Fiscal policy and interest rates: the role of financial and economic integration *



Similar documents
The Greek financial crisis: growing imbalances and sovereign spreads. Heather D. Gibson, Stephan G. Hall and George S. Tavlas

4. International Parity Conditions

BALANCE OF PAYMENTS. First quarter Balance of payments

II.1. Debt reduction and fiscal multipliers. dbt da dpbal da dg. bal

Vector Autoregressions (VARs): Operational Perspectives

A Note on Using the Svensson procedure to estimate the risk free rate in corporate valuation

Cointegration: The Engle and Granger approach

11/6/2013. Chapter 14: Dynamic AD-AS. Introduction. Introduction. Keeping track of time. The model s elements

Chapter 8: Regression with Lagged Explanatory Variables

Measuring macroeconomic volatility Applications to export revenue data,

How To Calculate Price Elasiciy Per Capia Per Capi

Appendix D Flexibility Factor/Margin of Choice Desktop Research

Supplementary Appendix for Depression Babies: Do Macroeconomic Experiences Affect Risk-Taking?

Risk Modelling of Collateralised Lending

Morningstar Investor Return

PROFIT TEST MODELLING IN LIFE ASSURANCE USING SPREADSHEETS PART ONE

Chapter 7. Response of First-Order RL and RC Circuits

CHARGE AND DISCHARGE OF A CAPACITOR

Principal components of stock market dynamics. Methodology and applications in brief (to be updated ) Andrei Bouzaev, bouzaev@ya.

MACROECONOMIC FORECASTS AT THE MOF A LOOK INTO THE REAR VIEW MIRROR

DOES TRADING VOLUME INFLUENCE GARCH EFFECTS? SOME EVIDENCE FROM THE GREEK MARKET WITH SPECIAL REFERENCE TO BANKING SECTOR

The Real Business Cycle paradigm. The RBC model emphasizes supply (technology) disturbances as the main source of

Usefulness of the Forward Curve in Forecasting Oil Prices

Determinants of Capital Structure: Comparison of Empirical Evidence from the Use of Different Estimators

Relationships between Stock Prices and Accounting Information: A Review of the Residual Income and Ohlson Models. Scott Pirie* and Malcolm Smith**

Journal Of Business & Economics Research September 2005 Volume 3, Number 9

BALANCE OF PAYMENTS AND FINANCIAL MA REPORT All officiell statistik finns på: Statistikservice: tfn

Why Did the Demand for Cash Decrease Recently in Korea?

Real long-term interest rates and monetary policy: a cross-country perspective

Individual Health Insurance April 30, 2008 Pages

Duration and Convexity ( ) 20 = Bond B has a maturity of 5 years and also has a required rate of return of 10%. Its price is $613.

WORKING PAPER SERIES STOCKS, BONDS, MONEY MARKETS AND EXCHANGE RATES MEASURING INTERNATIONAL FINANCIAL TRANSMISSION NO.

Can Austerity Be Self-defeating?

Analysis of tax effects on consolidated household/government debts of a nation in a monetary union under classical dichotomy

Internal and External Factors for Credit Growth in Macao

Determinants of Bank Long-term Lending Behavior in the Central African Economic and Monetary Community (CEMAC)

Market Liquidity and the Impacts of the Computerized Trading System: Evidence from the Stock Exchange of Thailand

Migration, Spillovers, and Trade Diversion: The Impact of Internationalization on Domestic Stock Market Activity

Chapter 6: Business Valuation (Income Approach)

Chapter 8 Student Lecture Notes 8-1

The Asymmetric Effects of Oil Shocks on an Oil-exporting Economy*

ARCH Proceedings

The Identification of the Response of Interest Rates to Monetary Policy Actions Using Market-Based Measures of Monetary Policy Shocks

Economics Honors Exam 2008 Solutions Question 5

Monetary Policy & Real Estate Investment Trusts *

When Is Growth Pro-Poor? Evidence from a Panel of Countries

GOOD NEWS, BAD NEWS AND GARCH EFFECTS IN STOCK RETURN DATA

CRISES AND THE FLEXIBLE PRICE MONETARY MODEL. Sarantis Kalyvitis

Small and Large Trades Around Earnings Announcements: Does Trading Behavior Explain Post-Earnings-Announcement Drift?

is a random vector with zero mean and Var(e

Measuring the Effects of Exchange Rate Changes on Investment. in Australian Manufacturing Industry

Estimating the immediate impact of monetary policy shocks on the exchange rate and other asset prices in Hungary

Working paper No.3 Cyclically adjusting the public finances

The Interest Rate Risk of Mortgage Loan Portfolio of Banks

The Grantor Retained Annuity Trust (GRAT)

Debt Accumulation, Debt Reduction, and Debt Spillovers in Canada, *

Chapter 9 Bond Prices and Yield

Hedging with Forwards and Futures

Evidence from the Stock Market

The impact of Federal Reserve asset purchase programmes: another twist 1

Foreign Exchange Market Microstructure

Do Credit Rating Agencies Add Value? Evidence from the Sovereign Rating Business Institutions

Bid-ask Spread and Order Size in the Foreign Exchange Market: An Empirical Investigation

Flight-to-Liquidity and Global Equity Returns

LECTURE: SOCIAL SECURITY HILARY HOYNES UC DAVIS EC230 OUTLINE OF LECTURE:

Migration, Spillovers, and Trade Diversion: The Impact of Internationalization on Stock Market Liquidity

Mathematics in Pharmacokinetics What and Why (A second attempt to make it clearer)

LEASING VERSUSBUYING

JEL classifications: Q43;E44 Keywords: Oil shocks, Stock market reaction.

Estimating the Term Structure with Macro Dynamics in a Small Open Economy

Why does the correlation between stock and bond returns vary over time?

Index funds and stock market growth

Can Individual Investors Use Technical Trading Rules to Beat the Asian Markets?

USE OF EDUCATION TECHNOLOGY IN ENGLISH CLASSES

Behavior and Importance of Bank Loan Components after Monetary and Non-Monetary Shocks

Debt Relief and Fiscal Sustainability for HIPCs *

A Note on the Impact of Options on Stock Return Volatility. Nicolas P.B. Bollen

Investor sentiment of lottery stock evidence from the Taiwan stock market

When Do TIPS Prices Adjust to Inflation Information?

Default Risk in Equity Returns

Stability. Coefficients may change over time. Evolution of the economy Policy changes

MEDDELANDEN FRÅN SVENSKA HANDELSHÖGSKOLAN SWEDISH SCHOOL OF ECONOMICS AND BUSINESS ADMINISTRATION WORKING PAPERS

The Impact of Surplus Distribution on the Risk Exposure of With Profit Life Insurance Policies Including Interest Rate Guarantees.

Does Stock Price Synchronicity Represent Firm-Specific Information? The International Evidence

Estimating Time-Varying Equity Risk Premium The Japanese Stock Market

The Impact of Surplus Distribution on the Risk Exposure of With Profit Life Insurance Policies Including Interest Rate Guarantees

DYNAMIC MODELS FOR VALUATION OF WRONGFUL DEATH PAYMENTS

Title: Who Influences Latin American Stock Market Returns? China versus USA

Dynamic co-movement and correlations in fixed income markets: Evidence from selected emerging market bond yields

AP Calculus BC 2010 Scoring Guidelines

DNB W o r k i n g P a p e r. Stock market performance and pension fund investment policy: rebalancing, free f loat, or market timing?

CURRENT ACCOUNTS IN THE EURO AREA: AN INTERTEMPORAL APPROACH. José Manuel Campa Angel Gavilán

NATIONAL BANK OF POLAND WORKING PAPER No. 119

Terms of Trade and Present Value Tests of Intertemporal Current Account Models: Evidence from the United Kingdom and Canada

WORKING PAPER SERIES GLOBAL CRISES AND EQUITY MARKET CONTAGION NO 1381 / SEPTEMBER 2011

News Intensity and Conditional Volatility on the French Stock Market

The Transmission of Pricing Information of Dually-Listed Stocks

VIX, Gold, Silver, and Oil: How do Commodities React to Financial Market Volatility?

Transcription:

Research Insiue of Applied Economics 2008 Working Papers 2008/10, 29 pages Fiscal policy and ineres raes: he role of financial and economic inegraion * By Peer Claeys, Rosina Moreno and Jordi Suriñach AQR Research Group - IREA. Deparmen of Economerics, Universiy of Barcelona. Avda. Diagonal, 690. 08034 Barcelona, Spain. Email: Peer.Claeys@ub.edu. Absrac: I is commonly believed ha a fiscal expansion raises ineres raes. However, hese crowding ou effecs of deficis have been found o be small or non-exisen. One explanaion is ha financial inegraion offses ineres rae differenials on globalised bond markes. This paper measures he degree of inegraion of governmen bond markes, using spaial modelling echniques o ake his spillover on financial markes ino accoun. Our main finding is ha he crowding ou effec on domesic ineres raes is significa bu is reduced by spillover across borders. This spillover is imporan in major crises or in periods of coordinaed policy acions. This resul is generally robus o various measures of cross-counry linkages. We find spillover o be much sronger among EU counries. Key words: fiscal policy, spillover, ineres raes, crowding ou, spaial models. JEL codes: C21, E43, E58, E62, F42.. * We would like o hank Anna Giribe for excellen research assisance. Peer Claeys acknowledges suppor by a Marie Curie Inra-European Fellowship wihin he 6h European Communiy Framework Programme. 1

Research Insiue of Applied Economics 2008 Working Papers 2008/10, 29 pages 1 INTRODUCTION A governmen running a defici needs o urn o financial markes o place his addiional public deb. Newly issued public bonds compee for financing wih bonds issued by privae agens. The addiional demand creaed by he fiscal expansion pushes up ineres raes, and evenually crowds ou privae invesmen. No all economiss agree ha consolidaing public finances would immediaely reduce pressure on ineres raes, however. Despie a vas lieraure esing crowding ou, here is acually surprisingly lile robus empirical suppor for his hypohesis. 1 Ineres raes are insulaed from fiscal policy under wo alernaive condiions. The firs explanaion for a zero impac of deficis on aggregae macroeconomic variables is ha economic agens anicipae paying down currenly high deficis wih higher axes in he fuure. Under Ricardian Equivalence, privae saving fully offses he effec of higher public consumpion (for a given level of axaion). Few economiss would consider he assumpions underlying he Ricardian Equivalence null as realisic, however. More elaborae macroeconomic models ha depar from he baseline Ricardian assumpion easily find real economic effecs of fiscal policy. There is by now also a large body of empirical evidence ha clearly refues he Ricardian hypohesis (Blanchard and Peroi, 2002). A second explanaion for he lacking crowding ou effec is capial mobiliy. Fiscal deficis need no be financed by domesic financial resources only. Capial flows beween economically inegraed economies, offseing any ineres rae differenials ha follow upon an increase in he domesic supply of governmen bonds. Under full capial mobiliy, domesic and foreign agens alike diversify heir asse porfolio across borders. As a consequence, he budge decision of one governmen affecs he financing condiions of all oher governmens on inernaional capial markes. Domesic ineres raes rise in proporion o he amoun of bonds issued worldwide. For a small open economy, he crowding ou effec would end o zero. In pracice, capial mobiliy is far from complee, as imperfec informaio risk aversion and imperfec subsiuabiliy of domesic and foreign bonds inroduce a home bias in porfolio decisions. As a consequence, he spillover is likely o be less han complee. The empirical models ha are used o assess he crowding ou effec of fiscal policy fail o accoun for his spillover. A baseline es for crowding ou ypically regresses a domesic ineres rae on some domesic fiscal indicaor. Even simple exensions of his model o include he effecs of fiscal policy in open economies require quie resricive assumpions on parameerisaion. Ofe one simply conrols for a se of addiional (foreign) explanaory variables. Usually, only a paricular subse of counries is examined, or idenical resricions are imposed on he ransmission of fiscal 2

Research Insiue of Applied Economics 2008 Working Papers 2008/10, 29 pages policy across all counries. Due o he dimension of open economy models, even simple exensions quickly exhaus he available degrees of freedom. In pracice, he ineracions are much more complex. Spillover works ou on global financial markes, and affecs a large group of counries conemporaneously. In his paper, we use spaial echniques o accoun for his spillover. We deliberaely keep he baseline model as simple as possible o make he sronges possible case for spillover. We es a panel model ha explains ineres raes by fiscal variables o analyse crowding ou of fiscal policy. The spaial model simply exends his baseline model for he spillover effec in all nearby foreign economies. In paricular, we es he spillover of ineres raes on financial markes in a spaial lag model. We hen conrol also for spaially disribued economic linkages in a spaial error model. We es his model on a large cross secion of OECD and emerging marke economies over he period 1990-2005. Our main finding is ha he domesic crowding effec of fiscal policy is sizeable. Bu he spillover on financial markes offses he significan effecs of larger deficis on ineres raes. We canno idenify wheher his spillover is he direc effec of financial marke inegraio or he by-produc of he economic inegraion of counries. Spillover is much sronger in he midnineies when here were major crises, or policy acions were being coordinaed beween governmens. Various measures of cross-counry linkages give broadly similar resuls. The main findings are also robus o alernaive specificaions and daa definiions. Finally, we find he spillover o be quie srong among EU counries. The paper is srucured as follows. In secion 2, we provide a simple heoreical model for esing crowding ou, and he effecs of financial and economic inegraion. We consequenly specify he spaial model for including his spillover in secion 3. We hen coninue in secion 4 by discussing he resuls of he baseline model of spillover of fiscal policies, and provide several robusness checks. The final secion summarises he main resuls, and discusses some policy implicaions. 2 CROWDING OUT AND SPILLOVER Crowding ou of ineres raes is ypically analysed in a parial equilibrium loanable funds model (Barro, 1992). This model deermines he ineres rae on asses from he equilibrium beween he demand and supply bonds. Boh he privae secor and he governmen urn o financial markes o look for finance. Firms inves he capial raised on sock or bond markes (b) o inves in new capial sock. Addiional governmen financing is necessary when he governmen runs a defici (d). Financial inermediaries channel he demand for bonds of he privae secor, boh a home (f) and abroad (f*) secor, o mach he oal supply of domesic bonds (b+d): b d f f * (1) 1 See he conrasing argumens of he European Commission (2004) and he Bush Adminisraion (Gale and Orszag, 3

Research Insiue of Applied Economics 2008 Working Papers 2008/10, 29 pages We can illusrae all he major poins of he analysis wih his simple parial equilibrium model. 2 Consider he case in which he governmen runs a higher defici d. For a given demand for bonds f+f*, he increased supply of deb will pu downward pressure on he price of governmen bonds. Ceeris paribus, his rise in bond yields is making i more difficul for he privae secor o seek finance on capial markes: governmen finance crowds ou privae bonds b on capial markes (Cebula, 1998). However, he demand for addiional bonds is likely affeced by he governmen s decision o lend. Under wo alernaive heoreical condiions, privae secor savings fully offse he addiional supply of bonds. Firs, if economic agens anicipae he pay down of higher deficis hey se aside savings for he higher ax burden in fuure periods. In he limi, domesic privae saving fully offses he effec of he higher public dissaving d. Under his Ricardian Equivalence hypohesis, a higher defici d does no have an impac on aggregae macroeconomic variables a all. Many economiss consider Ricardian Equivalence as a reasonable saring poin for he analysis of fiscal policy. Few would endorse i as a realisic descripio however. The view ha privae savings offse he change in public savings is no based on a firm empirical rejecion since Ricardian Equivalence is no direcly esable. Bu pleny of empirical sudies have examined he alernaive hypohesis ha fiscal policy has any real economic effecs. Recen evidence seems o converge on a leas some expansionary effecs on major economic variables (Blanchard and Peroi, 2002). More elaborae macroeconomic models ha depar from he baseline Ricardian assumpion easily find suppor for hese real economic effecs of fiscal policy. I herefore seems a safe assumpion o rejec he Ricardian Equivalence hypohesis. Second, when financial markes are inegraed across borders, he foreign demand for bonds f* can shif as well. In open economies ha are economically inegraed and do no impede rade or financial flows, capial flows move massively so as o offse any ineres rae differenial. Under hese circumsances, he supply of savings is very ineres rae elasic: even a small rise in d is likely o rigger a large increase in f*. The higher is capial mobiliy, he weaker will be he reacion of domesic ineres raes o a change in he supply of domesic bonds. Under he null of full capial mobiliy, he rise in ineres raes is simply proporional o each counry s oal indebedness on he global bond marke. Consequenly, due o he spillover on inernaional bond markes, he crowding ou effec of deficis is only a fracion of he oal rise under auarchy. In realiy, his diluion is likely o be less han complee as capial mobiliy is parial: domesic privae agens prefer o inves in domesic financial asses. As a consequence, domesic savings and invesmen are highly correlaed (he Feldsein-Horioka puzzle). This home bias depends on informaion 2003). 2 These parial equilibrium models have been exended for ineremporal saving behaviour (Laubach, 2003; Engen and Hubbard, 2004). Dynamic macroeconomic models ha include boh deb non-neuraliy and long erm ineres raes have no been developed ye, due o heir complexiy. 4

Research Insiue of Applied Economics 2008 Working Papers 2008/10, 29 pages imperfecions on foreign financial markes. Neiher are financial asses in differen counries perfec subsiues, due o exchange rae, inflaion and defaul risk. Due o differences in regulaion across counries, risk averse agens may prefer o inves in domesic asses only. Moreover, privae agens are likely o hold a larger porion of domesic public deb. Governmens ofen prefer o place deb only domesically in order o avoid having o pay exchange rae premia, and as a commimen no o defaul on deb pu wih is own ciizens. As a consequence, complee spillover is unlikely, ye i is hard o pu a precise size on he spillover effec. The spillover is likely o be sronger beween economies ha are more closely inegraed. 3 A SPATIAL TEST FOR CROWDING OUT The mos common es for crowding ou based on he loanable funds model akes a very simple form: i basically explains domesic ineres raes by domesic fiscal balances (in his case, he surplus s ). i. (2a) s We measure by he coefficien he degree of crowding ou. The large number of sudies ha have employed various definiions of he governmen surplus, ineres raes, economeric approaches and daa ses o es (2a) can basically give suppor for any view. 3 There are wo cases in which we would no rejec 0. Firs, we would no find a significan crowding ou effec under Ricardian Equivalence. Second, we may no be able o rejec 0 if here is capial mobiliy beween open economies. However, a specificaion like (2a) does no accoun for hese spillover effecs of financial markes. Basically, he es for crowding ou only considers domesic variables, bu does no accoun for he inegraion of he domesic economy wih foreign economies. Exising empirical evidence on he effec of inegraion is lile, and uses alernaive ways for neing ou he inernaional linkages from he domesic crowding ou effec. Due o he omission of hese foreign variables ha explain he spillover effec, would be biased downward. The reasoning is ha he domesic effec of fiscal expansions will likely be larger once a proxy X for he foreign demand for bonds f* is inroduced. i s X. (2b) One alernaive is o condiion he relaion beween ineres raes and deficis on foreign capial inflows (Cebula and Koch, 1994). Anoher is o assess direcly he effecs of deficis on ineres 3 See he references in Barh e al. (1991) and he overview aricle by he European Commission (2004). 5

Research Insiue of Applied Economics 2008 Working Papers 2008/10, 29 pages raes abroad. In (2c), we explain ineres raes in counry A by he crowding ou by deficis in counry A, and in addiion a direc crowding ou effec of deficis in counry B. i A, s A, s B,. (2c) The coefficien measures he size of he spillover effec. In order o esimae a specificaion similar o (2c), pleny of idenifying resricions are necessary ha severely reduce he dimension of he open economy model. Usually, only a paricular subse of counries can be examined, or idenical resricions are imposed on he ransmission across all counries in a panel. Cohen and Garnier (1991) find a posiive effec of US deficis on ineres raes in several G7 counries. Ardagna e al. (2007) find significan crowding ou effecs from boh domesic and foreign fiscal expansions in a panel of OECD counries. Marcellino (2002) or Giuliodori and Beesma (2005) consider he impac of shocks o German fiscal policy on he French and Ialian economy. Paesani e al. (2006) ake a somewha differen approach by idenifying spillover from shocks o bond markes on inernaionally linked capial markes. This allows hem also o consider he direcion of he spillover. An alernaive conrol ha models he linkages beween domesic and foreign bond markes, is o include he level of foreign ineres raes in (2b). i A, s A, i B,. (2d) Quie a few papers include foreign ineres raes in he analysis of crowding ou. Chinn and Frankel (2007) ake he German long erm rae as he benchmark in heir sudy of US fiscal policy. Caporale and Williams (2002) or Paesani e al. (2006) reduce heir sample o a few G7 economies and use in urn he ineres rae from he oher counry as a benchmark. The assumpion ha domesic ineres raes direcly depends on a single foreign benchmark rae, is raher srong. Ideally, one would like o conrol for he level of ineres raes in various counries. Mos papers consruc as he benchmark ineres rae an aggregae world ineres rae. Tanzi and Luz (1993) argue ha a he world level all spillover effecs should cancel ou and a significan crowding ou effec of fiscal policy resored. They aggregae all domesic deficis and examine he effec on global ineres raes. Ford and Laxon (1999) and De Haan and Kno (1995) do he same for OECD and EU counries respecively. Faini (2006) calculaes an average euro area ineres rae, and considers he fiscal effec on ineres raes a home and a EMU level conemporaneously in a panel framework. In empirical applicaions, modelling he ransmission across financial markes requires quie resricive assumpions. Assumpions like hese likely bias he direcion and he srengh of he spillover effec. We would expec ha on financial markes, spillover is a work beween differen markes conemporaneously. Spillover should also be sronger beween markes ha are more closely conneced. 6

Research Insiue of Applied Economics 2008 Working Papers 2008/10, 29 pages A convenien way o hink of hese complex linkages is wih an exogenously specified marix W ha specifies he srucure and inensiy of he closeness of differen observaions. The elemen w ij of W represens he proximiy beween wo observaions i and j. A common specificaion for his weigh marix W is physical coniguiy. Bordering regions are believed o have closer links. I is sraighforward o find oher W s ha reflec eiher economic disance beween counries, such as geographic disance, rade, level of economic developme culural or insiuional differences. A paern of spaial ineracion in a variable implies ha he disribuion of his variable across observaions is no random, and herefore he co-movemen of ineres raes on inegraed financial markes will bias he OLS esimaes of (2a). Wih spaial spillover, parameer esimaes are biased, inefficien and inconsisen (Anseli 1988). This bias may explain he mixed findings in he empirical lieraure esing crowding ou and spillover. By inroducing spaial lags (i.e., ineres raes in neighbouring counries) we direcly conrol for he ineracion wih he level of ineres raes in close by unis. We can rewrie (2d) more generally as a spaial auoregressive model: i s Wi (3a) In specificaion (3a) we conrol he crowding ou effec in counry n for he ineracion wih ineres raes in all neighbouring counries (he ermwi ). This is a weighed measure of ineres raes in n he counries wih which a counry has economic links. This spaial lag erm has o be reaed as an endogenous variable; equaion (3a) can be esimaed wih ML-echniques. A posiive (and significan) coefficien indicaes spillover. We can also recas he es of significance of as a es for he degree of financial marke inegraion: he larger, he more inegraed are financial markes. A significan spaial lag also reduces he bias in he esimae of he direc crowding ou effec. Financial inegraion is no he only facor behind he co-movemen of bond markes. Real economic inegraion affecs macroeconomic condiions globally, and here is quie some evidence for increased synchronisaion of business cycles (Doyle and Faus, 2002). Insead of a direc spillover on financial markes, he spillover could alernaively be due o a co-movemen of some omied economic variables, unrelaed o financial marke inegraio ha vary across space. There are wo possible ways o accoun for hese effecs of real economic inegraion. Firs, and as in equaion (2b), we proxy he economic co-movemen across counries by inroducing oher (domesic) macroeconomic variables. Alhough an OLS regression of (2b) sill yields unbiased esimaes, inference may be misleading since he precision of he esimaes is affeced. Alernaively, we can inroduce such spaial links in he residuals of he empirical model. Hence, 7

Research Insiue of Applied Economics 2008 Working Papers 2008/10, 29 pages we qualify model (2a) o include a spaial correlaion srucure of he error erm n. We hen rewrie (2a) as follows: i s W (3b) In his spaial error model, is a whie noise error erm. The parameer in (3b) is he spaial auoregressive parameer, and reflecs he spillover across counries due o economic inegraion. Financial and real economic inegraion probably work in he same direcion. Co-movemens in macroeconomic facors also drive he synchronisaion of financial markes. I is likely ha economies wih close economic (rade) links also have more ighly linked financial markes. As a consequence, he spillover effecs of financial markes and economic inegraion may be similar, and hard o disinguish. This observaional equivalence may cause he spaial lag and spaial error model o give very similar resuls (Kaminsky and Reinhar, 2000). 4 3.1 Specificaion We argue ha we can recover significan crowding ou effecs of fiscal policy on ineres raes if we use a spaial exension of he simple baseline model (2a) o analyse he spillover effec of fiscal policy. Hence, we depar from a model as (2a) and use spaial echniques o es and model he crowding ou and spillover effec of fiscal policy. We are ineresed in wo effecs in he spaial model. In firs insance, we es he crowding ou effec of domesic deficis on domesic ineres raes ( ). We expec ha a higher surplus (lower defici) has a significanly negaive effec on ineres raes. Secondly, he spaial model allows esing for he effec of ineres raes abroad. The more ineres raes rise in oher counries, he higher will be he domesic ineres rae as well. If ineres raes are very close, his suggess ha global credi markes are fairly inegraed. The pool of loanable funds any governmen draws from exceeds he available funds in he domesic credi marke only. We expec ha his spillover effec will be significa and posiive. The inerpreaion of he spillover effec is differen in he spaial lag or error model. Firs, in he spaial lag model, we look a he aggregae effec of all oher counries ineres raes on he home counry s ineres rae. We inerpre a significan spaial lag as evidence of a direc spillover of fiscal policy on financial markes. The spaial lag parameer is he slope of his reacion 4 Models (3a) and (3b) can be combined in a general specificaion like ha encompasses boh effecs. We do no es a general spaial model in his paper. 8

Research Insiue of Applied Economics 2008 Working Papers 2008/10, 29 pages funcio and measures he degree of financial inegraion. Posiive spaial correlaion in ineres raes exiss if 0, whereas here is evidence of negaive spaial correlaion if 0. 5 The economic inerpreaion of he model (3b) wih spaial links in he errors is slighly differen. A shock o he domesic ineres rae, which is no explained by fiscal policy, spills over o all oher observaions ha are close. Spaial correlaion in he error erm reflecs a similar reacion of counries ineres raes o shocks, because of omied variables ha are spaially correlaed. This indicaes imporan economic channels of spillover, bu is no relaed o fiscal policy per se. A posiive indicaes posiive spaial correlaion of he shocks; negaive shows ha shocks are of opposie sign. 3.2 Daa We esimae hese wo spaial models on a panel of 101 counries, for which we have annual daa on ineres raes and fiscal policy covering he period 1990-2005. A panel model allows combining he ypical analysis of domesic crowding ou in a ime series model, wih he spillover of ineres raes in he cross secion dimension. We ie boh dimensions and he srucure of linkages across counries, wih a weigh marix W ha reflecs geographical disance. 6 Counries ha are more disan are assumed o have weaker economic links. Geographic disance usually is a good proxy for economic linkages (as in he graviy model, for example). We show ha our resuls do no depend on his specific assumpio and we check our resuls for differen definiions of W. Spaial panel daa models have only recenly been developed, and no all heir properies have been examined. Our saring poin is he fixed effec panel model in which subsequenly spaial dependence is accouned for by including a spaially lagged erm of he dependen variable. This is he expression given in (3a), including a counry specific fixed effec. The sandard esimaion mehod for he fixed effec model is o eliminae he inercep erm of he regression by expressing all variables as a deviaion from heir ime average, and hen using sandard OLS esimaors. In presence of spaial auocorrelaion i is common pracice o use maximum likelihood mehods o esimae he demeaned equaion. These spaial auoregressive models are esimaed hrough he maximum likelihood mehod of esimaion developed by Elhors (2003). 7 5 A spaial es for financial marke inegraion could be equally applied o oher financial asses. We look a he spillover effecs of governmen bonds on ineres raes for wo reasons. Firs, spillover on governmen bond markes is policy relevan. In conras o privae bond issues, fiscal policy could inroduce disorions on governmen bond markes. I.e., here are no only pecuniary implicaions for oher domesic or foreign issuers. Second, we can include many counries in our sample as governmen bonds are he mos comparable asse across counries. They are usually he leas risky asse and are raded on he mos liquid marke. 6 We assign a cenre o each counry, and use is coordinaes o calculae he disance beween hese cenroids. We use a GIS sofware for hese calculaions. 7 Consisen esimaion of he individual fixed effecs is no possible as n grows large, due o he incidenal parameer problem. Anselin and Le Gallo (2008) argue ha "since spaial models rely on he asympoics in he cross-secional dimension o obain consisency and asympoic normaliy of esimaors, his would preclude he fixed effecs model from being exended wih a spaial lag. However, Anselin and Le Gallo (2008) show ha for consisen esimaes of, 9

Research Insiue of Applied Economics 2008 Working Papers 2008/10, 29 pages We use a nominal long erm ineres rae (5 o 10 years) as our dependen variable. Deficis are usually argued o affec long erm ineres raes. No ha many counries ouside he OECD have been able o issue long erm bonds, however. Mos emerging marke economies have financed deficis wih shor erm bonds a a 5 years horizon a mos. Similarly, fiscal daa for many counries are available only over recen years. The surplus o GDP raios all come from he IMF Governmen Saisics Daabase. As we prefer working wih balanced panels over he full sample period, we consequenly had o eliminae a large number of counries from he sudy. Due o variable daa qualiy, we also decided o remove some oulier observaions. We firs run a simple pooled esimae and quied he observaion if he sandard error exceeds hree imes he residual variance. Evenually, we have 496 observaions on 31 counries over 16 years. This keeps in he sample mosly OECD counries and a few emerging marke economies. 8 4 RESULTS FOR THE BASELINE MODEL 4.1 Spillover on financial markes: he spaial lag model We firs esimae a simple pooled OLS regression of (2a), wihou any furher resricions on he spaial effecs, o replicae he crowding ou effec of similar sudies. Table 1 reveals a very significan and quie srong crowding ou effec: a rise in he surplus of 1% of GDP leads o a 109 basis poins fall in he ineres rae. This resul coninues o hold even if we impose more srucure on he pooled model. A priori, we would prefer o use a fixed effecs esimaor. Firs, we include a specific group of counries. Even if we draw a number of counries from he global sample of economies, his draw is no random (Balagi, 2001). Second, he specificaion (2a) is raher basic, and we do no conrol for oher relevan deerminans of ineres raes. The seing of fiscal policy is raher heerogeneous across counries. As a consequence, he counry-specific effec is likely correlaed wih he explanaory variable (he surplus). The Hausmann es indicaes ha a fixed effecs esimaor is indeed preferable. I has also been more common in his lieraure o use simple pooled esimaes or panel fixed effecs esimaors (Frenkel and Chin 2005; Kinoshia, 2006). We repor boh, and find ha he panel esimaes of in he fixed or random effecs model are very similar (Table 1). Our esimae is on he high end of he range of esimaes found in he lieraure. In he overview sudy of he European Commission (2004), he crowding ou effec varies beween abou 20 and he demeaned spaial regressions from ML esimaion like in Elhors (2003) are appropriae. One complicaion wih his is ha he variance covariance marix of he demeaned error erm is differen from he usual one. Alernaive approaches o he Elhors esimaion are sill a opic of ongoing research. 8 The sample includes he following EU counries (Ausria, Belgium, Denmark, Finland, France, Germany, Hungary, Ireland, Ialy, Luxembourg, Neherlands, Porugal, Spai he UK), some oher OECD counries (Ausralia, Canada, Japa Korea, New Zealand, Swizerland, Turkey, he US) and emerging markes (Colombia, Lebano Mexico, Pakisa Peru, Philippines, Singapore, Souh Africa, Thailand). In he Appendix, we provide a deailed descripion of he daase and is sources. 10

Research Insiue of Applied Economics 2008 Working Papers 2008/10, 29 pages 100 basis poins. 9 For he US, he crowding ou effec is usually esimaed o be around 40 basis poins (Canzoneri e al., 2002; Laubach, 2003; Engen and Hubbard, 2004). This resul is also confirmed by VAR sudies on US daa, like Dai and Phillipon (2005). For EU counries, he crowding ou effec is mosly smaller in magniude (Faini, 2006). Wha explains he srong crowding ou effec in our esimaes hen? Mos oher papers examine he crowding ou effec of deficis in a single counry. In conras, panel sudies, like ours, have found much sronger effecs. Chinn and Frankel (2007) esimae a crowding ou of ineres raes beween 150 and 200 basis poins in a panel of he US and he larges EU counries. Similarly, Ardagna e al. (2007) use panel VAR echniques o look a he impac of deficis on ineres raes in a panel of OECD counries and find a rise of 150 basis poins afer 10 years. De Haan and Kno (1995) reach similar conclusions for he large EU counries. Hence, he inclusion of more counries in a crosssecion analysis of deficis and ineres raes ypically delivers sronger crowding ou effecs. We acually observe a similar effec in single counry sudies in which conrol variables for inernaional capial flows are included. Cebula and Koch (1994) find ha ineres raes rise by more han 60 basis poins afer a 1% increase in he defici raio, whereas capial flows reduce he effec by abou 24 poins. Chinn and Frankel (2007) find a sronger impac on raes, once foreign ineres raes are conrolled for. Tanzi and Luz (1993) aggregae daa for he G7 and find a rise in long erm raes of abou 150 basis poins. These resuls sugges ha a conrol for he spillover effec from oher counries is imporan. Omission of linkages on inernaional financial markes biases he findings of crowding ou. As he esimae of is quie likely biased, inefficien and inconsise we now inroduce he spaial exension. In he second panel of Table 1, we presen he esimaes of differen versions of he spaial lag model. The baseline esimae is he spaial lag model wih fixed effecs. We find ha he crowding ou effec halves in case a spaial lag is included: a defici of 1% of GDP pushes up ineres raes by 45 basis poins. The spillover effec is very significan and quie large: a 1% rise in ineres raes abroad also raises domesic raes by abou 0.55%. The consequence is ha an increase in he defici of 1% will cause domesic ineres raes o rise by 45 basis poins. Consequenly, he second round effec of he defici is o push up ineres raes abroad by a furher 25 ( 0.55%*45pp) basis poins. A governmen creaing a defici sill faces a quie seep increase in domesic raes, bu par of his increase spills over abroad. The crowding ou effec in he spaial lag model is more in line wih he resuls of he empirical sudies of single counries. This suggess ha he conrol for he spaial links indeed correcs he iniial panel esimaes. As regards our findings on spillover, i is slighly harder o compare is size. Mos sudies simply repor ha he domesic crowding ou effec is larger han he foreign spillover effec. Caporale and Williams (2002) find his resul for he US; and Faini (2006) repors similar resuls for he EU counries. Ardagna e al. (2007) repor ha he aggregae (world) defici affecs 9 See in paricular he overview able in European Commission (2004) on pp. 153-55. 11

Research Insiue of Applied Economics 2008 Working Papers 2008/10, 29 pages domesic ineres raes, bu is impac is less relevan han ha of domesic fiscal policy. In differen seings, oher sudies have found close connecions beween ineres raes across borders (Minford and Peel, 2007). Noneheless, counry-specific facors sill play a role in explaining he deviaion of domesic ineres raes from he evoluion in worldwide ineres raes (Breedon e al., 1999). One of he main reasons is a change in he fiscal policy sance. 10 Few sudies repor he impac ha inernaional capial flows or foreign ineres raes have on domesic ineres raes. Cebula and Koch (1994) find a similarly srong reducion in ineres raes (24 pp) as we do. The co-movemen of ineres raes may no jus reflec he inegraion of financial markes. Economic inegraion makes counries suscepible o global economic developmens. Trade, financial inegraion and similar economic srucures raise he co-movemen of business cycles inernaionally (Imbs, 2004). Economics shocks ha are common o a group of counries would display a close synchronisaion of economic variables. This migh show up in a significan spillover effec. We inroduce a ime period fixed effec in he spaial panel o absorb hese common shocks. We indeed find ha he spillover effec is much smaller in his case, whereas he crowding ou effec remains as srong (Table 1). 4.2 Financial and real economic inegraion An alernaive possibiliy is ha he co-movemens of economic variables are also spaially disribued. Anoher way o model hese economic links is o incorporae a spaial srucure in he residuals of he baseline model. The assumpion is ha hese economic facors, excep ineres raes, are spaially disribued across economies. We esimae his spaial error panel model (3b). By conrolling for hese spaial linkages, we pick up a significan crowding ou effec. Table 2 shows ha he resuls are very similar o hose of he spaial lag model. Moreover, he spillover effec causes a 1% rise in foreign raes o raise domesic raes by 56 basis poins. We can no idenify wheher spillover is due o eiher financial marke inegraio or he co-movemen of macroeconomic variables (Kaminsky and Reinhar, 2000). 4.3 Some conrol variables Alernaively, one may consider a correcion of he baseline model (2a) wih a spaial srucure for he errors oo naive. The facors ha deermine ineres raes are pleny of course, and he surplus is cerainly no he only deerminan of (long-erm) ineres raes. One ofen saed reason for he 10 Noe ha for oher asses han governmen bonds, mos empirical papers find similar resuls on he imporance of spillover. Ehrmann e al. (2005) find ha asse prices reac more srongly o domesic shocks, bu sill allows for a srong spillover beween he US and EU markes. 12

Research Insiue of Applied Economics 2008 Working Papers 2008/10, 29 pages ambiguous findings regarding crowding ou is he conemporaneous influence of moneary policy, auomaic fiscal sabilisers, ineres paymens on ousanding deb and any economic effecs of fiscal policy iself. 11 We es exensions of he spaial lag model ha conrol for hese addiional regressors X n,, as in (4) 12 i s Wi X (4) I is quie common in he empirical lieraure on crowding ou o direcly es he effec of public deb on (long erm) ineres raes, insead of using deficis. The argumen is ha public deb subsiues privae capial, and hence i he sock of deb ha has an impac on he level of ineres raes (Engen and Hubbard, 2004). Moreover, he iniial fiscal posiion of counries maers for crowding ou. Fiscal policy has non-linear effecs. A higher levels of deb, ineres raes ypically reac more srongly o higher deficis (Ardagna e al., 2007). In paricular, emerging marke economies sar paying a higher risk premium for fiscal indiscipline (Zoli, 2004). Table 3 repors he esimaes of he spaial panel lag model wih fixed effecs for a model augmened wih public deb. Conrolling for public deb gives an ineresing resul. The crowding ou effec of he surplus becomes less srong: ineres raes rise by a mere 16 basis poins afer a 1% rise in he defici. Ardagna e al. (2007) find a shor run effec of deficis of abou 10 basis poins, afer conrolling for deb. This effec accumulaes over ime o abou 100 basis poins, especially as he deb raio rises. The impac of deb albei significan - is very small. These resuls fall in a similar range as in he oher sudies. Single counry sudies find raher modes crowding ou effec of higher public deb. The consensus esimae ranges beween 2 and 7 basis poins for he US wih a variey of mehodologies (Ford and Laxo 1999; Canzoneri e al., 2002; Laubach, 2003; Engen and Hubbard, 2004). 13 As for he impac of he surplus on ineres raes, sudies ha conrol for links beween differen counries, or use a cross-secion approach, find slighly sronger effecs of deb. For example, he impac of deb is slighly sronger for he EU counries han in he US (Faini, 2006). Pooled esimaes for a group of OECD counries show a rise of abou 25 basis poins afer a rise in domesic deb (Ford and Laxo 1999; Orr e al., 1995). A similar effec is found by Tanzi and Luz (1993). 11 These effecs could cause some problems of endogeneiy in (2a), bu hese feedback effecs are likely small. IV esimaes are no considered in mos of he lieraure, however. Spaial panel models ha conrol for endogeneiy of he regressors have no been developed ye. 12 In he spaial economerics lieraure, he boom-up approach for searching an adequae specificaion prevails. The so-called Hendry approach is no common. Florax e al. (2003) demonsrae ha he specific-o-general approach slighly ouperforms he Hendry approach in he case of he esimaion of linear spaial models. 13 The only excepion is Friedman (2005), who finds ha a 1% rise in he deb raio increases ineres raes by 90 basis poins. 13

Research Insiue of Applied Economics 2008 Working Papers 2008/10, 29 pages Ineresingly, he spillover effec is much sronger and is esimaed very precisely. Three quarers of a 1% rise in he ineres raes spills over o close by counries. Afer all crowding ou and spillover effecs have worked ou, a 1% rise in he defici will push up ineres raes by 16 basis poins a home, and by 12 basis poins abroad. Long-erm ineres raes are very much influenced by moneary policy in he shor erm. 14 We conrol in wo differen ways for is effec. Firs, we include a shor-erm ineres rae in he specificaion. A shor horizons, moneary policy ses ineres raes o sabilise inflaion and oupu. Cenral bank decisions direcly influence he financing condiions of he governmen (and is ineres paymens on ousanding deb). The insignificance of he crowding ou effec confirms ha he shor run impac of a higher defici may be significan in raising ineres raes, bu i is no very imporan and i is blurred by he impac of moneary policy. Bu once a conrol for shor erm raes is included, he spaial lag coefficien is negaive. Such a negaive spillover effec can only be explained by a subsanial spaial ransmission of changes in shor erm ineres raes, which offse he co-movemen of long erm ineres raes beween neighbouring counries. Oher sudies also illusrae his co-movemen of shor erm raes across borders (Minford and Peel, 2007; Ehrmann e al., 2005). Second, we include also he inflaion rae. 15 Higher inflaion eases pressures on deficis as i erodes he real value of ousanding deb. We find ha he spillover is no really affeced by he spaial variaions in inflaion. 4.4 Time variaion in he crowding ou effec Financial and economic inegraion can explain why changes in asse markes have large effecs on oher financial markes. Globalisaion is ofen argued o have srenghened he spillover beween economies in wo differen ways. On he one hand, as inegraion is a gradual process, we are likely o observe a change over ime in he srengh of spillover. On he oher hand, here could be urbulence in he spillover channel due o financial or economic crises. Tranquil periods in which here is a normal degree of real and financial inerdependence suddenly swich o an environmen wih wild co-movemens during currency and financial crises (Claessens e al., 2001). Some auhors argue his disincion is only appare and inerdependence is deermined by real facors ha change only gradually over ime (Boyer e al., 1999; Forbes and Rigobo 2002). The resuls in Table 1 showed ha common shocks migh be more imporan in explaining inerdependencies across counries han a genuine spillover from oher economies. If ineres raes are indeed driven by some common facors in any given year, hen we would no expec o 14 Crowding ou is obscured by saic specificaion of he relaion beween deficis and ineres raes in (2a). The reason is ha governmen bonds are acually raded on financial markes. As financial markes are forward looking, i is he anicipaion of upcoming deficis, raher han he curren fiscal balance, ha resuls in higher long erm raes insanly. A few sudies include expecaions abou he defici or raings, and direcly analyse he effec of hese budge projecions on expeced ineres raes (Laubach, 2003). These daa are available for a limied sample only. Papers ha look ino he effec of defici announcemens by he governme or analyse he effec of deficis on risk premia usually ignore he spillover effecs of fiscal policy, wih he excepion of Kichen (1996). 14

Research Insiue of Applied Economics 2008 Working Papers 2008/10, 29 pages see a spillover effec in a year-by-year esimaion of he spaial lag model. All inerdependencies would be absorbed by he consan erm in his cross-secion model. Noe ha if spaial links are predominanly deermined by conagious crises across emerging economies, he annual frequency of fiscal daa may no pick up he high frequency movemens on financial markes due o sudden fiscal crises. We urn again o he sandard spaial lag model for explaining he variaion in ineres raes by fiscal variables bu esimae i a a cross-secional level for each year. Noe ha he efficiency of hese cross-secion esimaes is smaller han in he panel case. Figure 1 plos he coefficiens of an ML esimaion of he baseline regression over he sample 1990 o 2005. We have hree major resuls. Firs, here is a crowding ou effec of fiscal policy on ineres raes: a fall in he surplus (higher defici) raises ineres raes. Second, he spillover effec is no paricularly sable. The spaial lag coefficien varies in a raher large band beween 0 and 40 basis poins since he mid nineies, bu here are some srong drops in 1994 and 2004. Finally, boh effecs vary over ime. We can disinguish hree differen episodes. In he firs half of he nineies, fiscal policy has hardly go any crowding ou effecs. Foreign ineres raes end o go in he opposie direcion of domesic raes. In a second period, which goes from he mid nineies o he year 2000, crowding ou is significan and large. A he same ime, spillover becomes sronger as well. Saring in 1999, crowding ou and spillover boh become less pronounced. There is a gradual decline in he esimaed coefficiens and. These resuls are also corroboraed by he findings of a cross-secion esimaion of he spaial error model. These resuls each us some imporan lessons. Firs, if we compare hese findings wih our panel esimaes wih ime period effecs, we canno clearly aribue he smaller spillover effec o common shocks only. There is an imporan change in he crowding ou effec as well. I is no surprising ha spaial links are increasingly imporan in explaining he ransmission of ineres raes across borders. Increasing globalisaion is believed o have spurred capial mobiliy and increased rade flows. Linkages on inernaional markes have cerainly become much sronger han hey were a decade ago. Moreover, he nineies have seen several large crises ha have spread o oher counries. The 1994 Tequila crisis in Mexico was he firs big fiscal crash. The Asian Flu ha sared in 1997 in Thailand se off a series of problems in he Asian Tigers, bu spread globally. Russia defauled in 1998 afer Brazil had devalued he real a few monhs before. Argenina defauled in 2001 and Turkey experienced fiscal and moneary rouble in he same year. Since he no major emerging marke crisis has occurred. We find a break in spillover: here are no significan spaial links since 2001. In conras o he nineies, domesic crises have had much less impac abroad in recen years. There could be wo reasons for his. Firs, domesic crises are less serious now han hey were in he nineies. Second, even hough financial and economic 15 Daa on inflaion expecaions are no available for all counries in he sample. 15

Research Insiue of Applied Economics 2008 Working Papers 2008/10, 29 pages inegraion are progressing, conagion is now much weaker. Oher sudies have also found ha spillover has become much weaker in recen years. Forbes and Chinn (2004) also find evidence of sronger links over he period 1996-2000. Didier e al. (2006) show ha he co-movemen of emerging marke bond spreads and reurns have been declining since 2000. Mauro e al. (2006) presen similar resuls. The reasons for he changes in he crowding ou effec are unclear. Large inernaional crises can explain he large crowding ou effec in he mid nineies. In fac, some bu no all of he emerging marke crises sared wih domesic fiscal problems. High and rapidly growing public deb cas curren moneary policy sraegies ino doub, and mean high ineres raes o preven capial from fleeing he counry. This is only a parial explanaio however. For lack of daa, we have no been able o include many emerging markes in he sample. And even if hese economies in crisis had a global impac, he mere size of heir budge problems is probably oo small o affec ineres raes in indusrialised economies. Insead, fiscal policy in boh he US and he EU was much more focused on deb consolidaion in he nineies. The Clinon Adminisraion governed a 10% reducion in public deb in he span of five years, in par helped by he srongly booming economy. In addiio EU counries decided on a common fiscal rerenchmen and a sric moneary policy sance o prepare for EMU. EU counries had o abide by he rules of he Sabiliy and Growh Pac in order o qualify for he eurozone. Afer his join consolidaion effor, budge discipline has become less igh. I should no come as a surprise ha afer he enry in he EMU in 1999, crowding ou is much smaller. 5 SOME ROBUSTNESS CHECKS 5.1 Global or local linkages We immediaely pick up on he previous explanaion for he change in he crowding ou effec. Fiscal consolidaion in he EU counries migh indeed be responsible for he large crowding ou effec in he mid nineies. There are addiional reasons o expec a sronger spillover effec beween EU member saes over ime. Srong inerlinkages are he consequence of ongoing economic and financial inegraio and his mus have srenghened he spillover of economic policies beween hese counries. In addiio for hose EU counries paricipaing in moneary unio spillover may even be sronger. A common moneary policy has spurred financial inegraion and probably also rade links. If differen governmens borrow in he same currency, as in a moneary unio free riding makes each governmen disregard is own ineremporal budge consrain (Chari and Kehoe, 2007). A variey of reasons may be invoked for he lack of credibiliy of he no bailou clause ha prevens oher governmens (or he cenral bank) from rescuing he insolven governmen. The offseing ineres rae effecs do no need o maerialise he as 16

Research Insiue of Applied Economics 2008 Working Papers 2008/10, 29 pages defaul premia are spread ou over all members of he union. 16 In he absence of agreemens specifying he fiscal relaions beween governmens, he crowding ou effec depends ceeris paribus on he aggregae fiscal policy sance of all member saes. Could spaial links be sronger beween paricular groups of counries, or are capial markes ruly global? We run he same baseline model for some subsamples of counries. We are paricularly ineresed in he subgroup of EU counries. The resuls of he spaial ess mus be aken wih some cauion since we include only 13 EU counries. The properies of spaial panel ess are insead asympoically valid. Table 4 summarises he resuls of he differen specificaions of he spaial panel. Crowding ou is much less significan for an EU counry. In conras o he global sample, a 1% defici raises ineres raes by only 10 basis poins. Nowihsanding, he oal effec on ineres raes is much larger due o he spillover effec. Nearly 90% of an ineres rae rise is ransmied o oher EU counries. A 1% defici raising domesic raes by 10 poins will in a second sep cause a rise in foreign raes of abou 9 basis poins. Hence, deficis will raise ineres raes by nearly he same amoun a home as in anoher EU counry. As before, he panel wih fixed or random effecs gives very similar resuls. There is again some evidence ha common shocks are driving ineres raes. The resuls of he spaial panel model are somehow alered when accouning for ime period fixed effecs. The spillover effec is much weaker, and he crowding ou effec is compleely absen. As regards he source of he spillover, here is no much evidence o idenify he role of financial or economic inegraion in he ransmission of ineres raes across EU counries. The esimaes of he spaial error model show an imporan spillover effec, bu no crowding ou. Mos sudies have examined spillover beween OECD counries. We look a he indusrialised economies, bu exclude he mos recenly acceded member saes. Our baseline resuls for he global sample are no much affeced: crowding ou effecs are significan and spaial links are raher large. The esimaes are of he same order as for he global sample. 5.2 Differen weigh marices For all previous resuls, we have used a measure of geographical disance as a proxy for cross counry economic linkages. We check if he resuls are robus o oher definiions of he weigh marix W, and focus on he spaial panel lag model wih fixed effecs. 17 We firs ry ou some differen measures of disance. We alernaively measure he (invered) disance beween counries as he disance beween capial ciies, or he grea circle disance beween counry 16 Yardsick comparisons across governmens may parially undo his spillover, if he accumulaion of deb by one governmen increases he relaive crediworhiness of comparable governmens. 17 This resul is robus for he oher panel models. 17

Research Insiue of Applied Economics 2008 Working Papers 2008/10, 29 pages cenroids. 18 Table 5 shows ha he poin esimaes are very similar, and so is he significance of boh effecs. A more common choice of he weighing marix in spaial sudies is physical coniguiy. Counries ha share a common border are believed o ransmi effecs o heir direc neighbours, bu no effec a all o far-away counries. Under his ype of ransmission mechanism, crowding ou is only marginally sronger, bu he spaial effec is negaive. The reason is ha border links are an awkward choice, as here are pleny of missing observaions in our sample. Only a few European counries share a common border, bu mos oher economies are isolaed from each oher (i.e., here are many zeros in he weighing marix). This downplays he imporance of spaial ransmission. Physical disance is a bes a proxy for he inegraion of counries financial markes, bu sill gives lile economic conen o being close. Our esimaes of he spillover effec could be quie conservaive as a consequence. We experimen wih some more economic weigh marices. I is ofen argued ha rade is a major channel for economic ransmission across counries. We herefore use differen possible weigh marices incorporaing bilaeral expors and impors. We scale oal expors from counry i o counry j by oal expors of counry i. 19 Similarly, for impors of counry j we scale by oal impors of counry j. 20 We also weigh by oal rade, summing bilaeral expors and impors, and dividing by oal rade of he counry. As a consequence, hese weigh marices are asymmeric: he srengh of he ransmission depends on he size and imporance of each counry. For example, he US may rade a lo wih Colombia, ye he imporance of his rade volume for he US economy is iny. In conras, for Colombia, US rade is much more imporan. We would expec he spillover from he US o Colombia o be much sronger han vice versa. This weigh marix beer reflecs he srengh of ransmission from large o small economies. Surprisingly, none of he resuls of he baseline model is alered very much. The crowding ou effec is as large as before, and so is he spaial effec. Keleijan e al. (2006) similarly find lile differences beween he use of rade or disance marices in heir analysis of financial marke spillover. This resul confirms ha disance is a good proxy for rade and economic relaions in a graviy model. One migh argue ha rade is endogenous o he srengh of he economic links. We choose an alernaive weigh marix ha has a dummy if wo counries are in a free rade agreemen. This resuls in a slighly sronger crowding ou effec, and weaker spaial links. However, he resuls could be biased. There are a few counries only ha do no have some kind of bilaeral rade agreemen in our sample. As a consequence, he imporance of spaial links is probably undersaed. 18 A grea circle is he shores pah beween wo poins along he surface of a sphere. 19 All daa are in USD, rade daa are FOB or CIF. Spaial panel models canno handle ime varying weigh marices. We arbirarily fix expors and impors a a base year in he mid-of-sample (1998). Two counries are close if hey have srong bilaeral rade (relaive o he oher rading parners). In conras o he lieraure on conagio we do no use he compeiion for expor shares on hird markes (Forbes and Chin 2004). 20 The wo numbers do no mach for saisical reasons. This is known as he missing rade problem. 18

Research Insiue of Applied Economics 2008 Working Papers 2008/10, 29 pages The panel model provides an average effec of fiscal policy on ineres raes, while arguably hese crowding ou and spillover effecs may differ across counries. Changes in fiscal policy in he large indusrialised economies are likely o have a larger effec on smaller economies. The ransmission of economic evens is likely o run in one direcion. For example, measured by grea circle disance, Germany is equally disan from France and Hungary. The impac of changes in he German economy is likely o be large for boh counries. Ye, he inverse impac of he French economy on Germany is almos cerainly much larger han ha of he Hungarian economy. We conrol for he direcion of spillover and he imporance of ransmission beween economies by muliplying counry size (GDP in USD PPP erms) by physical disance. Noneheless, he resuls do no change much if we use his asymmeric weigh marix. Boh indusrialised and emerging marke economies are increasingly open o financial markes. Financial inegraion beween indusrialised economies is gradually proceeding wih economic inegraion. Insead, emerging marke economies could be subjec o conagious crises ha spread from a crisis in anoher emerging marke, bu are unrelaed o he economic fundamenals (and in paricular he fiscal posiion) of he counry iself. Economic crises may spread faser beween emerging markes ha are more exposed on financial markes, have similar macroeconomic characerisics or are prone o informaion asymmeries ha rigger sunspo crises. As a final robusness check, we ry o model hese various channels of conagion. We capure he heerogeneiy beween indusrialised and emerging economies by he difference in economic developmen. We use a weigh marix in which spaial links are sronger if he difference of (log) per capia income (in PPP USD) is smaller. We do no find significan differences in he crowding ou effec, and he spaial effecs remain as srong as wih he oher weigh marices. Our weigh marix is a raher rough aemp o disinguish links beween indusrialised and emerging marke economies. We have no aemped o model hese channels of conagion wih alernaive definiions of he weigh marix. 21 I has been argued ha hese alernaive channels may be less imporan han he classical ransmission channels, such as rade (see he findings of Gerlach and Smes, 1995). Eichengreen and Rose (2001) and Forbes and Chinn (2004) also find ha real and financial linkages are predominanly deermined by real rade inegraion. Kaminsky and Reinhar (2000) argue ha rade or financial links are hard o disinguish, and hence ofen boh specificaions give very similar resuls. 5.3 Alernaive daa definiions 21 Alernaive definiions could be: ineres rae spread, sock marke index, exchange rae regime, real exchange rae, compeiion for bank lending, or inernaional reserves. Daa on hese oher channels are unforunaely no compleely available for all counries in our sample. 19

Research Insiue of Applied Economics 2008 Working Papers 2008/10, 29 pages Fiscal policy is argued o affec ineres raes a long horizons. This can be esed in mos indusrialised economies ha are able o issue long erm bonds a a horizon of en or more years. No ha many counries ouside he OECD have been able o issue long erm bonds, however. Mos developing economies can only ge finance on capial markes a shor horizons, and have financed deficis wih shor erm bonds a a 5 years horizon a mos. We add o he sample hose counries ha issue governmen bonds a a horizon shorer han five years, and also use a shoro medium erm ineres rae (of 1 o 5 years) as our dependen variable for he counries in he iniial sample. Table 6 gives he resuls of he spaial panel lag model wih fixed effecs. For he baseline model, he resuls barely change. If we augmen he model wih some addiional conrol variables, he resuls sill do no change very much. This is surprising, given ha wih a few excepions (Barh e al., 1984), he lieraure usually does no find effecs of deficis on shor erm raes (Cebula, 2000). We also add several conrol variables o his specificaio and confirm he resuls of secion 4.3. The crowding ou effec is weaker when we add public deb, bu he spillover effec becomes much sronger. If we add inflaio only he spillover effec is imporan. When daa are available, we also add as a conrol he long erm rae. In his case, boh he crowding ou and spillover effec disappear. The level of ineres raes is deermined by many oher facors han fiscal policy. As argued in secion 4.3, i is he level of public deb ha deermines crowding ou. The addiion of new sock of public deb should insead pu addiional pressure on he change in ineres raes. Hence, insead of using he level of ineres raes as he dependen variable, one should use he firs difference insead. Table 6 shows ha his does no affec he esimaes of he crowding ou effec, nor of he spillover. The addiion of conrol variables does no change our conclusions. Finally, some oher sudies have used he yield o filer ou any of he shor erm effecs of fiscal or moneary policy, and analyse he impac of expeced deficis on ineres raes. The argumen is ha higher fuure deficis ranslae ino higher ineres raes in he fuure, and hence, a seepening of he yield curve. By using he yield as he dependen variable, here is no need o model he facors ha drive he level of ineres raes. We subrac from he long erm ineres rae he shor erm rae, and use his yield as a dependen variable. The esimaion of he baseline model does no show very differen resuls, even if we add he usual conrol variables. 6 CONCLUSIONS There is much discussion abou he effec of fiscal expansions on ineres raes. This variey in opinion is due o he lile robus empirical endorsemen for crowding ou. A lack of response of ineres raes can be jusified under wo differen heoreical condiions. Firs, under Ricardian Equivalence, deficis do no affec macroeconomic variables as economic agens anicipae he paydown of higher deficis wih fuure axes. Second, capial flows beween economically 20