Download This PDF is a selection from a published volume from... Economic Research Volume Title: NBER International Seminar on Macroeconomics 2007

This PDF is a selection from a published volume from the National Bureau of
Economic Research
Volume Title: NBER International Seminar on Macroeconomics 2007
Volume Author/Editor: Richard Clarida and Francesco Giavazzi, organizers
Volume Publisher: University of Chicago Press
ISSN: 1932-8796
Volume URL: http://www.nber.org/books/clar07-1
Conference Date: June 15-16, 2007
Publication Date: January 2009
Chapter Title: On the Welfare Implications of Financial Globalization without
Financial Development
Chapter Author: Enrique G. Mendoza, Vincenzo Quadrini, José-Víctor Ríos-Rull
Chapter URL: http://www.nber.org/chapters/c3012
Chapter pages in book: (283 - 312)
6
On the Welfare Implications of Financial
Globalization without Financial Development
EnriqueG. Mendoza, University of Maryland,IMF,and NBER
Vincenzo Quadrini, University of SouthernCalifornia,CEPR,and NBER
Jose-VictorRios-Rull,, UniversityofPennsylvania,CAERP,CEPR,and
NBER
".. . muchoftheMexican
byforeigners However,
banking
systemis nowcontrolled
Feesremain
hasnotledtocheaper
thisinfluxofforeigncapital
high,producing
banking.
inMexover30percent Banking
interest
ratesaverage
Credit-card
bumper
profits.
icoremains
expensive
for thecountry's
for banks,andintimidatingly
veryprofitable
poor."
TheEconomist,
Nov.25,2006
6.1 Introduction
The questionof whether financialglobalizationis beneficialfor the participatingcountriesis the subjectof a profound and divisive debate in
academicand policy circles.Criticslike Stiglitz (2002),Soros (2002),and
Bhagwati(2004)offerominous assessmentsarguingthat financialglobalizationat best has failed to produce the largebenefitsit promised and
at worst is a flawed policy leading to economic collapses. On the other
hand, supporterslike Mishkin(2006),Rajanand Zingales(2003),Frankel
(2007),and Obstfeldand Taylor(2004),make a strongcase in defense of
financialglobalization.Theircase is founded not just on traditionaltheoreticalargumentshighlightingthe gains fromglobalasset trading(international risk sharing,efficient reallocationof capital,etc.). They argue
that the development of domestic financial systems- and the social,
political, and economic institutions that anchor them is a necessary
conditionfor countriesto reapthe potentialbenefitsfromfinancialglobalization.
This argumentraises key unansweredquestions:Whatare the implications of going forwardwith financialglobalizationif domestic financial marketsremainunderdevelopedin some countries?Does this neutralizethe gains from globalizationwithout furtherimplicationsor are
there adverse consequences?In the case of adverse consequences, are
284
Mendoza,Quadrini,and Rios-Rull
those limited to the riskof financialcrashesor sudden stops, or arethere
systemicnegative effects?How largeare these effectsand can economic
policy mitigate them?
This paper aims to answer these questions by studying the effects of
financialintegrationamongs countries that differ in the degree of domestic financialdevelopment. We formalize cross-countrydifferences
in financialdevelopmentthroughthe tightnessof borrowingconstraints
in a multicountry,generalequilibriummodel with heterogenousagents
and incompleteasset markets.Weshow that,if financialglobalizationis
not accompaniedby domestic financialdevelopment,liberalizationcan
have sizable consequences on the distributionof wealth and adverse
welfare effects on some of the participatingcountries.In particular,we
show that, even though liberalizationleads to a significantincreasein
wealth inequality in the most financiallydeveloped country,the aggregatewelfare consequencesare still positive for this country.By contrast,
in the countrywith less developed financialmarkets,the aggregatewelfare consequencesare negative even though the distributionof wealth
does not change much. The welfare effects we estimatedwarf the small
gains from internationalasset trading in open economy real business
cycle models (see, for example, Backus,Kehoe, and Kydland 1992;and
Mendoza 1991), and they are comparableto existing measures of the
welfaregains obtainedby removingdistortionson capitalaccumulation
in open economies (e.g., Gourinchasand Jeanne 2006; Mendoza and
Tesar1998;and Quadrini 2005). Moreover,these adverse effects arise
with the gradualbuildup of large global financialimbalancesbut without the occurrenceof financialcrises.
The analysisconductedin this paperis motivatedby some of the findings from our previous work (see Mendoza, Quadrini,and Rios-Rull
2007,MQRhenceforth).Although the primarygoal of that study was to
investigate the emergence of global imbalances,we also found that financialintegrationcould resultin aggregatewelfarecosts for less financially developed countries.This suggestion was based on the quantitative predictionsof a multicountrymodel where agentsfacenoninsurable
idiosyncraticshocks to endowments and investments, and market incompleteness derives from the limited enforcementof credit contracts.
Thismodel was able to explaintwo key featuresof the globalimbalances
that startedto emerge in the early 1980swith the gradualprocess of financialglobalization:(a)a seculardeclinein the net foreignassetposition
of the United States,which reached-8 percentof world gross domestic
product (GDP)in 2006,and (b) a shift in the compositionof the U.S. ex-
On the Welfare Implications of Globalization without Development
285
ternal asset portfolio featuring a large negative position on risk-free
bonds and a positive position in riskyassets.
In this paper,we use a model with some of the characteristicsof the
MQRsetup but with modificationsthatsharpenthe focus of the analysis
on the consequences of financial integration for the distribution of
wealth within each countryand for the socialwelfare acrossindividuals
and nations. On one hand, we simplify the MQRmodel by abstracting
from shocks to investments and consider only idiosyncraticshocks to
earnings. Investmentshocks are importantfor capturing the portfolio
compositionof foreignasset holdings- which was one of the foci of our
previous paper- but they are not crucialfor the welfare implicationsof
capitalmarketsliberalization.We furthersimplify the model by assuming the absenceof state-contingentclaims.Cross-countrydifferencesin
financialmarketsderive from exogenous differencesin borrowingconstraints.On the other hand, we extend the MQRsetup by introducing
the accumulationof physicalcapital.Thisallows us to combinethe analysis of the distributionaleffects of globalization with the traditional
analysis of efficiencygains from capital reallocationinduced by financial integration.
Thepaperis organizedas follows. Section6.2 describesthe model and
defines equilibriaunder financialautarkyand under global financialintegration.Section6.2.2provides an intuitive characterizationof the two
equilibriaand the implicationsof moving from financialautarkyto financialglobalization.Section6.3 presents the quantitativeresults,with
particularfocus on the model's normativeimplications.Section6.4 conducts a sensitivity analysis. Section 6.5 examines some policy implications, and section 6.6 concludes.
6.2 The Model
Consider a world economy composed of I countries, indexed by i, all
with identical characteristicsexcept for the level of domestic financial
development (or the deepness of domestic financialmarkets).Financial
development is capturedby a parametera1as specified following. In the
quantitativeexercise conducted in the next section, we also allow for
heterogeneityin populationand productivityto matchthe relativesizes
of countriesin the model and in the realworld. However, differencesin
population and productivityact only as rescalingfactorsin the derivation of the equilibriumconditions with capital mobility and they are
not relevantfor the qualitativetheory of the paper.Therefore,through-
286
Mendoza,Quadrini,and Rios-Rull
out this section, we present the model assuming that countries differ
only in a\
Each country is populated by a continuum of agents of total mass 1,
and each agent maximizes expected lifetime utility E Z~=0PfLZ(cf
), where
ctis consumption at time t and |3is the intertemporal discount factor. The
utility function is strictly increasing and concave, with U(0) = -oo and
U"(c) > 0.
Agents are endowed with efficiency units of labor zt that they supply
inelastically to the domestic labor market for the competitive wage wr
The efficiency units of labor change stochastically according to a discrete
Markov process that is independent across agents (i.e., the shock is idiosyncratic). There are no aggregate shocks, and therefore, the issue of
cross-country risk-sharing, like in Clarida (1990), is not an issue here.
Each agent has the ability to operate the production technology yt =
A($t\~*y, where ktis the input of capital, lt is the efficiency units of labor
employed in production (i.e., an individual agent's labor demand), and
A is the Total Factor Productivity that is constant. We assume that 0 <
v < 1 so that there are decreasing returns to scale in production. Capital
depreciates at rate 8. For analytical convenience we define F(kt, lt) =
which represents the sum of output plus the capital
(1 - b)kt+ A(fcfZJ"e)v,
stock net of depreciation.
The assumptions that the production technology is individually operated and displays decreasing returns are not standard in heterogenousagents models with idiosyncratic risks. These assumptions allow us to
distinguish the portfolio choice of an individual agent between physical
capital and bonds. However, because there is no uncertainty in production, the aggregate properties of the model are similar to those of a standard model with an aggregate constant-returns-to-scale production
function.
We also assume that there is a cost in changing the stock of capital. The
role of this adjustment cost is to make the cross-country reallocation of
physical capital sluggish.1 This reflects the fact that financial capital is
more mobile than physical capital. However, as we will show in the
sensitivity analysis, the adjustment cost is not important for the welfare
results of the paper. The adjustment cost takes the form y(Kt, kt+1)=
(|)• (kt+1/Kt- 1)2,where kt+1is the individually-chosen input of capital for
the next period and Ktis the aggregate stock of capital in the current period. The assumption that the individual adjustment cost depends on
the aggregatecapital stock at t, instead of the individualcapital at t, makes
the formulation of the agent's problem simpler because we do not need
On the WelfareImplicationsof Globalizationwithout Development
287
to add kt as an individual state variable. However, we will show in the
sensitivity analysis that the use of the more standard adjustment cost,
(p(fct,
kt+1)- <|>• (kt+1/kt- 1)2,leads to very similar results.
In addition to capital, agents can trade non-state-contingent assets or
bonds, bt+vThe market interest rate on these bonds is rj.Define at as the
end-of-period net worth before consumption. The budget constraint is:
+
a, = c( + <p(K(/fc(+1)
fc(+1+-%lr.
(1)
Net worth evolves according to:
aM = et+1wt+1+ F(kt+VlM) lMwt+1+ bt+v
(2)
Thus, net worth is the sum of labor income, the value of operating the
production technology, net of wage payments, and bonds.
The degree of domestic financial development is captured by a limited liability constraint requiring net worth not to be smaller than a minthat is,
imum value aim,
(3)
«(+1>tf.
This constraint imposes an exogenous borrowing limit: the lower the
value of a1,the higher the agents' ability to borrow.2 The lower bound a{
is the only exogenousdifference among countries.
6.2.1
Optimization Problem and Equilibrium
be a (deterministic) sequence of prices in country i. A
Let {r;, w>i+1}"w
single agent's optimization problem can be written as:
Vfo, at) =
max
\U(ct)
ct'bt+l'kt+Vlt+l L
+ 0 ^X^e^,
e,+1
at+1)g(^ e»+i)|
(4)
J
subject to (1), (2), (3).
Notice that this is the optimization problem for any deterministic sequence of prices, not only steady states. This motivates the time subscript in the value function.
The solution to the agent's problem provides the decision rules for
consumption, cj(e,a), bonds bj+1(e,a), productive assets, kj+1(e,a), and labor / j+1(e,a). These rules determine the evolution of the distribution of
agents over e and a, which we denote by Mj(e, a). The definition of equilibria with and without international mobility of capital are as follows:
288
Mendoza, Quadrini, and Rios-Rull
1 (Autarky). Given the level of financial development, a\ initial
a general
aggregate capital, K\, and distributions, M;(e, a), for ie [!,...,!},
without
is
equilibrium
mobility of capital defined by sequences of: (a) agents'
policies {c;(e, a), fc;+1(e, a), fc;+1(e, a), Z;+1(e, a)};=t; (b) value functions
(e) distribuW\{E, a))%t; (c) prices [r\, w\+£=t; (d) aggregate capital {K;+1}T%;
tions {M^(e,fl)}7=f+rSuch that: (i) the agents' policies solve problem (4); (ii) the
markets for assets clear, fji+1(e, a)M;(e, a) = K;+1, JJ?T+1(e, a)M;(e, a) = 0,
and ie {!,..., 1); (in) the labor market clears J^+^e, a)M!T(e,a)
for alli>t
=
alli>t
and ie {!,..., I); (iv) the sequence of distributions is consis\,for
tent with the initial distribution, the individual policies, and the idiosyncratic
shocks.
Definition
2 (Financial
Given the level of financial develintegration).
initial
and
a1,
distributions, M|(e, a), for i e
opment,
aggregate capital, K\,
a
with
{1, ...,/},
general equilibrium
mobility of capital is defined by sequences
of: (a) agents' policies {c;(e, a), fc;+1(e,a), fc;+1(e,a), Z;+1(e,a)};=t; (b) value functions {V;(e, a)};=t; (c) prices \r\, w\+l)%t; (d) aggregate capital {X;+1}*=f;(e) distributions {M^(e, a)}*=t+1.Such that: (i) agents' policies solve problem (4); (ii)
the asset markets clear, /e^+1(e, a)M!T(e,a) = K^for all i>t,ie
{1, ...,/},
and Ejg^+ife' fl)^(e/ a) ~ 0//or cM 7 - t, and interest rates are equalized
across countries, r\ - rTfor alli^t
and ie {1, . . . , I); (Hi) the labor markets
=
clear /e>fl^+1(e,
allT>t
and ie [1, ...,/}; (iv) the sequence
a)Mir(z, a)
\,for
distributions
is
consistent
with
the
initial
distribution, individual policies,
of
and the idiosyncratic shocks.
Definition
The only differencein the definitionof the two equilibriais that with
financialintegrationthere is a global marketfor bonds, and therefore,
interestrates are equalized acrosscountries(condition[ii]).As a result,
countriesmay have nonzero foreign asset positions.
6.2.2 Characterizationof the Equilibrium
This section characterizesthe properties of the equilibriumwith and
without internationalfinancialintegration.Toshow the importanceof financialdevelopment,we also consideran alternativeeconomywhere asset marketsare complete.This alternativeeconomy would featuretrade
in a full set of claimsthatarecontingenton the realizationof the idiosyncraticshock.Firstwe look at the autarkyregimeand then the regimewith
financialintegration.Inboth caseswe will contrastthe completemarkets
setup with the environmentwhere marketsare incomplete.
Consider first the economy with complete markets in financial au-
On the WelfareImplicationsof Globalizationwithout Development
289
tarky. In this case, agents choose contingent claims b(et+1),physical capital kt+vand labor lt+v
The first-order conditions are:
U'(ct) = (1 + r()p{l7'[c(e(+1)]+ Me(+1)},
"'^
=
u^'m
Ve(+1
(5)
PE(t/'[c(e(+1)]+ Met+1)}
(6)
WM>h+l) = *>t+l
(7)
where X(£,+1)*s *e Lagrange multiplier associated with the limited liability constraint (3).
The first condition holds for any realization of et+1,which implies that
next period consumption, c(e,+1),must be the same for all et+1(full inlt+1)/[l +
surance). The first two conditions imply together that Fk(kt+1,
=
the
+
from
return
the
1
that
is,
productive asrt,
marginal
kt+1)]
q>k(Kt,
rate.
interest
to
the
is
set (net of the adjustment cost) equal
Together with
condition (7), this implies that all agents choose the same inputs of capital and labor.
The following lemma establishes that the condition p(l + r) = 1 holds
in the steady state equilibrium with complete markets.
Lemma 1. Consider the autarky regime and assume that there are complete
markets.Then the steady state interest rate satisfies r = 1/p - 1.
Proof 1. If P(l + r) = 1 is not satisfied, condition (5) implies that the
consumption growth of all agents will be either positive or negative.
This cannot be a steady state because aggregate consumption is not constant.
Q.E.D
Consider now the economy with incomplete markets, where only
nonstate contingent claims are traded. The first-order conditions are:
U'{ct) = (5(1 + rt) E[U'[c(et+1)]+ X(el+1)»
(8)
U'M= Jft+^\
(9)
F,(^W*+i)
= «W
+x(£<-»
PE«U'
W^iM
(10)
=
In this case we obtain again that Fk(kt+1,
lt+1)/[l + %(Kt,kt+1)] l + rt,
and the allocations of production inputs are the same for all agents. Individual consumption, however, is not constant but depends on the
290
Mendoza,Quadrini,and Rios-Rull
realization of the efficiency units of labor. This is akin to the typical
Bewley (1986)economy with uninsurablerisks and aggregateproduction. Because all agents use the same production inputs, they get the
same investmentincome.As it is known fromthe savings literature(see
Huggett 1993;Aiyagari1994;and Carroll1997),the uninsurabilityof the
idiosyncratic risk generates precautionarysavings and in the steady
state P(l + r) < 1.
Lemma 2. Considerthe autarky regimeand assume that only noncontingent
bondsare traded.Then the steady state interest rate satisfies rt < 1/(3 - 1.
Proof 2. Suppose that (3(1+ rt)> 1. BecauseU'(.) is convex, condition
(8) implies that for all agents the expected next period consumptionis
bigger thancurrentconsumption.Therefore,next period aggregateconsumption will also be bigger than today's consumption,which cannot
be a steady state equilibrium.Therefore,rt< 1/ p - 1.
Q.E.D.
Lemmas 1 and 2 establish that in autarkythe economy with incomplete marketshas a lower interestrate than the complete-marketseconomy, at least in the steady state. Suppose now that an economy with
complete markets (country 1) becomes financially integrated with an
incomplete-marketseconomy (country 2). The following proposition
characterizesthe steady state equilibriumwith capitalmobility.
Proposition 1. Suppose that country 1 has completemarketswhile in country 2 marketsare incomplete.In the steady state equilibriumwithfinancial integrationr < 1/p - 1 and country 1 accumulatesa negative netforeign asset
position.
Proof 1. Appendix A.
The case of financialintegrationbetween a complete-marketseconomy and an incomplete-marketseconomy allowed us to establishanalyticalresults.Fromthese resultswe can inferthe propertiesof the equilibriumwhen marketsare incomplete in both countriesand a1< a2.In
general,a highervalue of a leads to higher savings and thus reducesthe
equilibrium interest rate in the autarky regime. Figure 6.1 shows the
equilibriumof the model underautarkyand underfinancialintegration.
The figure plots the aggregate supply of savings in each country as an
increasing,concave function of r? Country 1 has deeper financialmar-
On the Welfare Implications of Globalization without Development
291
Figure 6.1
Steady state equilibria with heterogeneous financial conditions
kets (a1< a2),and hence lower supply of savings for each interestrate.
Becausethe demand for productivecapital,which is inverselyrelatedto
the interestrate,is the same in the two countries,country 1 must have a
higher autarkyinterestratethan country2; that is, r1> r2.
When the countriesbecome financiallyintegrated,the interest rates
are equalized. Comparedto autarky,the interestrate and the supply of
savings fall in country 1 and rise in country 2, and hence the country
with deeper financial markets ends up with a negative foreign asset
position. Moreover,the capitalstock rises relativeto its autarkylevel in
country 1 and it falls in country 2. Thus, financialglobalization leads
capital to flow from the less financiallydeveloped country to the more
developed country. Interestingly,this relocation of capital is akin to
those driven by country-specificimprovements in productivity or removal of distortions on capital accumulation,but here this is a byproductof global financialintegrationamongs countriesthat are identical except for a1< a2.
6.3 QuantitativeAnalysis
In this section, we study the quantitativeproperties of the model in a
two-countryversion calibratedto actual data. The first country is calibratedto representthe United Statesand the second the aggregateof all
otherOrganizationforEconomicCooperationand Development(OECD)
countriesplus emerging economies.4
The two countriesare assumed to differin populationand productivity so that we can match the actualpopulation size and per capitaGDP
292
Mendoza, Quadrini, and Rios-Rull
observed in the data (using the WorldBank'sWorldDevelopment
Indicain
1
The
mass
of
is
set
to
which
is
the
actual
0.064,
tors).
agents country
U.S. populationsharein the entireOECDand emergingeconomies.The
mass of agents in country2 is then set to 0.936.The productivityparameters A1and A2are chosen so that the per capita GDP of the two countries in the steady state with capital mobility match the relative per
capita GDP of the United Statesvis-a-vis the aggregateof other OECD
and emerging economies. In 2005 the per capita GDP of this group of
countries,compared to the United States,was 0.152.In the model, the
relativeper capitaGDPof the two countriescanbe expressedas y1/y2 =
Hence, normalizingA1 = 1, the model matches the rela(A1M2)1/(1"ev).
tive per capitaGDP of the countriesin the databy settingA2= 0.1521~8v.
After setting v = 0.9 and 0 = 0.289, as specified below, we get A2 =
0.248.
This calibrationof the relative population and productivity implies
that the model matchesthe U.S. share of total GDP for the aggregateof
OECDand emerging economies, which in 2005was about 31 percentat
currentexchangerates.Notice that it is the sharein total GDP that matters for the quantitativeresults. Whetherthe 31 percent share derives
from differencesin population size, productivity,or both, is irrelevant
for the economic consequencesof liberalization.5
Forthe calibrationof the productionfunction,we firstset the returnsto-scalecoefficientto v = 0.9 and then we choose 6 to matcha capitalincome shareof 0.36.Becausethe capitalincome shareis 1 - v(l - 6) = 0.36,
this requires8 = 0.289.The depreciationrate is set to 8 = 0.06 and the
subjective discount factor is chosen to produce a capital-outputratio
of 3 in the steady state with capital mobility.This requirementimplies
(3= 0.949.We set the adjustmentcost parameterto <|>= 0.6, which is the
value used by Kehoe and Perri (2002) to match the observed cyclical
variabilityof investmentin industrialcountries.6The coefficientof relative risk aversionparameterin the utility functionis set to a = 2.
The stochastic endowment of labor efficiency follows a two-state
Markovprocess, e = e(l ± Ae),with symmetric transitionprobability
matrix. We calibrate this process to match recent estimates of the
U.S. earningprocess by setting the persistenceprobabilityto 0.975and
Ae= 0.6.Thesevalues imply thatlog-earningsdisplay a first-orderautocorrelationof 0.95 and a standarddeviation of 0.3, which is in the range
of values estimatedby Storesletten,Telmer,and Yaron(2004).
The remaining parametersthat need to be calibratedare the lower
bounds for net worth, a1and a2.These are set to replicatethe volume of
OntheWelfareImplications
of Globalization
withoutDevelopment
293
domesticcreditto the privatesector,in percentageof GDP,observedfor
the United Statesand the aggregateof otherOECDand emergingeconomies. Accordingto the 2005 WorldDevelopment
Indicators,this variable
was 195percentfor the United Statesand 119percentfor the aggregate
of the othercountries.Wereplicatethese values in the steady state with
capitalmobilityby setting a1= -2.6 and a2= -0.02.
63.1 TheMacroeconotnicImplications of Financial Integration
We conduct a simple exerciseto study the consequencesof capitalmarkets liberalization.Wederivethe effectsof a completeand unanticipated
removalof barriersto financialasset tradingbetween countries1 and 2,
startingfroma preliberalizationequilibriumin which the two countries
are in the autarkysteady state.7
Figure6.2 plots the transitiondynamics of key macroeconomicvari-
Figure 6.2
Transition dynamics of macroeconomics variables
294
Mendoza, Quadrini, and Rios-Rull
ables from the autarkysteady state to the new steady state under capital markets integration.Except for the interest rate, the dynamics are
shown as percentdeviations relativeto the autarkysteady state in each
country.Country 1 startswith a higher interestrate than country2 (3.1
percentversus 2.6 percent).This is a consequenceof the lower preliberalization savings of country 1 because of the weaker borrowing constraint.As soon as capitalmarketsare opened, the two interestratesare
equalized at 2.76 percent. After that, the world interest rate rises only
slightly.As a resultof the lower interestraterelativeto autarky,the stock
of physical capitalin country 1 increases.The opposite occurs in country 2.8Thetwo stocksof capitalconvergeto the postliberalizationsteady
state,but the convergenceis gradualbecause of the adjustmentcost. As
shown in the next section, the convergencewithout the adjustmentcost
would be much faster.
The dynamicsof outputs and wages follow fromthe dynamicsof capital: as country 1 increases the stock of capital, output and wages increase.The opposite arises in country2. The dynamics of interestrates
and wages will be key for understandingthe welfare consequencediscussed following.
Figure 6.3 plots the transitiondynamics for the externalaccounts in
percent of each country'sdomestic output. The first panel shows that
country 1 accumulatesa large stock of net foreign liabilities.In the long
run these liabilities reach 81 percent of GDP.This is a gradual process
that takes more than forty years.9
The remainingpanels of figure 6.3 display the flow transactionsstarting with the currentaccount balance. The currentaccount falls into a
large deficitclose to 5 percentof GDP on impact and stays negative until it converges to zero in the long run. Net exports also decline sharply.
The tradebalanceremainsin deficit during the early stages of the transition but it shifts into a surplus in the postliberalizationsteady state.
This surplus is necessaryto service the foreign debt (see the plot showing the factor payments). Note that both the currentaccount and the
tradebalancesatisfy world equilibriumconditionsbecause country1 is
about 31 percentof the world's GDP (e.g., a currentaccountdeficit of 5
percentin countryl's GDPcorresponds,approximately,by a 2.2percent
surplus in country2).
Sincethe responsesof the externalaccountsto financialintegrationresult fromthe responsesof investmentand savings, the last two panels of
figure 6.3 decompose the currentaccount into these two components.
From the national accountingidentity we have that CA = S-I, where
CAis the currentaccountbalance,S nationalsavings, and I domesticin-
On the Welfare Implications of Globalization without Development
295
Figure 6.3
Transitiondynamicsof foreignaccountvariables
vestment. In terms of deviation from the preliberalizationsteady state,
this identity can be written as ACA= AS - AI,where AG4 is the deviation of the currentaccountfrom the steady state value (which is zero).
Similarly,AS and A/ are the deviations of savings and investment from
the steady state values.
Thebottompanels of figure6.3 plot these two variables.The fall (rise)
296
Mendoza, Quadrini, and Rios-Rull
in the currentaccountbalanceof country1 (country2) derives fromboth
an increase (fall) in investment and a decline (rise) in savings. This follows from the fact that the interestratein country 1 (country2) declines
(rises) after liberalization.Lower interest rates encourage investment
but discourage savings. The opposite occurs in country 2. During the
first few periods after liberalization,the contributionof savings to the
currentaccountimbalanceis of similarmagnitudeas the contributionof
investment.
The macroeconomicdynamics illustrated in figures 6.2 and 6.3 are
qualitativelysimilar to those obtained in internationalrepresentativeagent neoclassicalmodels in response to the unilateralremoval of distortionson capitalaccumulation(see Mendozaand Tesar1998),or when
financialintegrationleads to a change in tax structureas a consequence
of increasedtax competition(see Quadrini2005).Representative-agent
models, however, are silent about the distributionaleffects of financial
liberalization.Our setup instead allows us to investigate how the welfare consequences of globalization are distributedamong the population of each country.As we will see, some agents gain greatly while
others are negatively affected.
6.3.2 TheWelfareConsequencesof Liberalization
We study next the welfare implicationsof financialintegration.In the
model, agents are ex-ante identical but ex-post heterogeneous in asset
holdings and earningabilities.As a result,the welfareeffectsof financial
integrationare differentfor each agent depending on their position in
the initial distributionof wealth and earningabilities.
The top panels of figure 6.4 plot the steady state wealth distributions
in the autarkysteady state. Since there are two realizationsof earnings,
the distributionover net worth is plotted separatelyfor agents with low
and high earning abilities.As can be seen, a large share of agents have
very low levels of wealth, in particularamong those with low earning
ability.The lower level of financialdevelopment in country 2 prevents
agents in this countryfrombuilding largenet debt positions,but results
in a large mass of agents aroundthe lower bound for assets.
Thebottom panels of figure6.4 show the bond holdings bt+1as a function of net worth at in the autarkysteady state. Physical capital is the
same for all agentswithin each countrybecausethey all choose the same
kt+1.The importantpoint to note is that agents with lower assets choose
negative values of bt+1;that is, they borrow.It is then easy to see how
On the WelfareImplicationsof Globalizationwithout Development
297
Figure 6.4
Wealth distributions and portfolio composition
changes in the interest rate triggered by financial globalization affect
these agents:they gain froma reductionin the interestratebut they lose
when the interestrateincreases.
Figure6.5 plots the distributionsof the welfare effects of financialintegrationin countries1 and 2 as a functionof initialnet worth, a, and for
differentinitial earningabilities,e. The welfare effects are computed as
the proportionalincreasein consumptionin the autarkysteady state,g,
that would make each individual agent indifferentbetween remaining
in financialautarky(FA)or shiftingto the regime with financialintegration (FT).In the second case agents would experiencethe transitiondynamics from the preliberalizationsteady state to the steady state with
capital mobility.Formally,for each agent ;',who is identified by initial
states (e, a), the welfare gain from capital markets liberalizationis the
value of gj that solves:
Mendoza, Quadrini, and Rios-Rull
298
Figure 6.5
Welfare gains of liberalization
t=0
t=0
where (ftAis agent/'s consumptionunder the autarkysteady state and cfJ
is agent/'s consumptionwhen the economy is liberalizedat time 0, starting from the autarkysteady state. Because of the homotheticityof the
utility function,the previous equationcan also be written as:
(l+g>y-°VFA(e,a) = V»(e,a),
and V"are the value functionsunder financialautarkyand fiwhere VFA
nancialintegration.
As shown in figure 6.5, in country 1 the welfare gains decreasewith
On the Welfare Implications of Globalization without Development
299
initial wealth and at some point they become negative. The opposite
holds trueforcountry2, where (initially)pooreragentsexperiencea loss
fromcapitalmarketsliberalization.
As suggested previously,these welfare findings are a consequenceof
the changes in interestratesafterfinancialmarketsintegration.The decline in the interestrate relativeto autarkyin country 1 is beneficialfor
pooreragentsbecause,as shown in the previous figure,they arenet borrowers.At the same time, theirwages increasebecauseof the increasein
the input of capital.On the other hand, wealthy agents are net lenders
and a small part of their income derives from wages. It is then not surprising that they lose from a fall in the interestrate.The opposite is true
in country2 where the interestrateincreasesafterfinancialmarketsintegration.Pooreragents in country 2 are also borrowers,but for them
the increasein the interestrate raises the cost of servicing the debt. On
the otherhand, richeragents in country2 benefit from the higher interest payments.Becausea smallershareof income is derived fromwages,
the benefit from higher interestpayments more than compensates the
lower wages.
The question we ask next is whether the welfare consequences of
those gaining from liberalizationoffset the negative effects for those
experiencing losses. We address this question by assuming, for each
country,a social welfare functionthat assigns equal weight to all agents
residing in the country.This can be interpretedas the utility of a benevolent country planner that assigns the same weight to each agent. The
aggregatewelfare gain is computed as the proportionalincrease in the
autarkyconsumptionof all agentsthatmakesthe plannerindifferentbetween remainingin autarky (but with the consumption increase) and
liberalizing.For country i, the aggregatewelfare gain is the value of Q
that solves:
*)M'(e,
a).
a)= JMV£'(e,
V*(e,a)M'(e,
(1+ GO1"'
JM
In this aggregate welfare measure, the percentage increase in consumption is the same for all agents within each country.Therefore,this
is also the percentageincreasein aggregateconsumption.Theaggregate
welfare gains are 1.7 percentin country1 and -0.4 percentin country2.
Hence, financial integration produces a substantial welfare gain for
country1 and a nontrivialloss for country2.
Tounderstandthese aggregatewelfareresults,it is useful to look back
at the preliberalizationdistributionsof wealth plotted in figure 6.4. Because most of the agents are concentratedon the left-hand side of the
300
Mendoza, Quadrini, and Rios-Rull
distributions,the aggregatewelfare effects are dominatedby the gains
and losses of the poorer agents. As a result,country 1 gains on average
while country2 is on averageworse off.
These welfare results have some common elements with the normative results of Davila, Hong, Krusell,and Rios-Rull(2005),who study
the constrainedoptimal allocation chosen by a benevolent planner in
a one-sector neoclassic growth model with uninsurable idiosyncratic
shocks.In this environment,the plannercan only affectthe relativewellbeing of differenthouseholds using changes in relative prices and not
individualized transfers.They find that the plannerwould be willing to
induce large changes in the capital stock. In particular,in a calibrated
model where the poor households have an income composition that is
labor intensive, the plannerwould choose a stock of capitalthat is bigger than in the laissez-faireworld. In our environmentthereis a related
mechanismat work;a reductionin the capitalstock and the consequent
increasein the interestrate(whichis what happens in country2 afterliberalization)is undesirablebecause it tilts the prices againstthe poor.
Weconclude the welfare analysisby decomposing the welfare effects
caused by capital reallocation,which also arise in representativeagent
models, fromthose causedby the distributionaleffectsof financialglobalization.Tomake the decomposition,we conduct the following experiment. As in the previous section, we consider a full and unanticipated
liberalizationof capital marketsstartingfrom the autarkysteady state.
However, we do not allow agents to reoptimize the input of physical
capital and continue to use the same inputs they were using in the autarky equilibrium.This is equivalent to assuming that the adjustment
cost is infinitelylarge.
Although the stocks of capital in the two countriesremainthe same,
the interest rates are equalized immediately.This leads to a change in
savings similar to the one obtained in the previous section. Therefore,
country 1 startsborrowingfrom country2 and accumulatesforeign asset liabilities.The distributionof welfare gains among the populationof
the two countriesis also very similar:the poor in country 1 gain while
the poor in country2 lose. Theaggregatewelfaregain for country1 is 2.2
percentand for country2 is -0.74 percent.Therefore,without capitalreallocation,the welfare losses for country2 are larger(0.74versus 0.41).
By allowing capital to be efficientlyreallocated,the losses are reduced
somewhat but not completely.In country 1, instead, the gains are even
largerwithout capitalreallocation(2.2versus 1.55).Clearly,the welfare
On the WelfareImplicationsof Globalizationwithout Development
301
consequences coming from the redistributional impact of globalization
dominate the welfare consequences from capital reallocation.
6.3.3
The Distributional Consequences of Liberalization
In addition to the aggregate and distributional welfare effects of liberalization explored in the previous subsection, financial integration affects
the distribution of wealth within each country. Figure 6.6 plots the Gird
index for the distribution of net worth a in each country. Before liberalization, the Gird index is about 66 percent in country 1 and 48 percent in
country 2. The higher Gird in country 1 is explained by the lower borrowing limit: because of the higher ability to borrow, a significant fraction of agents in country 1 end up with large liabilities, leading to a high
concentration of wealth. Agents in country 2 have tighter borrowing
constraints, and therefore, they cannot accumulate very large liabilities.
As the two countries liberalize, the concentration of wealth increases
gradually in country 1 and in the long run the Gini index reaches the
Figure 6.6
Implicationsof liberalizationfor the dynamicsof the distributionsof wealth
302
Mendoza,Quadrini,andRios-Rull
value of 0.73. In country 2, instead, the Gini index remains almost the
same. These changes are a directconsequenceof the changes in interest
ratesafterliberalization.
Consider first country 1. Liberalizationinduces a fall in the interest
rate (relativelyto the autarkyrate),which reduces the incentiveto save.
As a result of the lower savings, a largerfractionof agents borrowup to
the limit, and therefore,more agents end up with large liabilities.This
patternis consistentwith the increasein inequalityobservedin the U.S.
economy in recentyears, at least for the partgeneratedby the higherindebtedness of the household sector.
Forcountry2, instead, liberalizationleads to an increasein the interest rate. Because of this, savings tend to raise and fewer agents hit the
borrowing limit. This should decrease inequality. At the same time,
however, a higher interestratereducesthe incomes of those borrowing;
that is, the poorer agents. This featuretends to keep more agents closer
to the borrowing limit. For our parametrization,the two effects are of
similarmagnitude and the Gini index does not change much. Interestingly,the limited dataavailableon the evolution of Ginicoefficientsover
the last twenty years also show small changes in the majorityof emerging economies. Using the income distribution data available in the
World Bank'sWorldDevelopmentIndicators,we found that the median
acrossemergingeconomiesof the changebetween the most recentvalue
of the Ginicoefficient(generallyfor2003)and the values reportedforthe
mid-1980s (mainly for 1984 through 1986 depending on the country)
was about 1 percentagepoint. In typical emerging countrieslike Brazil,
Malaysia,Mexico,and Turkey,the changesin the Gini coefficientsrange
between -0.9 and 0.5 percentagepoints.
6.4 Sensitivity Analysis
This section examines the robustnessof our findings to changes in the
specificationof capitaladjustmentcosts and the relativesize of the countries'borrowingconstraints(differencesin financialdevelopment).
In the baselinemodel we assumed thatthe capitaladjustmentcost depends on the aggregatecapitalstock.We want to show now that the results arerobustto using a standardadjustmentcost functioncp(fcf,
kt+1)=
~
•
<|>(kt+1/kt I)2'where ktis the individualstock of capital.The optimization problem is similar to problem (4) but the argumentsof the value
functionnow include kt;that is, V(et,at,kt).The first-orderconditionsfor
OntheWelfareImplications
of Globalization
withoutDevelopment
303
bond holdings and labordo not change. The first-ordercondition with
respectto physical capital,however, is now slightly different:
The last termcapturesthe effectof capitalaccumulationon futureadjustmentcosts. In the previous formulationthis termwas ignoredby individual agents because it was externalto their optimizationproblem.
This additionalterm makes the analysis more complicated.In particular,it is no longer the case that all agents choose the same input of capi> 0) will choose a lower adjusttal. Agents that are constrained(X[ef+1]
ment of capital to increasecurrentconsumption. As a result, they will
have a differentstock of capitalthanunconstrainedagents.This implies
that in the economy therewill be a nondegeneratedistributionof physical capitalin addition to net worth and earningabilities.
Figure6.7 reportsthe transitiondynamics aftercapitalmarketsliberalizationwhen we use this new specificationof the adjustmentcost. The
parameter§ is set to the same value used in the baselinemodel (i.e.,0.6).
The figure also reports the transitiondynamics for the baseline model
analyzed earlier and for the economy without adjustment cost. For
economy of space we reportonly the dynamics for country 1.
Thenew specificationof the adjustmentcost leads to a fasterreallocation of capital compared to the baseline model. This is because now
agents internalizethat higher reallocationtoday reduces the cost of future reallocations.Consequently,the adjustmentis faster.Without adjustmentcost the marginalproductivityof capitalis immediatelyequalized to the interestrate.Withadjustmentcosts, instead, the equalization
arisesonly in the long run.Butbeyond the differencesin the speed of adjustment,the qualitativepropertiesof the model do not change and, as
reportedin table6.1, the welfare numbersare very similar.
Anothersensitivity exercisewe conduct involves the lower bound a,
which representsdifferences in financialmarkets development. Table
6.2 reportsthe welfare consequenceswhen the two economies are characterizedby differentvalues of a. In the firstrow we reportthe baseline
model. In the second we decrease the bound for country 1, and in the
thirdwe decreasethe bound for country2.
Figure 6.7
Transitiondynamicsfor differentadjustmentcosts in country1
305
On the WelfareImplicationsof Globalizationwithout Development
Table 6.1
Welfaregains for differentadjustmentcost functions(percentof consumption)
Modelwithoutadjustmentcost
Modelwith individualadjustmentcost
Modelwith aggregateadjustmentcost (baseline)
Country1
Country2
1.55
1.60
1.67
-0.34
-0.37
-0.41
Table 6.2
Welfaregains for differentbounds on assets (percentof consumption)
Interestrate(autarky)
Welfaregains
Lowerbounds on assets
(a\a2)
Country1
Country2
Country1
Country2
(-2.6, -0.02) (baseline)
H.0,-0.02)
(-2.6,-0.30)
(-2.6,0.00)
(-2.6,0.23)
(-2.6,0.30)
(-2.6,0.46)
3.08
3.25
3.08
3.08
3.08
3.08
3.08
2.62
2.62
3.00
2.59
2.14
1.95
1.53
1.67
3.02
0.35
1.76
3.00
4.00
5.60
-0.41
-0.58
-0.13
-0.43
-0.50
-0.36
-0.08
Largerdifferencesbetween the lower bounds lead to largerwelfare
effects. For example, when we decrease the limit for country 1 to -4.0
(keepingthe limit of country2 unaltered),the gains of country1 and the
losses of country2 increaseby a factorof 1.8 and 1.4,respectively.When
we decreasethe limit for country2 from -0.02 to -0.3 (makingthe limit
closer to country 1), the welfare consequencesbecome less than half.10
So, in general, what matters is how much the borrowing limits differ
acrosscountries.
Weshould also point out thatthe welfareconsequencesfor the less developed countrydo not change monotonicallywith the tightness of the
financialconstraint.As shown in the lower section of table 6.2, after a
certainpoint, the welfarelosses for country2 startto decline with tighter
constraints.
Thisresultis easy to understandif we use the law of motion forwealth
(equation[2])to rewritethe financialconstraintas follows:
+ bt+1> a.
+ F(kt+1,
lt+1) lt+1wt+1
et+1wt+1
Fromthis expression,it is easy to see thatthe lower limit on at+1imposes
a lower limit on bt+rBecauseall agents choose the same productionin-
Mendoza, Quadrini, and Rios-Rull
306
puts kt+1and lt+1,high values of a reduce the maximumleverage.Then,
an increasein the interestrate in country 2 following the liberalization
of capital marketscannot have large welfare costs for poor agents because they are not very leveraged.Effectively,higher values of a reduce
the inequality in the distributionof wealth with consequent reduction
in the redistributionalconsequencesof capitalmarketsliberalization.In
the extremecase in which a is so high that bt+1is zero for all agents, the
welfare consequencesof liberalizationbecome positive also for country
2. In this case only the efficiencygains fromcapitalreallocationarise.Of
course, this is an extremecase of limited empiricalrelevance.
Regardingcountry 1, the welfare gains increasemonotonicallywith
the reductionin the financialdeepness of country2. This is because the
integrationwith less financially developed countries implies a larger
drop in the interestrate.
The last sensitivity exerciseis with respectto the volatilityof individual laborefficiencies.Laborefficienciescan taketwo values and the proportionaldeviation from the mean, Ae,was set to 0.6 in both countries.
Wenow allow the two countriesto have differentvolatilitiesof earnings.
Table6.3 reportsthe welfare consequenceswhen the two economies are
characterizedby different values of Ae.In the first row we report the
baseline model. In the second and third rows we change the earnings
volatility in country 1, and in the fourth and fifth rows we change the
earningsvolatility in country2.
Lower (higher)volatility of earningsleads to lower (higher)savings,
and therefore,a higher (lower) interestrate in the autarkyequilibrium.
The largerthe interestrate differentialbefore capitalliberalization,the
largerthe welfare consequencesfromliberalizing.In otherwords, liberalization brings higher welfare gains or lower welfare losses as we reduce the relativevolatility of individual earnings.
Table6.3
Welfaregains for differentvolatilityof earnings(percentof consumption)
„ . ... of,
Volatility earnings
(AJ,A*)
Interestrate(autarky)
Country1
Country2
Country1
Country2
(0.60,0.60)(baseline)
(0.55,0.60)
(0.65,0.60)
(0.60,0.55)
(0.60,0.65)
3.08
3.41
2.73
3.08
3.08
2.62
2.62
2.62
3.00
2.22
1.67
2.64
0.43
0.29
3.21
-0.41
-0.69
-0.10
-0.06
-0.83
Welfaregains
On the Welfare Implications of Globalization without Development
307
The results reportedin rows 2 and 5 are of particularinterest.These
are the cases in which agents in country 1 face lower volatility of earnings comparedto agents in country2 (in addition to a weaker borrowing constraint).This is anotherway of capturing,in reduced form, the
higher financialdeepness of country1. In the currentpaper we have assumed that financialmarketdifferencesare only capturedby the borrowing limits. More generally,however, financialdevelopment is also
associatedwith the availabilityof a largervariety of contracts,like derivatives,thatareessentiallystate-contingentand provide some formof
insurance.Becausegreaterinsurancereducesthe need of precautionary
savings, liberalizationleads to largerimbalancesand largerwelfareconsequences.This is the approachwe have taken in Mendoza, Quadrini,
and Rios-Rull(2007),where the availabilityof insuranceis determined
endogenously by the degree of contractenforceability.
The case of cross-countrydifferencesin the volatility of earnings is
also relevant for understanding the internationalflow of capital between industrializedand emerging economies. The economic transformation experiencedby emerging economies allowed these countriesto
experiencefaster growth. At the same time, rapid transformationsare
also likely to induce greateruncertaintyat the individual level. Therefore,we have two opposing effects on the internationalflows of capital.
On the one hand emerging economies should be net importersof capitalbecausefastergrowth induceshigherinvestments.On the other,they
should be net exportersbecause the economic transformationsthat allow for fastergrowth also induce highersavings. Togetherwith the poor
development of their financial markets, the saving effect is likely to
dominatethe investmenteffect in these countries.It is then not surprising to see a net flow of capitalgoing from emerging economies toward
industrializedcountries.See also endnote 9 on this point.
6.5 Policy Implications
One importantresultwe obtainedfrom the welfare analysis of financial
liberalizationis that in country 2- the country with less developed financialmarkets- the aggregatewelfare consequencesare negative. In
this sectionwe ask whetherthereis a simple domesticpolicy that,when
applied in conjunctionwith financialliberalization,can make the aggregate welfareconsequencespositive for country2 (in addition to country
1). Of course, there are many policies we can think of. Here we simply
considera once-and-for-allredistributivetax on wealth.
Mendoza, Quadrini, and Rios-Rull
308
Table 6.4
Welfaregains fromliberalizationwhen country2 implementsa one-timewealth redistribution (percentof consumption)
Taxrate(%)
0.0
1.0
2.5
5.0
Initialwealth Gird
afterredistribution
in country2
Welfaregains
country1
Welfaregains
country2
0.482
0.477
0.470
0.458
1.67
1.64
1.61
1.56
-0.41
-0.20
0.12
0.62
Suppose that at time zero, just before liberalization,the government
imposes a tax on net worth a and redistributesthe revenues as lumpsum transfers.Formally,V = j^YaM^e,a), where V is the lump-sum
transfer,t1the wealth tax rate,and i is the index for the country.This redistributivetax is applied only at time zero and it is not anticipated.We
tried differenttax rates for country 2 and the aggregatewelfare results
are reportedin table 6.4.
As more wealth is redistributedinitially,and hence there is less initial inequality,the aggregatewelfare consequencesimprove. A 2.5 percent tax rateis sufficientto turn the aggregatewelfare cost of liberalization in country 2 into a welfare gain. The welfare gain is further improved by increasingthe tax rate. It is also interestingto note that the
redistributivetax in country 2 has almost no effect on the welfare of
country 1.
6.6 Conclusion
This paper studied the welfare and distributionalconsequencesof capital marketsliberalizationamong countriesthat areheterogeneousin financial development. We found that if financialglobalizationdoes not
lead to financialdevelopment, it can result in adverse effects for the social welfare of the less developed countries.The quantitativeanalysis
shows that these effects are large and may justifypolicy intervention.
These conclusionswere obtainedby studying the quantitativeimplications of a two-countrymodel with heterogenousagents and uninsurable idiosyncraticrisks. Marketincompleteness derives from the unavailability of state-contingentcontracts and limits to the amount of
borrowing.Countriesare heterogenousin the degree of financialdevelopment capturedby the tightness of the borrowing limits. A baseline
On the Welfare Implications of Globalization without Development
309
simulationcalibratedto the United Statesand an aggregateof the rest of
the OECD countries plus emerging economies, showed that financial
globalizationwithout financialdevelopment leads to an aggregatewelfaregain of 1.7 percentin the United Statesand a loss of -0.4 percentin
the aggregateof the other countries.This occursbecause, relative to financialautarky,the integrationof capital marketsincreasesthe cost of
borrowingto poor agentsin the less financiallydeveloped countriesand
lowers it in the more financially developed country. Moreover, the
model predicts that financialglobalizationwithout financialdevelopment has adverse effects on wealth inequality.It worsens the distribution of wealth in the United States and fails to reduce inequality in the
othercountries.
If financialintegrationleads to fasterconvergencein the development
of financialmarkets,then therecouldbe net benefitsforthe countriesthat
are less financiallydeveloped. This is more likely to happen in the long
run. In the short run, fiscalpolicies can play a useful role in redistributing the gains and losses. We found that a redistributionof initialwealth
in countrieswith lower financialdevelopment can make the aggregate
welfareconsequencesfromliberalizationpositive for all countries.
Acknowledgments
This paper was prepared for the NBER's InternationalSeminar on
Macroeconomics(ISOM),Istanbul,June 15-16, 2007. We would like to
thank GiancarloCorsetti and Mehmet Yorokoglufor insightful comments.Financialsupportfromthe NationalScienceFoundationis gratefully acknowledged.
Notes
1. This assumption is typical in international real business cycle models and is used there
for the same reason; that is, to make the reallocation of capital sluggish.
2. The asset market structure and the treatment of financial heterogeneity in this model
are less general than the ones in the MQR setup. In MQR we allow for a full set of statecontingent assets that are constrained by an enforcement constraint. Here, instead, we allow only for nonstate-contingent assets. With earnings shocks only, the two market structures lead to similar results. With investment shocks, however, the two market structures
have different implications for the composition of portfolio. Because in this paper we
abstract from investment shocks, we decided to adopt the simpler specification.
3. The savings supply curves in figure 6.1 correspond to the well-known aggregate savings curve from the closed-economy heterogenous agents literature (e.g., Aiyagari 1994).
310
Mendoza, Quadrini, and Rios-Rull
Aggregate savings converge to infinity as the interest rate approaches the rate of time preference from below, because agents need an infinite amount of precautionary savings to
attain a nonstochastic consumption profile. The demand of savings is downward sloping
because of the diminishing marginal productivity of capital.
4. The emerging economies are Argentina, Brazil, Chile, China, Colombia, Costa Rica,
Ecuador, Egypt, El Salvador, Hong Kong, India, Indonesia, Malaysia, Morocco, Nigeria,
Pakistan, Peru, Philippines, Singapore, Sri Lanka, South Africa, Taiwan, Uruguay, and
Venezuela. In general, these are the developing countries that today are most active in international financial markets. Because the OECD and the emerging economies represent a
very large share of world output, the inclusion of all other countries would not make a big
difference for the quantitative results.
5. With proper rescaling of the limited liability constraints, a\ different combinations of
populations and productivities that keep the same relative aggregate GDP of the two
countries constant do not affect the autarky interest rates. The postliberalization interest
rate will be closer to the autarky interest rate of the country with the larger economy, independently of whether this derives from a larger population or a higher productivity.
6. Capital adjustment costs are widely used in international real business cycle (RBC)
models, but there is no consensus on the calibration strategy to parameterize them. For example, Baxter and Crucini (1993) calibrated the elasticity of investment relative to Tobin's
q to match investment variability in industrial countries. Chari, Kehoe, and McGrattan
to match the relative variability of consumption to out(2000) calibrated the parameter <(>
put. Kehoe and Perri (2002) targeted the variability of investment.
7. In reality, the process of financial globalization has been gradual. However, the effects
of gradual financial integration are easy to infer from those of the immediate liberalization.
The transitional dynamics of the gradual liberalization would spread over time the impact
of the immediate liberalization.
8. Interestingly, the investment rate in the emerging countries of South East Asia (excluding China) fell by 10 percentage points of GDP in 1998 and this decline has persisted for
a decade. The initial drop can be attributed to the Asian crisis, but the persistence of the
decline suggests that there was also some structural change at play.
9. The model's predictions are again in line with the current situation in the world economy where poorer and less financially developed countries are net lenders to more developed countries, particularly the United States. We should also notice, however, that these
countries are experiencing faster growth than industrialized countries. With growth
differences, the predictions of our model are similar to those of a typical open economy
neoclassical model; that is, fast growing countries should be net importers of capital.
Therefore, once we consider growth differences between emerging and industrialized
countries, the net flow of capital will be determined by two opposing forces: growth differences lead to an inflow toward emerging economies while financial markets differences
lead to an inflow toward industrialized countries. Given that the degree of financial development in emerging economies is so far behind the one in industrialized economies, we
believe that the financial channel still dominates the growth channel so that the net flow
continues to be directed toward industrialized countries.
10. The lower bound for country 2 is not directly comparable to country 1 because country 1 has a higher per capita income. To make the bound comparable, we should rescale it
by the per capita income ratio of the two countries. Because the ratio is 0.152, the bound
for country 2, a2 = -0.3, corresponds to -0.3/0.152 = -1.97 in country 1. Hence, financial
On the WelfareImplicationsof Globalizationwithout Development
311
integration would be neutral if a2 = -0.3952, because this corresponds to -0.3952/0.152 =
-2.6, which is the limit for country 1 in the baseline calibration.
References
Aiyagari, S. R. 1994. Uninsured idiosyncratic risk, and aggregate saving. QuarterlyJournal
of Economics109:659-84.
Backus, D. K., P.J. Kehoe, and F.Kydland. 1992. International business cycles. Journalof Political Economy100 (4): 745-75.
Baxter, M., and M. J. Crucini. 1993. Explaining saving-investment correlations. American
EconomicReview 83 (3): 416-36.
Bewley, T. 1986. Stationary monetary equilibrium with a continuum of independently fluctuating consumers. In Contributionsto MathematicalEconomicsin Honorof GerardDebreu,ed.
W. Hildenbrand, and A. Mas-Colell, pp. 79-102. Amsterdam: North Holland.
Bhagwati, J. 2004. In defenseof globalization.New York:Oxford University Press.
Carroll, C. D. 1997. Buffer-stock saving and the lifecycle/permanent income hypothesis.
QuarterlyJournalof Economics112 (1): 1-55.
Chart, V. V., P. Kehoe, and E. McGrattan. 2000. Sticky price models of the business cycle:
Can the contract multiplier solve the persistence problem? Econometrica68 (5): 1151-80.
Clarida, R. H. 1990. International lending and borrowing in a stochastic stationary equilibrium. InternationalEconomicReview 31 (3): 543-58.
Davila, J., J. H. Hong, P. Krusell, and J.-V. Rios-Rull. 2005. Constrained efficiency in the
neoclassical growth model with uninsurable idiosyncratic shocks. Cahiers de la Maison
des Sciences Economiques b05066, University Panth&m-Sorbonne.
Frankel, J. 2007. New perspectives on financial globalization. Panel presentation at International Monetary Fund (IMF) conference, 27 April.
Gourinchas, P.-O., and O. Jeanne. 2006. The elusive gains from international financial integration. Reviewof EconomicStudies 73 (3): 1-31.
Huggett, M. 1993. The risk free rate in heterogeneous-agents, incomplete insurance economies. Journalof EconomicDynamics and Control17 (5/6): 953-70.
Kehoe, P. J., and F. Perri. 2002. International business cycles with endogenous incomplete
markets. Econometrica70 (3): 907-28.
Mendoza, E. 1991. Capital controls and the dynamic gains from trade in a business cycle
model of a small open economy. IMF Staff Papers 38 (September): 480-505. Washington,
D.C.: International Monetary Fund.
Mendoza, E. G, V. Quadrini, and J.-V. Rios-Rull. 2007. Financial integration, financial
deepness, and global imbalances. NBER Working Paper no. 12909. Cambridge, MA: National Bureau of Economic Research, February.
Mendoza, E. G and L. L. Tesar. 1998. The international ramifications of tax reforms:
Supply-side economics in a global economy. AmericanEconomicReview 88 (1): 226-45.
Mishkin, F. S. 2006. The next greatglobalization.Princeton, N.J.: Princeton University Press.
312
Mendoza, Quadrini, and Rios-Rull
Obstfeld, M., and A. M. Taylor. 2004. Globalcapital markets:Integration,crisis, and growth.
New York:Cambridge University Press.
Quadrini, V. 2005. Policy commitment and the welfare gains from capital market liberalization. EuropeanEconomicReview 49 (8): 1927-51.
Rajan, R., and L. Zingales. 2003. Saving capitalismfrom the capitalists:Unleashingthepowerof
financial marketsto createwealth and spreadopportunity.New York:Crown Business.
Soros, G. 2002. On globalization.New York:Public Affairs Books.
Stiglitz, J. E. 2002. Globalizationand its discontents.New York:W. W. Norton and Company.
Storesletten, K., C. Telmer, and A. Yaron. 2004. Cyclical dynamics in idiosyncratic labormarket risk. Journalof Political Economy112 (3): 695-717.
Appendix
Proof of Proposition 1
In both economies we have that in the steady state Fk(K,L) = 1 + r. Because with capitalmobility there is a single worldwide interestrate,all
agents employ the same input of capital in steady state. We want to
show that the steady state interestrateis smallerthan the intertemporal
discount rate. Suppose, on the contrary,that ($(1+ r) > 1. Under this
condition agents in country 1 will have nonnegative consumption
growth (see Lemma 1) and agents in country 2 will have positive consumption growth (see Lemma 2). This implies that worldwide consumption growth is positive, which cannot be a steady state equilibrium. Therefore,the equilibriummust satisfy 0(1 + rt)< 1. Under this
condition, agents in country 1 will experience negative consumption
growth (see againLemma1).Therefore,consumptionin country1 keeps
fallinguntil the limited liabilityconstraint(3)binds for all agents;thatis,
the net worth equals a1under the lower realizationof the earningshock.
Becausecountry1 holds the same amountof Kas country2, the binding
constraintfornet worth implies thatcountry1 has a negative net foreign
asset position.
Q.E.D.