoutflows fromrisksharing, may haveanincentivetodefault onitsobligations from
risk sharing. Thiscanrestrictthe extent to which risks canbeshared. Toillustratethe severity
of
the problem: Japan's consumption in the 1990s was at 550% of the level in the195Os, whileaverage consumption of theG7countries was only at 270%. Assuming that this development
was unpredictable, full risk sharing (started in the 195Os) would imply that Japan now loses
roughly 50% of its yearly endowment due torisksharing! Moral hazardarises because inter-national risk sharingreduces the benefitsacountry receives from its income. The reason for
that is that due to
the complexity of country-specific risks, international risk sharing takesplacebasicallyvia aggregates (such as national income or production) and not viacontracts
that condition on
the risks themselves. Most economic policies, however, affect productionand income, and in response torisk sharing the government may find the policy mix chosen
under autarchy not optimal anymore. For example, one reasonfor Japan's higher growth in
thepost-war period are high investmentsin infrastructure. Would Japanstill havemade these
tremendous investments, hadit known that
it
would have tosharethe benefits from it with themay then make alargedegreeof international risk sharing undesirable.
While there is a substantial literature on the explanationsoutlined above for the lack of international risk sharing,contributionsbasedongovernmental incentive constraints are rare. The limitations moral hazard andthe non-enforceabilityofcontracts place on contractual
re-lationships have beenextensivelystudied ina generalcontext,however, with few applications
to international risksharing. Anexemptionarestudiesthatexaminetheimplicationsofincen- '
tive constraintsforinternational capital flows. They findthatincentiveconstraintscanexplain somepuzzlingpropertiesof capitalflowsbetweencountries (Gertlerand Rogoff, 1990, and
Ak-teson, 1991). Inthe RBC-literature, it has beenmoreoverfound thatthe non-enforceability of
contracts helps to resolve important anomalies in international macroeconomic data (Kehoe
and Perri, 2000). Governmental moral hazard iscommonly named asacause forthe absence
of a large degreeofinternational risk sharing (e.g., Obstfeld and Rogoff, 1996, p.416f). Yet, formalcontributionsthatanalyze thetrade-offbetweeninternational risksharing anddistorted
incentives posedby moral hazardare absent.
The main aim of this thesis is to fill this gap and to study the implications ofincentive
constraints for international risk sharing. Questions that will be central to the analysis are:
How dooptimal risk sharing allocations under incentiveconstraints look like? Are there
still
substantial welfare gains from theimplementation oftheseallocations? Is there anyevidence
that
incentive constraints may, in practice,limit
international risk sharing? And, vice versa,isthe presenceof international risk sharingimportant for governments' incentives? What are
the welfare consequences of additional risk sharing in the presence of incentive constraints?
Is the degree of risk sharing chosen by individuals socially optimal? Is there any reason for
the government to interfere in the risksharing markets? What can besaid about the welfare
propertiesof distorted incentives dueto internationalrisk sharing inthepresenceof additional
distortions, such as those arising from tax competition? Furthermore, this thesis has the
ambition to
gain insights into the effectivenessof
existing international risk sharing, whichraises questions such as: Is the current level of international risk sharing efficient? And, if
not: How canrisksharing among countriesbeimproved? What is the roleofgovernments and
individualsin improving international risksharing?
These issues are
dealt with in
thefirst part of
the thesis. The second part considers adomesticrisk sharingpuzzle. Ashouseholdsaresubject to country-specific risk, entrepreneurs
have to bear firm-specific risk. Since there are no markets that allow to adequately deal with firm-specific risk (this is particularly true for small firms), the only possibility for an
entrepreneur to reduce her exposure to
firm-risk is to sell a part of the firm and
to investsthe proceeds in otherassets. The paradigm of full risksharing would, therefore, predict that
entrepreneurs sell alarge part of
their firm at
thefirst opportunity (since the size ofa singlefirm
is negligible compared to the whole market portfolio,the entrepreneurshould essentiallysell the whole firmin order to achievemaximal diversifcation). However, thereissubstantial
evidence against such behavior. Entrepreneurs tend to keep alarge stake in their firm, even
1. Introduction 5
the firm in pieces and not all atonce. Analogous to theinternational risk sharing puzzle, this
evidencesuggests a 'finance risk sharing puzzle'. Thispuzzle comes in two dimensions. First,
the evidence shows that entrepreneurs are able to sell the firm. Why doentrepreneurs then
not sellthe complete firm (or at least alarge stake) as required by full risk sharing? Second,
given that the entrepreneur wants to sell a certain stake
during the life of the firm, why is
this amount soldinpieces? Efficientrisksharing would suggest that the entrepreneur sells the
wholestake at theearliest opportunity inorder toobtainthe benefits from diversification as
soonas possible.
The aim ofthesecond part of thisthesis is to examine whether entrepreneurial incentive constraints can explainthispuzzle. Thefirst incentive constraintfaced byentrepreneursarises
from her
inability
to commit herself to futureactions. In particular, entrepreneurs may notbe able to commit themselves to future financing policies. Second, entrepreneursaresubject
to moral hazard, i.e., the effort an entrepreneur puts into the firm depends on how much
the entrepreneur benefits from the effort. This again poses atrade-offbetween risk sharing
and distorted incentives, which, ina static setting, is by now well understood in the finance
literature. In thisthesis, the dynamicimplications ofincentive constraintsareexplored. The
focus ofthe analysis
will
thereby be, beyond the above-mentioned puzzle, also on whether incentiveconstraints canexplain stylized facts of the life cycle of the firm.The remainder ofthe thesis is structured as follows. The rest of this
introduction will
give an overview ofthe chapters in this monograph and
will
outline the main conclusions.For readability, the text in this introduction only contains the mostimportant references. At
the end of the introduction, there is a
literature list that
contains the references and someadditional reading for the issuesraised inthis introduction. Part I ofthe thesis, whichdeals
with international
risk sharing, consists of Chapters 2-5, where Chapter 5 has a concluding character. Entrepreneurial risksharingis dealt with inChapter 6 (Part II). The chapters allstand ontheirown (beyondthis introduction), andare based onthe followingpapers: Eijffinger
and Wagner, 200lb, (Chapter2),Wagner, 2000 and200lb(Chapter3),Eijffingerand Wagner,
200la (Chapter 4), and Wagner, 200la (Chapter 6).
PART I The main aimofChapter 2 is
to
studywhether incentive constraintscansolve theinternational risksharing puzzle,i.e.,whetherthewelfare gainsfrom international risksharing are still largeonceincentive constraintsaretakenintoaccount. The analysis starts byoutlining
the benchmarkofefficient international risk sharing in aworld withoutincentive constraints,
which is theempiricallyrejected case of fullrisksharing. It is thenshownthat optimalincentive
compatiblerisksharing canbeunderstoodasconsumption smoothing;2 consequently,incentive
constraints can explain asignificant lack
of
international risksharing. Thisresult is obtainedby requiring that risksharing does not create any additional distortions ineconomic policies
(this circumventsthe modelling of the trade-offbetweenrisksharing and incentives, which is 2There issomeambiguityabout thedefinition ofrisksharing and consumption smoothing in theliterature, which is becauseboth concepts cannot befullyseparated fromeachother. Throughout thistliesis a broad
postponedto Chapter3). Sincetheoptimalcontractgenerally entails some kind
of
distortionin incentives, the welfaregains from consumption smoothing
will
constitutealower bound onthe gains fromincentive compatiblerisksharing.
The methods used inthe previous literature to estimate welfare gains from international
risk sharing have the disadvantage that the results aresensitive to assumptions on countries'
endowment processes and/or whether standard preferences or preferences
that
match assetprice moments are used (Van Wincoop, 1999). For theestimation ofthe welfare gains, a new
method hasbeendeveloped that does notrequire assumptions on the endowment process and
directlyusesassetprice datatocompute the welfare gains. The mainresult fromtheestimation
is that
the welfare gains from consumption smoothing are very large and hence incentiveconstraints(despite being ableto explain the lackofinternationalrisk sharing) do not solve the
internationalrisk sharingpuzzle. Moreover, itisfound thatthe welfare gains fromrisksharing
beyondconsumption smoothingarelikely tobesmall. Theinternationalrisk sharing puzzle is,
therefore, essentiallyaconsumption smoothing puzzle. The results furthermore indicate that
previous studies have substantially understatedthewelfare gains from risk sharing, possibly
dueto inappropriate assumptions on the stochastic process for countries' endowments. The
estimates alsosuggest that for the purposeof measuring welfaregains, standard preferences
satisfactorilyreflect the informationcontainedin asset price moments, andcan, therefore, be
safely used for thecomputation ofwelfaregains.
Chapter 3 and Chapter 4 study the interaction between governmental incentives and in-ternational risk sharing. Chapter 3 examines the implications of portfolio diversification for distortions ineconomic policies, andderivesthe optimaldegreeofdiversification under govern-mentalmoralhazard. It isfirstshownthat portfoliodiversificationcan cause severedistortions in basically all economic policiesthat affect production ofthe domestic firms. However, these
distortions do not afTect the degree of portfolio diversification chosen by individuals (since
they behave atomistically). Hence, individuals
will
diversify as predicted by standard risksharing models. Inthe model, diversification creates negative externalities among (domestic)
households since it increases aggregate distortions (which areborne by all households). The
decentralized equilibrium may, therefore, be inefficient. Then, the analysis considers the
so-cially optimal degreeofdiversification, which balances the welfare gains froma reduction in
risk with
the welfare losses from distortedincentives. For plausible parameters, the optimaldegree of diversification
implies a home bias that is in line with
the existing home bias inindustrial countries. Thissubstantiatestheresultsof Chapter 2 in that incentivesconstraints cannotonlyexplain a lackof international risksharing butalsoimplyanoptimal extent of risk
sharing that is close to the observed one. If a home bias is indeedoptimal, the decentralized
equilibrium willbeinefficient,i.e.,individuals
will
chooseanexcessive degreeof
diversification.While moralhazardcannot directlyexplain the home biasinindividual portfolios (individuals
choose the level ofdiversification regardless
of
incentive problems), it rationalizes the homebias as being the result ofgovernment interference inrisk sharing markets (for example due
1. Introduction 7
equalization orstructuralfunds) in order to avoid excessivediversification.
Chapter 4 studies an application of distorted economic policies due to international risk
sharing. This is done inanenvironment withaseconddistortion,arising from
tax
competition.Tax competitionamonggovernmentsinorder to attract foreigncapital resultsin inefficiently low taxes on capital and counteracts the distortions in incentives stemming from risk
ing.
Thelatter
arise in the modelbecauseinternational diversificationgives governments anincentive to levy excessivecapital taxesbecause a part of the taxburden has to beborne by
foreign investors. In thisframework, distorted incentives duetointernational risk sharing do
not necessarily reduce welfare:
if
taxes are inefficiently low (owing to tax competition), anincreasein governments' incentives to levycapital taxes due toan increaseddiversification is
beneficial for welfare. Next, it is analyzed which ofthese counteracting motives for capital
taxation is likelyto dominate. The results suggest that for boththe currentsituation in the
U.S. and the case offullfinancialintegration, taxes on corporationsareinefficiently high, i.e., the motive for taxation arising from diversification is dominant. Full financial integration is
thereby characterized by theabsence ofanyfrictionsspecific to cross-bordercapitalmovements and completediversification inthesmall country case (i.e., the share of theportfolioinvested
in domestic assets is zero). These results run counter to the commonly expressed concern
that
financial integration, by increasingcapital mobility, lowersgovernments'ability
toraise revenue from capital taxes.Chapter 4 furthermore provides empirical evidence for the importance of international
diversificationfor governments' incentives. A panelanalysis ofU.S. states from 1977 to 1998
showsthatthere isasignificantpositiveand economicallyimportantcausalitybetweenforeign
ownership of capital and profits (which is brought about by diversification) and taxes on
capital. Moreover, the resultssuggest
that
foreign ownershipwill
become amajordeterminantof
capitaltaxesif
diversificationacrosscountriesbecomes morecomplete.Chapter5contains theconclusions to Part I. Themain points areasfollows. Theanalyses
indicate inseveral waysthatincentive constraints provideaconsistentexplanation for the lack
of international risk sharing. However, they donot solvethe international risk sharing
puz-zle. International risksharinghasimportant implications for economic policies, whichfurther
researchshould address more extensively. Increasingrisk sharing beyond current levels may
reducewelfare and is at bestinefficient (as long as it isnotrestricted to consumption
smooth-ing). Further financial integrationand financial innovation may in that view be undesirable
becauseitallowsindividualstobettershare country-specificrisks.
PART II
Chapter6 analyzes thelife cycle of the firm underincentive constraints, andin, particular,the
ability
ofincentive constraints to account for the financerisksharingpuzzle. Asoutlinedabove, entrepreneurs haveamotive forsellingtheir firmsin order to reducetheirexposure to firm-specific risk. Chapter6showsthatincentiveconstraints affect thismotive as
follows. First, when a part of the firm is sold, an entrepreneur's incentives to exerciseeffort
according toherstake in thefirm. Rational investors will beaware oftherelation between an
entrepreneur'sstake in the firm andthe amount
of
effortexercised. Hence, their valuation ofthe firm
will
decreasewiththe stake the entrepreneursells of the firm. This, in turn,willlimittheentrepreneur's benefits from selling the firm; therefore, the entrepreneur may not want to
sell completely.3 This explains the
first part of
the finance risk sharing puzzle ('entrepreneurs keep asignificant stake in the firm'). Asecondconstraint tothe entrepreneur's desire to sell herfirm
arises because an entrepreneur hastypically several opportunities to sell the firm in thecourse of time. Whilethe efficiency losses ofan initial sale ofastake in the firm maybe
fully
internalized bythe entrepreneur (becauseofinvestors being aware ofthe relationship between
the entrepreneur'sstakeandexercisedeffort), subsequent sales bytheentrepreneurreduce her
efTort further and pose a negative externality on the existingshareholders. Therefore, if the
entrepreneurcannot commit herself to keep the stake retained initially, investors
will
reducetheir valuation of the firm further in order to be compensated for the additional efficiency
lossesstemming fromselling the firmat laterstages.
It is then shown that, as a consequenceofthese constraints, entrepreneurs may after the
start-upoftheir firm wait until they take the firm public (and thus postpone the sale of the
firm) and, once being public, stagger the sale of the firm. Thisrationalizes the second part
of the finance puzzle ('entrepreneurs spread the sale of the firm over time'). The analysis
furthermoredemonstratesthat incentive constraintsimplyarealisticlife cycle of the firm that is consistent with the useofventurecapital in the private phase of the firm as well as with a
change in control inthepublicphase (when the firmislisted at anexchange). It isalsoshown
that the sale of the firm by the entrepreneurisinefficient beyond the moral hazard constraint:
the entrepreneur sells too much of the firm and sells too late. This reduces the benefits of
listing a firm at an exchange, and as a result, too few firms go public in
equilibrium. The
chapter concludesby presenting financial regulations
that
removethis inefficiency.3 Itisassumed that due to a lossofentrepreneurial humancapital,theentrepreneur will not findit worthwhile
1. Introduction 9
References
and
Additional
Readings Papersthat
study the effects ofrisk sharing oninvestment are Obstfeld (1994b), Acemoglu and
Zilibotti
(1997), Devereux (1997) and in thefinance strand Stulz (1995). The lackof internationalrisk sharingisdocumented in Stockman
and Tesar (1995) (volatility of output and consumption), Sorensen andYosha (1998) (extent
ofincome insurance across countries), Stulz (1994) andTesarand Werner (1995) (home bias).
The
link
betweenthe home bias and the lackof international risksharingisdiscussed inLewis(1999), pages 604f. Surveys of theliterature onthe welfaregainsfrom internationalrisk sharing
are containedin Tesar (1995), Lewis (1999) and Lewis (2000). A recent study that compares
different approaches for the estimation of welfare gains and discusses the international risk
sharing puzzle is Van Wincoop (1999).
'Itansaction costs asan explanation for the home bias havebeenexaminedin Cooper and
Kaplanis (1986, 1994) and Botazzi et. al. (1996). The higher turnover
of
equity held byforeigners is reported in Tesar and Werner (1995). Better informationabout the risk-return
characteristics
of
domestic stock have been usedinGehrig (1993) and Brennan and Cao (1997)to explain the home bias. Gordon and Bovenberg (1996) and Razin et. al. (1998) consider
lemon problems asareason for the homebias. Cooper andKaplanis (1986), Stulz (1981), and
for developing countries Demirglig-Kunt and Huizinga (1995), study therationale for a home
bias stemming from taxdifferencesbetween foreign and domestic income.
Inflation risk as acause for the home biasisrejected
in
Cooper andKaplanis (1994),whileGordon and Gaspar (2001) present a framework in which the home bias isjustifiedby price
hedging motives. Botazzi et. al. (1996) find that human wealth can
partly
explain the homebias, Baxter andJermann (1998) conclude that the inclusion
of
human wealth into portfolioconsiderations worsens the homebias puzzle. Furtherexplanations for the lackof international
risk sharing or the home biasin portfoliosare presentedinAdlerandDumas (1983) (exchange
raterisk), Kocherlakota (1996b) (commitment problems), Lewis (1999) (mismeasurement of
consumption) and Obstfeld(2000) (costsof internationaltrade). Stulz (1994)andLewis (1999)
surveytheseveral explanations for the home bias.
Optimal contracts under moral hazard and self-enforceability are (among others) studied
by Spear and Srivastave (1987) and Phelan and Townsend (1991). Gertlerand Rogoff (1990)
and Atkeson (1991) provide applications to international capital flows. In Kehoe and Peri
(2000),self-enforceable contracts are used to explainthe momentsofmacroeconomic variables.
The implications ofentrepreneurial moral hazard for domesticrisksharing hasbeenstudied
in Kahn
(1990), Kocherlakota (1998) and Magill andQuinzii (1999, 2001). The fact that
entrepreneurs remainasubstantialstakein their firmevenafterseveralyears of the foundationof the firm isdocumented in Jain and Kini (1994) and Mikkelson et. al. (1997). Evidence that
the listing of the firm takesplace lateiscontainedin Pagano et. al. (1995) andRydqvist and
Hogholm (1995). Mikkelson et. al. (1997) document
that
entrepreneurssell significant stakesChapter 2
The Welfare Gains from Incentive
Compatible International Risk Sharing
2.1 Introduction
The significant lack of internationalrisk sharing,asoutlined intheintroduction to thisthesis,
raised thequestion whetherhouseholds forego sizeable gains from risk sharing. Most studies
find that
thewelfare gainsfrom additional international risk sharingare indeed substantial,1which has led to the 'international risk sharing puzzle'. In this chapter we consider an
ex-planation for this puzzle based on governmental moral hazard and the non-enforceability of international contracts. The main aim of this chapter thereby is to provide an estimate for
the welfare gains from international risk sharing that is restricted through these incentive
constraints.2
We first show that such risk sharing reduces to international consumption smoothing if
one requires thatrisk sharing does not create anyadditionaldistortions. Therefore, incentive
constraints can explain a lackof international risksharing;in particular, it isargued that they
explain the
pattern of
risk sharing among groups ofcountries ofdifferent legal integration.In order to obtain a reliable measure for the welfare gains from international consumption
smoothing,weemploy anewmethod
that
computes actual welfare gains frompastconsumptiondata. Actual welfare gains answer the question of what the gains for countries would have
been hadtheyengaged in risk sharing in the past. Van Wincoop (1999) and Athanasoulis and
Van Wincoop (2000) have shown that welfare gains arevery sensitive to assumptions on the
stochastic processgoverning countries' consumption. This has led toa confusionabout the
true magnitude ofthe welfaregains.3 Computingactual gains has the advantage that it does
1
Surveys are containedinTesar (1995), Van Wincoop (1999 ) and Lewis (2000).
2Wewill focus on the welfare gainsthat arisedirectly from the reductioninconsumptionvariability. If
thesegains are not small enough to explain theinternationalrisk sharing puzzle, thentaking intoaccount the possibleindirectwelfare gains (which are more difficultto quantify) willonlyworsenthe puzzle.
3In particular, Van Wincoop(1999) showsthatdifferent assumptions on the stochasticprocess (eventhough the parameters are estimated from the same data) call lead to welfare gains from less than 0.1% to more than 5% over a 100-year horizon.
not requireanyassumptions on the endowmentprocess.
A
further uncertainty about the sizeof
thewelfare gains intheliteraturestems from the factthatstandard preferences are not able to match asset price data, mostnotably theequity premium. Preferences that areconsistentwith
asset price data have been found to imply much higher welfare gains.4 Therefore, wedevelop an approach that computes welfare gains from market interest rates. The basic idea
of this approach is that due tointertemporaloptimization ofhouseholds, interest rates reflect
ratesof intertemporalsubstitution and can be used tovalue consumption streams.
We find thattheactual welfaregainsfrom consumption smoothingare large. Hence
incen-tiveconstraints cannotresolvetheinternationalrisk sharing puzzle. For the representative G7
country during 1956-1992, the welfare gains are equivalent to a2.8% increase in permanent
consumption, assuming standardpreferences. Market interest rates imply similargains. This F
number is abovethe welfare gains from full risk sharing from studies that assume a specific
consumptionprocess. Van Wincoop (1999), for example, finds welfare gainsbetween 0.5 and
1.5% for a comparable sample and horizon. Tesar (1995) estimates the welfare gains from
introducingaworld bondmarket (and thus fromconsumption smoothing) andfinds that they
are below 0.5% forastationaryprocess. Recently, Athanasoulis and VanWincoop (2000) have
developedatechnique thatissimilar to ours in thatitmeasureswelfaregainswithout assuming
a consumption process. Theyfind welfaregains that are consistent withourestimates. This
suggests thatthe consumptionprocesses used inthe literatureunderestimate the true welfare
gains. We furthermore find
that
consumption smoothing reduces the welfare costs of a unitconsumptionvariability by96.4%, whichindicates thatthe welfaregainsbeyond consumption
smoothingarelimited.
The remainder of thischapter is organized as follows. The nextsection demonstrates the
incentive problems arising from international risk sharing, derives the optimal non-distorting
risk sharing contract andcomparestoexisting risksharing. Section2.3 describesthe
method-ology for the measurement of the welfaregains. Section 2.4presentsthe welfare estimates for
the G7 countries. Thefinalsection summarizes.
2.2
Incentive
Compatible
Risk
Sharing
In order to study the implications
of
incentive constraints for international risk sharing, westart upby defining the benchmark offull internationalrisk sharing. Consideraworld made up
of n (e»ante) identical countries. Eachcountryreceives everyperiodastochastic endowment
yi whosedistributionisknown to allcountries. The representative household'stimeseparable
expected
utility is
Eo[E #u(4)]
t=l4Obstfeld (19948) and Van Wincoop (1994) havedemonstratedtliatusingpreferenceswith habit formation and/or timenon-separability resultsinlarger welfaregains. Moreover, Lewis(2000) shows thatstock niarket
2, Welfare Gains fromIncentive Compatible InternationalRiskSharing 15
Full risksharing requiresthat countriesperfectlypooltheirrisks(Lucas, 1982).
It
follows thateach country consumes a fixed shareof world output cw in each period given by
n
4 - a'cr with cw := E
1/; (2.1)i=1
where the a"sareindependent of therealization ofthe countries' endowments.
Full risksharingmay,however, notbe feasible
if
internationalcontracts arenot enforceable.Countries can then notbe forcedto provide the transfers (y;
-4)
necessarytoimplement (2.1).Furthermore, distortions arising from governmental moral hazard may make full risk sharing
undesirable. Moral hazard arises because acountry's production isnot exogenously given as
considered above but partly depends on governments policies. Toillustrate the moral hazard
problem, assume that a country's output y depends on some policy parameter e chosen by
the government. Thisparameter can, forexample, representinvestment in publicinputgoods
(such asinfrastructure andresearch), productionsubsidies orthelevelofcorporatetaxation.
The domestic costs ofthe policy are given by z(e) and can be interpreted as the welfare
loss for the representative household (expressedin units ofthe output good) caused by a tax
necessary to finance thepolicy e. Under autarchy, domestic consumptionequals production
and the policy e will be chosenoptimally by the government such that y'(e) = z'(e). Under risk sharing,the benefits from additional production do only partlyaccrue tothe country. In particular, under full risk sharing the country's consumption only increases according to its
share inthe worldoutput.
If
countriesaresymmetric, then a = 1/n in (2.1) and only 1/n-thof
every additional unit ofdomestic output remains inthe country. The government's policychoice isthen determined by y'(e)/n =z'(e), which leads toadistorted provision of e. The characterization ofthe optimal contract under moral hazard and enforceability
con-straints in adynamic frameworkiscomplex (see for example Atkeson, 1991) and the optimal allocationgenerally entails some kindof distortion. No modelssuitablefor empirical
applica-tion do yet exist, whichis partly owed to the difficulty ofspeciB,ingthe costs
of
distortions that arise from moral hazard. Inthe following werestrict risk sharing such that it does notcreateanyadditional distortions(compared to autarchy) and show that asimple and
intuitive
optimalincentive compatible contract exists.
For thispurposewe changetheaboveframeworkasfollows. The numberofperiods is now
finite and given by m.5 The countries'output is y; = 11'f (4) +
4,
where e; is chosen by the
government and summarizes all policies
that
affect output. The policye; causes furthermoresocial costs z(ej) to the country (measured in
units of
the output good). The ej 's and thecountry-specific productivity parameters 9, are unobservable. The shocks v; are identically
and independently distributedacross countries with Eo[u;] =0. Generally, more assumptions
onthe distribution of the u;'s are necessary because alsothe predictability of
shocks at t>0
matters for the allocationofconsumption. Consistent withour notion ofwelfaregains from
5 This isconsistent with most ofthe literature, which onlyconsidersrisk sharing over afinite number of
periods. Tlie reason for that is thatthe welfare gains fromrisksharing may notbebounded asthenumber of
periodsincreases (seeVan Wincoop, 1999).
16 RiskSharing underIncentive Constraints
risk sharing (see next section) we assume inthe following that all uncertaintyis immediately
resolvedafterthecontract iswritten. Acontractthereforespecifies anamountofconsumption
c; ineachperiod thatcandepend onthe realizationofcurrent and past output {ys},st and all
shocks {u}.6
The government
will
default on the contract (or allow the domestic household todefault) ifat any time tthe discountedfuturenetpayments toother
countries (Elt #(yi-<))
arelargerthanthe costsofdefaulting.7 The specification ofthe
latter
is problematic. Inthe literatureit hasoftenbeen assumedthatcountriesare forever excludedfrom contracting afteradefault (see, for example, Eaton, 1981). The disadvantage ofthisassumption is that it may not be
time-consistent: after a default there is no incentive to exclude countries from the contract
sincethe gains from risk sharing
will
generally be the larger the larger the numberofcountriesparticipating inrisk sharing. Moreover, it is notempiricallysupported. For example, countries
that
havedefaultedontheirgovernment bondscantypicallyreturn tothecapitalmarketsaftera few years (or do even neverstopborrowinginternationally). Other commonlyused costs are
costs ofdefault that are exogenously given or costs in term of reputation losses (which may
matter forothercontractual relationships). The difiiculty withtheseassumptionswithregard
to obtaining contracts that are useful for empirical work is to actually specify the costs in
quantitative terms.
We assume here that all net payments that have been made to a
country in the past
can be collateralized. In the case ofadefault, these payments can then be recovered. Such
collateralization can, for example, takeplace ondeposits atinternationalorganizations (such as
thespecialdrawingrights at the IMF) but alsoonassetsofcountriesheldabroad,e.g., private
property. The advantage ofthis assumption is that itallowsto achievea time-consistent and
tractablesolution tothecontractingproblem. The disadvantage is, ofcourse, that it assumes
the existence ofamechanismforcollateralization,which requiresacertaincoordinationamong
countries (however, the alternative punishments considered above also require coordination
in order to enforce the punishment). The assumption can furthermore be motivated with reputational arguments: it
rules out that
a country defaults on current payments if it hasreceived equal payments in the past but allows a country to default if, for example, it has
received more favorable endowments than othercountries. We furthermore assume that the
collateralization includes interest (where the interest rate is based on the steady state rate,
i.e., (1 + r)6 - 1). A
country's participation constraint isthen given bym t-1 , i i\ t-1
E 6•(1/1 - c:) S E l; : - 14, =E 8 (cl - v;) for all l S t s m
s=t s==1 /(1 + r)' 3=1
6Tothe extentthatconsumption is not predictable, theoptimalconsumptionallocationex-post (which is
studied here) may not beachievedex-ante.
2. Welfare GainsfromIncentive CompatibleInternationalRisk
Sharing 17
which canbereduced to
63(Yl-
cl) ;
0 (2.2) S=1In order for risk sharing not to create any additional distortions, the government must find
it optimal
to chose the same e;'s as under autarchy. The autarchypolicy el fulfills y'(e ) =
gif'(e t) - Z'(€ 1).
Thepolicy choicegiventhecontract c; =4({14}.g, {vi}) is
m
4 = arg illax E 6-4(vi(e;)) -
8*z(4) (2.3)4 S=t
Thenecessary condition for ej = e;t is
fe
ads--2
(2.4)1=t Dy:(4)
which follows from thefirst ordercondition for (2.3) after imposing ' e'A - Z'(e ).
Consider nowtheallocationofconsumption independentofcountries'output. The
incentive-constraint (2.4) is thennot binding andtheoptimalallocation has to
fulfill
0,6tu'(4) - At - Aie =
0, (2.5)where A isthecountry's weight in thesocialplanner function and At and A, are the
Lagrange-multipliers associated withtheresourceconstraint
n 71
Ed=
EY:
(2.6)i=1 i=1
and participation constraint (2.2), respectively. Since the participation constraint (2.2) is
identical for all countries, we have Ai = M fortwo countries i and j and
it
follows from (2.5) that11'(4)/11'(4+k) = 11'(4)/u'(Ci+k, (2.7)
Hencethe intertemporal ratio
of
marginal utilitiesisequalizedacross countries.Proposition 2.1
A contmctfulfilling
(2.2), (2.4), (2.6) and (2.71 is given by4 = atic:7 + 1,4 - vt - (U; -1, ) (2.8)
with cw := Eli y(e;) and ai := (E;11 bt('lif(el) + vi))/(E;-1 8tc7).
Proof. (2.4): a) el = ek: Eli 6'cl = E;116'(aic:'ty;-1/1-(vi-vi)) = Eli
bev . Differ-entiating wrt. to14gives 6. b) 4 = 4 for t > 1: E:1: 83(1 = Elt 63(aicr+Yj -yl- (U; -vii)),
differentiating wrt. toJjgives bt. (2.2): see (2.4) a).(2.6): Ili ai =
1,hence Eli 4 - cr =
El,1 y(e;). (2.7): For u'(c)homogenous,
u'(4)/u'(4+k) = 11'(aicr)'11'(a,cr+k) = u'(cr)/u'(cr+k),
whichisindependent of i. •
and depend on the realization ofthe shocks. This is in
contrast to full
risk sharing, wherea depends on apected permanent income. Since the incentive compatible contract requires
E11 8'cl = 211 6314 (2.2),
noe»postredistributiontakes place. However,according to (2.7) anoptimalcontract equalizesthe intertemporal ratiosofmarginal utilities. The welfaregainsfrom incentive compatiblerisk sharing arisetherefore solely due tointernationalconsumption
smoothing.
How does theoptimal contract compare to an allocation
that
emergesif
households cantradeariskless international bond? Fora country that haszero
initial
bond holdings and zerobond holdings after thelast period we have
m-1
(722 - c;) (1 -t- Ta) t ivt, - cl, =
O (2.9)t=l S=t
wherert denotesthe interest rate ona one-yearbond. Assumingas above (1 + rt)6 = 1 gives
E eM- 4) =
0 (2.10)t=1
which is identical to participation constraint (2.2). If we also continue to assume that all
uncertaintyresolves inthefirst period, we get fromthe Euler-equations
k-1
11'(4)/Et[11'(4+k)1 = 11,(4)/tt,(4+k) = 61' II(1 +ra) (2.11)
S=t
and the intertemporal ratios ofmarginal
utility
are again equalized acrosscountries. Hence,the optimal contract under incentive constraints implies the same allocation as in a bond
economy (incentive compatible welfare gains can, for example, alsobe interpreted as thegains
from
full
capital market integration). This result also gives support to the specification of the participation constraint: international borrowing and lending has so far beenenforceable among G7 countries(on which we basetheestimation inSection 2.4). If at all, theparticipationconstraintcanthereforeonlyunderstate the costsofdefault.8
Is the current lack of international risk sharing a result
of
incentive constraints? If so,risk sharing amongjurisdictions that face less incentive problems should be more complete.
Sorensen and Yosha (1998) find that 66% of all incomeshocks areshared among U.S. states
compared to only25%among the OECD countries. WhileU.S. states and the OECD countries
maybecharacterized bysimilardegreesoffinancial marketintegration,incentive problems are
limited within theU.S. First,federal lawensuresthe enforcementof inter-statecontracts.
Sec-ond, moral hazard is restrictedto policies onthestatelevel and canbe easierdealt with due
tothe common legal system. Hence, the relative lack of risk sharing among OECD countries
compared to U.S. statesis consistent
with
incentiveconstraints. Moreover, sinceProposition 8An understatement of thecostsofdefault leads to more restrictions oninternational risksharing(because fewer contracts canbeenforced) and possibly lower welfare gains. This possiblebias (which may also arise2. Welfare Gains fromIncentive Compatible International Risk
Sharing 19
1 shows
that
consumption smoothing is feasible in the presenceof incentive problems,con-sumptionsmoothing should be used moredominantlyamong the OECD countriesthanamong
U.S. states ifincentive constraints are important. Quite strikingly, S0rensenand Yosha find
that within the OECD 80% ofrisk sharing takesplace due to consumption smoothing, while
among the U.S. states consumption smoothing only accounts for 11% ofrisk sharing.' The
evidencetherefore indicates a role for incentive constraints inexplaining the current degree of
risk sharing. If thewelfare gains from consumption smoothing beyond the current degree of
risk sharingare moreover small, thiswould suggest
that
incentive problems can also accountfortheinternational risk sharingpuzzle.
2.3
Measuring
Welfare
Gains
from
Consumption
Smoothing
Since current consumption {ct}
will
already display some degree ofrisksharing (i.e., it will
deviate from {Vt}), the gainsfrom consumption smoothingaregiven by the welfare gains for
a representative consumer that moves from current consumption {ct} to consumption under
optimal smoothing {c;}.
In
expectedterms, thesegains aredefined bym m
0 - Eo[E 6:u(ci) - E btu(c:(1 +
gain))] (2.12)t=1 t=1
wherethe welfaregains (gain) are expressed in termsofpermanent consumption. The
domi-nant approach intheliterature has beento proceedby specifyingastochasticprocess for {ct }
and {c;} (with
{c;}
typicallybeing a share ofworld consumption) and thento compute (orsimulate) the ezpected welfare gains according to (2.12). On the contrary, we
will
compute actual welfare gains defined bym
6tu(c;) - E eu(c:(1 +
gain)) (2.13)t=1 t=1
where {ct} and {4} are
past consumption dataseries. Actual gains areidentical to expected gains intheabsenceof
uncertainty (i.e., Eo[ct] = ct and Eo[4] = ci). Theyanswerthe followingquestion: What would thewelfare gain for acountry have been had it made aconsumption
smoothing contract
promising {4}
inreturn for {ct} at t= 0?
The advantage ofmeasuringactual gains is that it does not require assumptions on the underlyingstochastic process for
{c;} and {ct}. Its
shortcoming is that it doesnotallow theterms ofthecontract (the a's) todepend on expected differences between countries. However, the
latter is for
our purpose oflimited relevance. Since
with
consumptionsmoothing E -1 6tc; = I;li 8tct (2.6)
and hence there isno ex-postredistribution,expected differencesonlymatter tothe extentthatcountriesgThe extensive use consumption smoothing among OECD countries (which isinabsolute termsevenlarger
have a different
volatility
of consumption (i.e., a country with more volatile consumptionshould get, ceteris paribus, alower sharea). Moreover, expected differences willonly distort
country-specificgains but notthe welfare gains forarepresentativecountry.
We define consumption under optimal smoothing {c'} analogous to Proposition 1 by
c; - ac .10 Equation (2.13) gives then the total welfare gains from additional
consump-tion smoothing. Besides, we arealso interestedin computingthe welfare gains from marginal
increases in consumption smoothing. Consider, for example, a householdthat exchanges
do-mestic consumption {ct} for the consumption
stream {Et} = (1 - q){c} +
q{c-},
where q isthe degree of (additional) consumption smoothing. The welfaregains from moving from {c}
to {E}
areanalogous to (2.13) defined bym m
0 = Z btu((1 - q)4 + q(<) - E Btu(ce(1 +
gain(q))) (2.14)t=1 t-1
Differentiating (2.14) with respect to q and solving for again/aq gives the maminalgainsfrom
a unit of consumption smoothing at q
Ogain/Oq = EZ,1 6tu'(Et)(c; - ct) (2.15)
E;116tu'(Ct)Ct
For q=0 and q=1, equation (2.15) becomes
Ogain/Oq(0) - E;-16tu'(Ct)c; -
1 (2.16) E;11 Ottz'(Ct)CtE-m #u'(ci)(c; - ct)
again/aq(1) =
A-t=1 (2.17) E;11 #11'(Ct)CtWhat is therelationbetweenagain/Bq(0), again/aq(1) and gain? If gain > 0 thenan increase
in q
reduces consumption variability, which, in turn, reduces the welfare gains from a unitreductioninconsumption variability (forahousehold
with
completely smoothed consumptionthe change in welfare due to a small variation in consumption variability is zero). Hence,
the marginal gainsevaluated at autarchy consumption (Ogain/Bq(0))
will
overstate thetotal
welfare gains, while the marginal gains evaluated at {c;} (again/aq(1)) will understate the
total
gains. On thecontrary, if the move from {c} to{c*}
increasesconsumption variability (and hence gain < 0) then Ogain/Dq(0)will
understate and Ogain/Oq(1)will
overstate thetotalwelfaregains.
The
Welfare Gains with CRRA-Utility. Foru(ct) =
4-7/(1-7),
equations (2.13),(2.16)and (2.17) canbesolvedexplicitly forthe welfaregains
gain = (2518£4*1-7)1/(1-7)-
1 (2.18) Et,11 btc:1-7 r-m rt 1-7 * Ogain/Bq(0) = 6--1 0 7 1_-1
(2.19) E;116 ct Ogain/Dq(1) = E:11 6241-7(4 -ct) (2.20)E;11 04-7
2. Welfare GainsfromIncentiveCompatible International Risk Sharing 21
The Welfare
GainsImplied by Intertemporal Optimization.
TheEuler equationfor the representative household beforeadditionalconsumption smoothing is
u'(ct) = 6(1 +
rt)Et[u'(ct+i)]
(2.21)where rt is the domestic one-year interest rate at t.
Approximating u'(ct+1) by u'(ct-1) t
u"(ct-1)(c,+1 - 4-1) in
(2.21) givesu'(ct) = 6(1 + rt)(1 -
7t-1E:[ct+11 - Ct-1)11'(Ct-1) (2.22)Ct-1
where 7t = -11"(Ct) ' ct/u' (ct) is the coefficientofrelativeriskaversion. Subsequent substitution
of lt'(Ct-1) in (2.22) gives forconstant 711
Es[c,+11 - c,-1
11'(Ct) = e 11(1 + rs)(1
i
)u'(Co) (2.23)S=1 Cs-1
Using (2.23) to
substitute for it'(ct) in
theequation for the marginal welfare gainsevaluatedat autarchy (2.16)yields
again/Dq(0) = 221 62t(1111(1 + rs)(1 - 7E,Ic.:.1lic.-1)4) -
1(2.24)
Em-1 62t(f[11(1 + Ta)(1 - 7E.Ic.td-ic.-1)Ct)
With
rationalexpectations, the expectation errorEs[c,+11 - C,+1 isindependent ofallvariablesknown at s, hence one can obtain an unbiased estimator
of
nominator and denominator in(2.24) by setting E, [c,+11 - CS+F
An
estimator for Dgain/Oq(0) is given by
; 62t(Ill= (1 + rs)(l- 744.3-1)4)
Dgain/Oq(0) = -1 (2.25)
E;11 52t (Hi-1 (1 + rs)(1 - 7 c'+2-1 )Ct
2.4
Results
We computethe actualwelfaregainsfrom consumption smoothing for the G7 countries(United
States, Germany, Canada, France, UnitedKingdom,
Italy
and Japan) during1956-1992. Themain ingredients for the computation arethe aggregate (real) consumption data series {c'},
which are
taken from the
Penn World Table. World consumption {cw} is the per capitaconsumption of the G7 countries (where all countries are weighted equally). Estimation of
the welfare gains implied byEulerequationsfurthermorerequires realinterest rate data. We
chose, whenever available, market rates onone-yeartreasurybills,otherwiseshortermaturities
(fromOECDMacroeconomicIndicatorsand Datastream). Real interest rates areobtained by
deflating
with
CPI-inllation. We furthermore assume a discountfactor of 6 = 0.95 and a
degree
of
relativerisk aversion 7 -3. Table2.1 containsthe resultsforequations (2.18)-(2.20)(CRRA-utility) and
(2.25) (Euler-equations).11 Fortheconsequences ofthisassumption for theinterpretation ofthe welfare gains implied by interest rates,
Table 2.1: The actual welfaregainsfrominternationalconsumption smoothing CRRA-utility Euler-equations
Marginal gains Marginalgains Marginalgains
Total gains underdomestic under world underdomestic
consumption {C} consumption
{Cl
consumption {C}United States -5,5% -4,9% -7,3% -6,5% Germany 1,5% 1,7% 1,2% 1,0% Canada -2,0% -1,6% -2,5% -1,6% France 0,8% 0,9% 0,7% 2,1% United Kingdom -6,2% -5,0% -9,0% -7,2% Italy 12,8% 14,6% 7,6% 19,3% Japan 28,9% 36,5% 10,2% 43,9% Averagegains 2,8% 4,2% 0,15% 5,0%
Notes: Welfare gains are expressed asper:entageincreaseinpennanentconsumption for G7-countries during 1956-1992, average gains are computed from consumption-weighted country gains
The
total
welfare gains (first column in Table 2.1) differ widely across countries, from -6.2% for the U.K. to 28.9%for Japan. Thisdifferenceissurprisinglylarge given thatthere is no redistribution (due to the participationconstraint 2.2). The heterogeneity across coun-tries therefore arises solely due to households' risk aversion in thepresence ofdifferences incountries' consumption
volatility
before consumptionsmoothing (for 7=0
the welfaregainsare zero for eachcountry).12 For example, the standard deviation
of
consumption divided bymean consumption for Japan is60%higher than for world consumption, while forthe
'loser'-countries U.S. and U.K. domestic consumption is less volatile than world consumption. The
distribution ofthe welfaregains among countries will change if the contract would take into
account expected differences in countries' consumption variability (for example ifcountries
would bargain over theterms ofthe
contract at t = 0)
and should therefore be interpretedwith caution.
It
neverthelessshows that the incentives tosmooth consumption can be quitedifferent acrosscountries.
We compute the welfare gains for a (fictitious) representative country from the average
ofthe country-specific welfare gains. The gains are 2.8%
if
countries areweighted accordingto their share in permanent world consumption (a). How does this number compare to
estimates for fullrisksharing from theliterature? Studies
that
assume astationaryprocess forconsumption typically findwelfare gains of less than 0.5% (for example Obstfeld, 1994a and
Tesar, 1995). Tesar (1995)alsoestimatesthe welfare gains fromrisk sharing thatislimited to
trade in an international bond (andhence there is only consumption smoothing). Assuming
a stationary process, she finds that the welfare gains can be much smaller than for full risk
sharing. Estimates obtainedfromotherendowmentprocesses arehigher. Van Wincoop (1999),
for example, estimates foracomparable sample and horizon gains of 0.5% fora random walk
process and of 1.5% foraconsumption process that has autocorrelatedgrowth rates. Yet, this
numbersare
still
substantiallybelow theactual gainsfromconsumption smoothing. Since the12Negativewelfare gains do not contradicttheparticipationconstraint (2.2) because we have assumedthat