Fiscal Limits, External Debt, and Fiscal Policy in Developing Countries



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

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

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

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

BALANCE OF PAYMENTS. First quarter Balance of payments

Economics Honors Exam 2008 Solutions Question 5

Why Did the Demand for Cash Decrease Recently in Korea?

4. International Parity Conditions

PROFIT TEST MODELLING IN LIFE ASSURANCE USING SPREADSHEETS PART ONE

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

How To Calculate Price Elasiciy Per Capia Per Capi

LEASING VERSUSBUYING

CRISES AND THE FLEXIBLE PRICE MONETARY MODEL. Sarantis Kalyvitis

Working Paper No Net Intergenerational Transfers from an Increase in Social Security Benefits

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

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.

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

Chapter 6: Business Valuation (Income Approach)

Risk Modelling of Collateralised Lending

Individual Health Insurance April 30, 2008 Pages

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

The Grantor Retained Annuity Trust (GRAT)

NBER WORKING PAPER SERIES IS THE U.S. CURRENT ACCOUNT DEFICIT SUSTAINABLE? AND IF NOT, HOW COSTLY IS ADJUSTMENT LIKELY TO BE?

DYNAMIC MODELS FOR VALUATION OF WRONGFUL DEATH PAYMENTS

MSCI Index Calculation Methodology

Interest Rates and the Market For New Light Vehicles

Chapter 9 Bond Prices and Yield

Morningstar Investor Return

Debt Relief and Fiscal Sustainability for HIPCs *

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

Measuring macroeconomic volatility Applications to export revenue data,

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

Working Paper Monetary aggregates, financial intermediate and the business cycle

Mika Kortelainen. Adjustment of the US current account deficit

A One-Sector Neoclassical Growth Model with Endogenous Retirement. By Kiminori Matsuyama. Final Manuscript. Abstract

The Real Exchange Rate, the Current Account, and Foreign Debt in a Stochastic Optimizing Model with Imperfect Capital Flows. Jafar A.

Capital Flows and Current Account Sustainability: The Ghanaian Experience

Real exchange rate variability in a two-country business cycle model

I. Basic Concepts (Ch. 1-4)

Fiscal consolidation in an open economy with sovereign premia

Optimal Investment and Consumption Decision of Family with Life Insurance

Premium Income of Indian Life Insurance Industry

Information technology and economic growth in Canada and the U.S.

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

Working paper No.3 Cyclically adjusting the public finances

USE OF EDUCATION TECHNOLOGY IN ENGLISH CLASSES

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

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

Estimating Time-Varying Equity Risk Premium The Japanese Stock Market

Fiscal Consolidation in an Open Economy

Working Paper Capital Mobility, Consumption Substitutability, and the Effectiveness of Monetary Policy in Open Economies

TREASURY WORKING PAPER 01/32

The Interest Rate Risk of Mortgage Loan Portfolio of Banks

What drives the German current account? And how does it affect other EU member states? (*)

Chapter 1.6 Financial Management

Price Controls and Banking in Emissions Trading: An Experimental Evaluation

EUROPEAN ECONOMY. What drives the German current account? And how does it affect other EU member states?

UNDERSTANDING THE DEATH BENEFIT SWITCH OPTION IN UNIVERSAL LIFE POLICIES. Nadine Gatzert

Lecture Note on the Real Exchange Rate

Chapter 8: Regression with Lagged Explanatory Variables

The Effect of Public Expenditure Shocks on Macroeconomic Variables in a Real Business Cycle Model. Case Study: Iran

Appendix D Flexibility Factor/Margin of Choice Desktop Research

CURRENT ACCOUNTS IN THE EURO AREA: AN INTERTEMPORAL APPROACH. José Manuel Campa and Ángel Gavilán an. Documentos de Trabajo N.

Public finance: structural deficit, sensitivity and long-term sustainability

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

Does International Trade Stabilize Exchange Rate Volatility?

Chapter 10 Social Security 1

Present Value Methodology

Working Paper. Income inequality and Germany s current account surplus. February Patrick Grüning 1, Thomas Theobald 2 and Till van Treeck 3

Can Austerity Be Self-defeating?

Finance and Economics Discussion Series Divisions of Research & Statistics and Monetary Affairs Federal Reserve Board, Washington, D.C.

ARCH Proceedings

Real Business Cycles Theory

Capital Budgeting and Initial Cash Outlay (ICO) Uncertainty

Energy prices and business cycles: Lessons from a simulated small open economy. Torsten Schmidt, RWI Essen * Tobias Zimmermann, RWI Essen *

INTEREST RATE FUTURES AND THEIR OPTIONS: SOME PRICING APPROACHES

DEMAND FORECASTING MODELS

The Interaction of Public and Private Insurance: Medicaid and the Long-Term Care Insurance Market

Depreciation and Corporate Taxes

Research. Michigan. Center. Retirement. Behavioral Effects of Social Security Policies on Benefit Claiming, Retirement and Saving.

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

Financial Globalization and Exchange Rates. Philip R. Lane IIIS, Trinity College Dublin and CEPR

NBER WORKING PAPER SERIES INTERGENERATIONAL TRANSFER, HUMAN CAPITAL AND LONG-TERM GROWTH IN CHINA UNDER THE ONE CHILD POLICY

The Interaction of Guarantees, Surplus Distribution, and Asset Allocation in With Profit Life Insurance Policies

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

Debt and Fiscal Sustainability in Sri Lanka

SUBJECT SA0 OF THE INSTITUTE AND FACULTY OF ACTUARIES

2009 / 2 Review of Business and Economics. Federico Etro 1

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

Hiring as Investment Behavior

Investigation of the effect of the degree of openness of the economy on real effective exchange rate Volatility: case study (the Iran economy)

Analysis of Pricing and Efficiency Control Strategy between Internet Retailer and Conventional Retailer

Tax Externalities of Equity Mutual Funds

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

Dynamic Hybrid Products in Life Insurance: Assessing the Policyholders Viewpoint

The real interest rate gap as an inflation indicator

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

Is China Over-Investing and Does it Matter?

MIKLÓS SZABÓ: GENERATIONAL ACCOUNTING FOR HUNGARY

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

Transcription:

WP/14/49 Fiscal Limis, Exernal Deb, and Fiscal Policy in Developing Counries Huixin Bi,Wenyi Shen, and Shu-Chun S. Yang

2014 Inernaional Moneary Fund WP/14/49 IMF Working Paper Research Deparmen and Sraegy, Policy, and Review Deparmen Fiscal Limis, Exernal Deb, and Fiscal Policy in Developing Counries * Prepared by Huixin Bi,Wenyi Shen, and Shu-Chun S. Yang Auhorized for disribuion by Andrew Berg and Caherine Paillo March 2014 This Working Paper should no be repored as represening he views of he IMF or DFID. The views expressed in his Working Paper are hose of he auhor(s) and do no necessarily represen hose of he IMF, IMF policy, or DFID. Working Papers describe research in progress by he auhor(s) and are published o elici commens and o furher debae. Absrac This paper sudies fiscal policy effecs in developing counries wih exernal deb and sovereign defaul risks. Sae-dependen disribuions of fiscal limis are simulaed based on macroeconomic uncerainy and fiscal policy specificaions. The analysis shows ha expeced fuure revenue plays an imporan role in he low fiscal limis of developing counries, relaive o hose of developed counries. Exernal deb carries addiional risks since large devaluaion of he real exchange rae can suddenly raise defaul probabiliies. Consisen wih majoriy views, fiscal consolidaions are counerproducive in he shor and medium runs. When an economy approaches is fiscal limis, governmen spending can be less expansionary han in a low-deb sae. As more revenue is required o service deb in a high-deb sae, higher ax raes raise he economic cos of increasing consumpion, reducing he fiscal muliplier. JEL Classificaion Numbers: E62; H30; H60 Keywords: fiscal limis; fiscal policy; sovereign defaul risk; exernal deb; developing counries; sae-dependen fiscal muliplier Auhor s E-Mail Address: hbi@bankofcanada.ca, wenyi.shen@oksae.edu, syang@imf.org * We hank Andrew Berg, Valerie Cerra, Fei Han, Eric Leeper, Nicolas Magud, Caherine Paillo, Guido Traficane, Andreas Tudyka, Todd Walker, and paricipaes in various seminars for helpful commens. This paper is par of a research projec on macroeconomic policy in low-income counries suppored by U.K. s Deparmen for Inernaional Developmen. This paper should no be repored as represening he views of he Bank of Canada, of he IMF, or of DFID.

Conens Page I. Inroducion... 3 II. Model... 5 A. Households... 6 B. Firms... 7 C. Governmen... 8 D. Aggregaion and Marke Clearing... 9 III. Esimaion and Calibraion...10 IV. Fiscal Limi Disribuion...11 A. Simulaing Fiscal Limi Disribuion...12 B. Revenue Collecion Capaciy...13 C. Devaluaion and Balance Shee Effecs...13 V. Fiscal Policy in a Highly Indebed Economy...14 A. The Economy in a High-Deb Sae...14 B. Fiscal Consolidaion...15 C. Governmen Spending Effecs in Differen Saes of Deb...16 VI. Conclusion...18 VII. References...32 Appendices I. Equilibrium Condiions...20 II. Bayesian Esimaion...23 III. Simulaing Fiscal Limi Disribuions...23 IV. Solving Nonlinear Model...25 Tables 1. Parameer Calibraion...27 2. Cumulaive Mulipliers of Governmen Consumpion...27 Figures 1. Esimaed CDF of he baseline fiscal limi disribuions for Argenina: uncondiional disribuions...28 2. Esimaed CDF for differen revenue mobilizaion capaciy for Argenina....28 3. Esimaed CDF of wih more flucuaions in he real exchange rae...29 4. Esimaed CDF of large devaluaion of he real exchange rae...29 5. Transiion dynamics of fiscal consolidaion...30 6. Esimaed CDF of a high-deb condiional disribuion...30 7. Impulse responses o governmen spending shocks in a high-deb sae...31 8. Impulse responses o governmen spending shocks in a low-deb sae...31

3 I. INTRODUCTION Sovereign deb is generally perceived riskier in developing han developed counries wih he excepion of he recen European deb crisis. Developing counries wih relaively low deb-o-gdp raios (by developed counries sandard) can have much lower credi raings han developed counries wih higher deb raios. 1 For example, Belgium, Unied Kingdom, and he Unied Saes all have ne governmen deb-o-gdp raios exceeding 0.8 in 2012 and sovereign raings a or above AA. A he same ime, Argenina and Ecuador have much lower deb-o-gdp raios (wih he gross deb-o-gdp raios a 0.48 and 0.22 in 2012, respecively), bu he raings are only B and B (Sandard & Poor s (2013)). 2 Fiscal limis defined as he maximum deb level a governmen is able and willing o service are generally lower in developing han developed counries. Using a dynamic sochasic general equilibrium (DSGE) model wih sovereign defaul risks, his paper sudies imporan facors ha shape fiscal limi disribuions of developing counries. I also analyzes fiscal policy effecs agains a backdrop of differen governmen indebedness. Fiscal limis in he model (and mos likely in realiy) are uncerain and forward-looking. Since sovereign defaul is ulimaely a poliical decision, which may or may no be grounded in economic raionales, our approach absracs from he complicaed facors underlying defaul decisions. 3 Insead, we assume ha wheher a governmen defauls each period depends on if he curren deb exceeds an effecive fiscal limi realized a ha period, drawn from a disribuion simulaed based on economic fundamenals. Because fiscal limis are based on he expeced discouned sum of maximum primary surplus ha can be generaed in he fuure, our approach emphasizes repaymen abiliy in sovereign defauls, as in Bi (2012) and Juessen e al. (2012). 4 Sovereign risk premia in our model arise endogenously and nonlinearly as a funcion of governmen indebedness as observed in daa. 5 1 The relaive high risk of sovereign deb in developing counries has been recognized, e.g., Alvarado e al. (2004), Hausmann (2004), and Reinhar e al. (2003). 2 The ne and gross deb o GDP raios are aken from he World Economic Oulook Daabase of he Inernaional Moneary Fund. 3 Borenszein and Panizza (2009) find ha, among various coss considered o make defaul decisions, economic coss (losing access o inernaional capial markes, rade exclusion, and disurbances hrough financial sysems) are shor-lived and he poliical cos is high. 4 Our approach o modeling defaul differs from he sraegic sovereign defaul lieraure, in which a uiliarian governmen accouns for some economic coss in making defaul decisions, e.g., Aguiar and Gopinah (2006), Arellano (2008), Derasmo and Mendoza (2012), Eaon and Gersoviz (1981), Mendoza and Yue (2012), and Yue (2010). Our model reains he DSGE framework convenien for incorporaing several economic and policy shocks and conducing fiscal experimens. 5 Using a sample of 26 emerging markes, Belhocine and Dell Erba (2013) find ha he sensiiviy of sovereign risk premia o he difference beween primary balances and deb sabilizing balances doubles as public deb increases above 45 percen of GDP.

4 The paper consiss of wo pars. The firs par simulaes fiscal limi disribuions for Argenina o demonsrae how our framework can assess fiscal limis, as well as o explain imporan facors affecing he disribuions. The second par sudies fiscal policy effecs in differen saes of deb. To make empirical relevance of simulaed disribuions, he shock processes are esimaed, fiing he linearized model (assuming no sovereign defaul) using Argenina s pos-defaul daa from early 2000s. When simulaing fiscal limi disribuions and sudying fiscal policy effecs, he model is solved nonlinearly by he monoone mapping mehod. Our resuls highligh wo facors imporan in explaining he relaively low fiscal limis in developing counries. Revenue collecion characerized by he maximum effecive ax rae can be implemened and he poliical risk facor in he model plays an imporan role in he level of fiscal limis. The lieraure has recognized ha a smaller ax base conribues o higher sovereign defaul risks in developing counries (see Hausmann (2004), and Mendoza and Oviedo (2004)). Due o inefficien ax collecion sysems, ax evasion, and large informal secors, developing counries on average have much lower effecive ax raes han developed counries (Inernaional Moneary Fund (2011)). Callen e al. (2003) esimae ha he effecive ax rae for emerging markes ouside easern Europe is only 10 percen, much lower han he average of indusrial counries, which is above 30 percen. Anoher facor is real exchange rae flucuaions, which increase he dispersion of a fiscal limi disribuion. As many developing counries rely on exernal borrowing o a large exen, a subsanial devaluaion elevaes defaul risks hrough he balance-shee effec. Among he explanaions for relaively high risks of sovereign deb in developing counries, Eichengreen e al. (2003) emphasize a counry s inabiliy o borrow in is own currency, he so-called original sin. From foreign crediors prospecive, flucuaions in real exchange raes increase he uncerainy associaed wih a counry s abiliy o repay is deb. Since revenues a governmen can collec are mosly denominaed in local currency, he problem of currency mismach adds addiional risk for a given size of deb (Bordo e al. (2006), Cespedes e al. (2004), and Krugman (1999)). In explaining Argenina s 2001 defaul, Calvo e al. (2004b) argue ha seep real depreciaion led by sudden sops urned an oherwise susainable fiscal posiion ino an unsusainable one in an economy wih heavily dollarized liabiliies. 6 When sudying fiscal policy effecs, we focus on fiscal consolidaion and governmen spending effecs in a high-deb sae. Consisen wih majoriy views, faser consolidaions reurn he risk premium o he seady-sae level more quickly, bu hey are more counerproducive in erms of reduced oupu han slower ones. Upon implemening a fiscal consolidaion hrough income ax hikes, households face higher curren and fuure ax raes. Higher curren axes suppress consumpion and labor. Moreover, expecing higher fuure ax raes discourages curren invesmen. 6 Anoher imporan facor explains he relaively high risk of sovereign deb in developing counries is deb inolerance resuling from poor credibiliy and a defaul hisory, as emphasized in Reinhar e al. (2003). Our analysis overlooks his facor.

5 Nex, we invesigae how governmen indebedness affecs spending effecs. To have he model-implied effecs of governmen consumpion in line wih mos empirical evidence, governmen consumpion in our model eners he households uiliy funcion as a complemen o capure is shor-run simulaive effec. 7 We find ha fiscal mulipliers are smaller when an economy is near is fiscal limis, alhough he difference is moderae. In our hough experimen, a high-deb sae is associaed wih a higher income ax rae because he governmen requires addiional resources o cope wih higher deb service paymen. Since he economic coss of raising consumpion in erms of leisure foregone increase (due o he lower afer-ax wage rae), governmen consumpion becomes less expansionary in a high-deb sae han in a low-deb sae. Our paper is relaed o several sudies ha assess fiscal susainabiliy. Celasun e al. (2007) propose a fan-char algorihm o simulae deb disribuions based on an empirical framework ha capures ineracions of deb dynamics wih macroeconomic shocks. The disribuions are hen used o assess deb susainabiliy by a somewha arbirary indicaor. Moivaed by Bohn (1998, 2008), Ghosh e al. (2011) and Osry e al. (2010) esimae fiscal space (he disance beween fiscal limis and curren deb level) in developed counries based on hisorical fiscal reacions o deb, wihou explicily modeling specific shocks ha can affec primary surplus. The approach is less informaive abou how a paricular policy or economic shock can affec sovereign risk premia and defaul probabiliies. Our fully-specified macroeconomic model allows us o sudy he ineracions beween he various shocks and fiscal limis and accouns for he sovereign risk channel of fiscal policy effecs. Similar o our srucural approach, Buffie e al. (2012) and Mendoza and Oviedo (2004) assess fiscal susainabiliy in a general equilibrium model. Mendoza and Oviedo inroduce he naural deb limi, capuring he maximum deb level ha a governmen remains able o fully service, bu he ineres rae in heir analysis is fixed. Buffie e al. (2012), insead, consider an exogenous risk premium bu do no allow sovereign defaul, similar o Corsei e al. (2013), García-Cicco e al. (2010), and Uribe and Yue (2006). Our model consrucs a general equilibrium framework ha endogenizes risk premia. II. MODEL The model is a small open, real economy wih wo producion secors for nonradables and radables (denoed by N and T, respecively). As one of our ineress is o see how macroeconomic uncerainy affecs he disribuion of he fiscal limi, he model feaures 7 Mos empirical evidence finds posiive governmen spending effecs on consumpion (e.g., Blanchard and Peroi (2002) and Peroi (2008)), bu Ramey (2011) finds he opposie for miliary spending. For developing counries, Ilzezki e al. (2013) find a posiive consumpion response o a governmen spending increase when he exchange rae regime is fixed. In addiion o complemenariy beween privae and governmen consumpion, oher heoreical explanaions, such as liquidiy-consrained households (Galí e al. (2007)) and deep habis (Zubairy (forhcoming)), have been proposed o generae co-movemen beween privae and governmen consumpion.

6 several imporan shocks ha drive business cycles in developing counries, including shocks o oal facor produciviy (TFP), spending and ax policy, and erms of rade. A. Households Households derive uiliy from effecive consumpion ( c ) and leisure (1 l ). Following Bouakez and Rebei (2007), effecive consumpion is assumed o be a consan-elasiciy-of-subsiuion (CES) index of privae consumpion (c ) and governmen consumpion (g ): c = [ω (c ) ν 1 ν ] + (1 ω)(g ) ν 1 ν ν 1 ν, (1) where ω is he weigh of privae consumpion in effecive consumpion, and ν > 0 is he elasiciy of subsiuion beween privae and governmen consumpion. When ν = 0 (ν ), c and g are perfec complemens (subsiues). 8 A represenaive household chooses privae consumpion (c ), labor (l ), and invesmen and capial in he wo secors (i N, k N, i T, k T ) o maximize he expeced uiliy over an infinie horizon [ E β log c φ l 1+σ ], (2) 1 + σ =0 } {{ } U subjec o he budge consrain c + i N + i T + κ ( ) i N 2 δ k N 2 k 1 N 1 + κ ( ) i T 2 δ k T 2 k 1 T 1 =(1 τ ) ( w l + r N k 1 N + r T k 1) T + z. (3) β (0, 1) is he discoun facor. σ is he inverse of he Frisch elasiciy for labor supply. Capial is secor specific, and r N and r T are reurns o capial in each secor. τ is he income ax rae, and z is governmen ransfers o households. 9 Invesmen is subjec o adjusmen coss wih he adjusmen parameer κ. The law of moion for capial is Aggregae invesmen is i = i N + i T. k j = (1 δ)k j 1 + ij, j {N, T }. (4) 8 Bailey (1971) is he firs o consider he relaionship beween he uiliy derived from privae consumpion and publicly provided goods and services. Subsequen papers allow governmen spending o affec household preference in sudying fiscal policy effecs, e.g., Barro (1981), Bilbiie (2011), and Finn (1998). 9 Governmen ransfers are kep consan hroughou he analysis. I is calibraed o close he governmen budge in he seady sae for a given governmen consumpion-o-gdp raio, he income ax rae, and he exernal deb-o-gdp raio from sample averages.

7 Privae consumpion and invesmen are CES aggregaes of nonradables and radables wih he inra-emporal elasiciy of subsiuion χ and he degree of home bias ϕ. Thus, x = [ ( ) ϕ 1 χ 1 χ x N χ + (1 ϕ) 1 χ ( ) χ 1] χ χ 1 x T χ, x { c, i N, } it. (5) Households supply labor o boh secors. Aggregae labor is l = [ (ϕ l ) 1 χ l ( l N ) 1+χ l χ l + (1 ϕ l ) 1 χ l ( l T ) 1+χ l χ l ] χ l 1+χ l, (6) where ϕ l is he seady-sae share of labor in he nonraded good secor. While capial is specific o each secor, we allow some labor mobiliy across secors, and χ l > 0 is he elasiciy of subsiuion beween secors. From he cos minimizaion problem, he aggregae wage index can be derived as w = [ ϕ l ( w N ) 1+χ l + (1 ϕ l ) ( w T ) 1+χ l ] 1 1+χ l. (7) We normalize he price of composie consumpion (or one uni of local goods) o 1. Le p N be he relaive price of nonradables o composie consumpion, and s be he relaive price of radables. Then, 1 = [ ϕ(p N ) 1 χ + (1 ϕ)(s ) 1 χ] 1 1 χ. (8) s is also he real exchange rae. B. Firms Firms in boh secors are perfecly compeiive, producing by Cobb-Douglas echnology, y j = a j ( ) k j 1 α j ( 1 l j α j ), j {N, T }. (9) a j is TFP of each secor, subjec o he common echnology shock ε a : ln aj a j = ρ a ln aj 1 a j + ε a, (10) where ε a N(0, σ 2 a). Variables wihou a ime subscrip indicae heir seady-sae values. A each period, a represenaive nonradable firm chooses labor and capial o maximize he profi p N y N w N l N r N k N 1. Similarly, a represenaive radable firm maximizes he profi p x y T w T l T r T k T 1, where p x is he relaive price for expors. To inroduce erms-of-rade shocks, he model assumes ha radable firms only produce for expors, and domesic

8 demand of radables is solely me by impors, priced a s. The erms of rade ξ px s follows an exogenous process where ε ξ N(0, σ 2 ξ ). ln ξ ξ = ρ ξ ln ξ 1 + ε ξ ξ, (11) C. Governmen Denoe he uni in foreign goods by. A each period, he governmen collecs axes and issues exernal bond (b ) o pay for expendiures, including governmen consumpion (g ), ransfers, and deb services. 10 Governmen consumpion is also a CES baske of nonradables and radables wih a degree of home bias ( ϕ G) and he inra-emporal elasiciy of χ. The relaive price of governmen consumpion is p G = [ ϕ G ( p N ) (1 χ) + ( 1 ϕ G ) (s ) 1 χ ] 1 1 χ. (12) A ime, he governmen sells b unis of bond a a price q, which raises q s b unis of local goods. A + 1, he governmen pays one uni of foreign goods for each uni of b if here is no defaul. In he case of defaul, i pays a fracion (1 +1 ) of he liabiliies. Le b d be he pos-defaul liabiliies. The governmen s flow budge consrain is ( ) τ w l + r N k 1 N + r T k 1 T +q s b = s (1 ) b 1 +p G g + z. (13) } {{ } } {{ } T, revenue b d Foreign crediors are assumed o be risk-neural. 11 Their demand for governmen bond is The governmen s ineremporal budge consrain is q = βe (1 +1 ). (14) (1 )b 1 = i=0 β i E 1 s +i ( T+i p G +ig +i z ). 12 (15) 10 In realiy, mos counries also issue domesic deb. Given our nonlinear soluion mehod, adding one more deb insrumen would dramaically increase compuaional ime. Since our focus is on exernal deb, we assume he governmen does no issue domesic deb for simpliciy. 11 We follow he common assumpion in he lieraure o assume foreign crediors are risk neural. Having risk adverse crediors in he model would accelerae he increase in risk premia when a governmen becomes more indebed. 12 To derive (15), we use (14) in (13), ierae i forward, and impose he ransversaliy condiion for governmen deb, lim j E β j (1 +j )b +j 1 = 0.

9 1. Defaul Scheme Following Bi (2012), defaul decisions depend on a realized effecive fiscal limi, B max, drawn from a fiscal limi disribuion B max (S ), condiioned on he sae S. If he governmen s liabiliies a he end of 1 are less han B max, i fully repays is deb ( = 0); oherwise, i reneges a fixed fracion of is liabiliies ( = d). Specifically, = { 0 if b 1 < B max d if b 1 B max The simulaion of B max (S ) is o be described in Secion IV. }, B max B max (S ). (16) 2. Fiscal Policy Governmen spending as a share of GDP in developing counries is generally low; reiring deb hrough cuing governmen spending may be difficul. In his model, we assume ha income axes adjus o mainain deb susainabiliy. To capure procyclical fiscal policy observed in developing counries (e.g., Alesina e al. (2008), Gavin and Peroi (1997), and Kaminski e al. (2004)), governmen consumpion responds o oupu wih a one-quarer delay (y 1 ). Thus, ax and governmen consumpion rules are specified as ln τ τ = ρ τ ln τ 1 τ + γ ln bd b + ετ, γ > 0 (17) ln g g = ρ g ln g 1 g + η g ln y 1 y + εg, ε τ, ε g N(0, σ 2 i ), i {τ, g}. (18) D. Aggregaion and Marke Clearing Oupu in unis of local goods is y = p N yn + ξ s y T. (19) The marke clearing condiion for nonradables is [ y N = (p N ) {ϕ χ c + i + v ( ) i N 2 δ k N 2 k 1 N 1 + v ( ) i T 2 ] } δ k T 2 k 1 T 1 + ϕ G (p G ) χ g. (20)

10 Finally, he balance-of-paymen condiion is c +i + v 2 ( i N k N 1 ) 2 δ k 1 N + v ( ) i T 2 δ k T 2 k 1 T 1 +pg g [ y = s q b (1 )b 1]. (21) Appendix I liss he equilibrium condiions of he model. III. ESTIMATION AND CALIBRATION To show how our framework can be used o assess fiscal limis of a counry, he model is calibraed o he recen economic condiions of Argenina as an example. I has had subsanial exernal public deb and a hisory of sovereign defaul. Bayesian echniques are applied o obain parameer values of hose characerizing economic uncerainy and fiscal policy rules. The esimaion is performed on he log-linearized model assuming no defaul. 13 The linearized model is solved by Sims s (2001) mehod. Four observables are used: real GDP, governmen spending, revenues, and he real exchange rae. Appendix II describes daa sources and he esimaion deails of he poserior mode. Table 1 summarizes he values of parameer calibraion. The model is a a quarerly frequency. Consisen wih he annual calibraion for Argenina in García-Cicco e al. (2010), he quarerly discoun facor β is se o 0.98 and he depreciaion rae δ o 0.03. Bursein e al. (2005) esimae ha he radable share in he consumer price index for Argenina is 0.53. For governmen consumpion, since a large proporion of governmen spending goes o pay services of public servans, ϕ G is se o 0.6, bigger han he degree of home bias in privae consumpion. To calibrae effecive consumpion c, we follow Bouakez and Rebei (2007) o se he weigh of privae consumpion o ω = 0.8. Since he elasiciy of subsiuion beween privae and governmen consumpion ν is no convenionally esimaed, we back ou ν = 0.49 o have he model-implied fiscal mulipliers roughly mach he esimaes for average developing counries in Ilzezki e al. (2013). 14 The elasiciy of subsiuion beween radables and nonradables in c and g (χ) is se o be 0.44, as esimaed by Sockman and Tesar (1995). To calibrae secoral mobiliy for labor, we follow Horvah s (2000) esimae using he U.S. secoral daa o se ξ l = 1. Following Gourio (2012), he invesmen adjusmen parameer κ is se o 1.7. 13 Esimaing he nonlinear model presened here is challenging, if possible. Bi and Traum (2012) and Bi and Traum (forhcoming) show how o esimae simple nonlinear DSGE models using he paricle filer. 14 Based on a sample 24 developing counries, Ilzezki e al. (2013) esimae ha he peak spending muliplier is slighly above 0.2 and he long-run muliplier is 0.63, alhough boh are insignifican. Also, ν = 0.49 implies ha privae and governmen consumpion are complemens, in line wih he conclusion of Karras (1994).

11 Our defaul scheme assumes a consan haircu rae d. Based on Surzenegger and Zeelmeyer s (2008) esimaed haircu raes of sovereign deb resrucures in emerging marke economies beween 1998 and 2005, Bi (2012) calculaes ha 90 percen of he annual haircu raes (as a share of all sovereign deb) fall below 0.3. For boh counries, we assume a consan quarerly haircu rae of 0.07 (equivalen o 0.28 for he annual rae). To calibrae he labor income share in each secor, we use Frankema s (2010) esimae of he labor shares in he naional income of Argenina, which has he labor income share near 0.55 in 2000. Since nonradable secors end o be a leas as labor-inensive as radable secors (Obsfeld and Rogoff (1996)), we se α N = 0.6 and α T = 0.55. The labor disuiliy weigh φ is se such ha he seady-sae fracion of ime devoed o work is 0.2. Fiscal policy in he seady sae is calibraed o he average of he sample used in Bayesian esimaion (2003Q1:2012Q2): he governmen spending share of oupu is 0.1476, and he ax rae, measured by he raio of oal revenues o GDP, is 0.227. Since he model only has exernal public deb, he deb-o-annual oupu raio is calibraed o he average share of exernal deb issued by he non-financial public secor and he cenral bank in GDP, equal o 0.24. For esimaed parameers, since no much informaion is available o guide our prior choices, all he priors are raher dispersed. The priors for all of he AR(1) coefficiens (ρ s) have a bea disribuion wih mean 0.5 and sandard deviaion 0.2. The priors for he sandard deviaions (σ s) of all shocks have an inverse gamma disribuion wih mean 0.1 and sandard deviaion of infiniy. For he cyclical fiscal parameer, he prior for η g follows a normal disribuion wih mean 0.5 and sandard deviaion 0.2, which imposes more weigh on procyclical spending policy. The fiscal adjusmen parameer γ has a gamma disribuion wih mean 0.05 and sandard deviaion 0.02. Since he income ax rae is he only insrumen for fiscal adjusmens, resricing γ > 0 is necessary o yield an equilibrium. The poserior mode suggess ha governmen spending is weakly procyclical wih η g = 0.1, and he response magniude of income ax rae o deb is γ = 0.06. IV. FISCAL LIMIT DISTRIBUTION The defaul scheme in he model requires simulaing fiscal limi disribuions. We firs simulae he uncondiional baseline disribuions (i.e., he disribuion wih an iniial sae a he seady sae) for Argenina and hen show how revenue collecion can affec disribuions. To see he role of curren economic shocks in affecing fiscal limis, a sae-dependen disribuion is also simulaed condiional on an iniial large erms-of-rade shock.

12 A. Simulaing Fiscal Limi Disribuion We define fiscal limis as he maximum level of deb in unis of local goods ha a governmen is able and willing o service. In erms of abiliy o pay, he maximum deb level equals he sum of all fuure discouned maximum primary surpluses. When compuing he maximum surplus of each period, he ax rae is se o he maximum ax rae τ max, chosen o be slighly above he highes revenue-oupu raio in he sample. 15 In he baseline simulaion, we se τ max = 0.29. 16 In erms of willingness o pay, we proxy i by a poliical risk facor 0 < θ 1. Sae-dependen fiscal limis are compued as B max (S i ) i=0 β θ 1 ( T max s max +i p G +i g +i z ), (22) +i } {{ } primary surplus where he sae of he economy is S = { ε a,ε g, ε ξ, k 1, N k 1 T }, and he superscrip max indicaes variables compued under τ = τ max. The expression of (22) is modified from he ineremporal governmen budge consrain (15). Firs, we assume ha he governmen does no defaul in he iniial period ; hence, = 0. For he analysis conduced under an iniial high-deb sae, we choose he deb-o-oupu raio such ha he defaul probabiliy is moderae. Second, he expecaion operaor is dropped as each draw of a fiscal limi B max (S ) from he disribuion B max (S i ) is condiional on he curren sae S and paricular sequences of realized shocks, using he Markov Chain Mone Carlo simulaions. Finally, he infinie sum of he maximum primary surplus is discouned by he poliical risk facor θ. To calibrae θ, we resor o he Inernaional Counry Risk Guide s (ICRG s) index of poliical risk. 17 The average raing for he sample periods is 66.5 ou of 100 for Argenina. We se θ = 0.67 for Argenina. 18 Appendix III describes he procedure o simulae fiscal limi disribuions. Figure 1 plos he cumulaive densiy funcion (CDF) of he baseline fiscal limi disribuions for Argenina. The x-axis plos fiscal limis in he raio of governmen deb o seady-sae annual GDP. The disribuion exhibis he propery ha when he sovereign defaul risk rises, 15 Bi (2012) ses τ max o he peak of a Laffer curve, which implies a maximum ax rae around 0.4 or higher. In developing counries, ax raes in his high range are rarely seen, and hus he maximum ax rae implied by he peak of a Laffer curve is less suiable. 16 The maximum revenue-o-oupu raio for Argenina in he sample is 0.261. 17 Area and Galina (2008) show ha ICRG s index significanly affecs he amoun of exernal credi in emerging markes. The index s poliical risk raing includes componens of governmen sabiliy, socioeconomic condiions, inernal and exernal conflic, corrupion, law and order, bureaucracy qualiy, ec. The range of raing is from 0 o 100, and a high raing indicaes low poliical risks. 18 Using ICRG s index o capure poliical risk is a shor cu. To properly model a counry s willingness o service deb requires o model a srucural poliical economy and is beyond he scope of his paper.

13 i ends o rise quickly. The defaul probabiliy is roughly zero when he deb-o-gdp raio is below 0.45. However, he probabiliy climbs o almos 1 when he deb-o-gdp raio reaches 0.7. Our esimaed fiscal limis for Argenina cover he acual deb levels in recen wo sovereign defaul episodes. The exernal deb-o-gnp raios a he year of defaul were 0.55 in 1982 and 0.53 in 2001 (Table 3 in Reinhar e al. (2003)). B. Revenue Collecion Capaciy The simulaed fiscal limis in Figure 1 generally are smaller han observed in many developed counries. One imporan facor driving he difference in fiscal limis beween developed and developing counries is revenue collecion capaciy. This capaciy is relaed o he maximum ax rae a governmen can implemen, subjec o poliical willingness and insiuion qualiy. Figure 2 compares he baseline disribuion for Argenina (τ max = 0.29, θ = 0.67, solid line) o wo alernaive assumpions. The dashed-solid line has τ max = 0.35 and θ = 0.67, in which he maximum ax rae is more in line wih he average effecive ax rae in he developed counries (Callen e al. (2003)). The dashed line has τ max = 0.35 and θ = 0.8, in which he poliical risk facor his he lower bound of ICRG s index for developed counries. Wih he same poliical risk facor, raising he maximum ax rae from 0.29 o 0.35 increases he mean of he fiscal limi disribuion from 0.61 o 1.02. If he poliical risk facor furher rises o 0.8, he mean of he disribuion rises o 1.22. Our simulaion shows ha he maximum ax rae a governmen can implemen has a large impac on fiscal limis: a one-percenage poin increase in he ax rae can raise he mean of fiscal limis by almos 7 percen of GDP for Argenina. The formulaion of fiscal limis (22) indicaes ha governmen spending is also imporan. Since governmen spending as a share of GDP for Argenina (or developing counries in general) is low, he room o increase fiscal limis hrough cuing governmen spending may be small. On he oher hand, developed and developing counries differ grealy in revenue receips, suggesing ha srenghening revenue collecion can be effecive in raising fiscal limis in developing counries. C. Devaluaion and Balance Shee Effecs Relaive o domesic deb, exernal governmen deb carries addiional risk due o flucuaions in he real exchange rae. In our baseline, he volailiy of he real exchange rae maches he daa (as he real exchange rae is one of he observables). Since he sample only covers he recen, pos-defaul period, i is likely o undersae he flucuaion of he real exchange rae for a longer period. Figure 3 compares he baseline disribuion for Argenina o he one wih a wice as large sandard deviaion of he erms-of-rade shock (σ ξ = 5.74 vs. 2.87 in he baseline). Comparing he wo disribuions, we noice ha a higher volailiy of erms-of-rade shocks produce a more dispersed disribuion. When he deb-o-gdp raio is 0.55, he defaul probabiliy raises from 0.06 under he baseline o 0.25 in he alernaive disribuion.

14 Anoher perspecive o show addiional defaul risk carried by exernal deb is o examine he condiional disribuion wih a large devaluaion in he real exchange rae. We subjec he esimaed Argenina economy o a 30-percen erms-of-rade shock, which leads o a real depreciaion of 20 percen from is seady sae iniially. Figure 4 compares he CDF of he baseline disribuion (solid line) o ha of he condiional disribuion (doed-dashed line). I shows ha a large exernal shock subsanially shifs he disribuion o he lef. Wih a deb-o GDP raio a 0.55, he defaul probabiliy increases from 0.06 o abou 0.4, urning a susainable fiscal pah o a risky one. Alhough a negaive erms-of-rade shock of 30 percen is rare, a sudden devaluaion of he real exchange rae by 20 percen or more is no uncommon around crisis imes. The implicaion of large negaive erms-of-rade shocks can be exended o oher shocks. For example, capial flow shocks may also be imporan in affecing fiscal limis of some developing counries (Calvo e al. (2004a)). Condiional disribuions highligh he impac of an iniial sae on fiscal limis and defaul risks. Even hough he fundamenal economic srucure and fiscal policy remain he same, emporary disurbances can move a disribuion and suddenly change he percepion abou fiscal susainabiliy in he shor run. V. FISCAL POLICY IN A HIGHLY INDEBTED ECONOMY Wih simulaed fiscal limi disribuions, he model is used o analyze wo fiscal issues ofen debaed in highly indebed economies: fiscal consolidaion and governmen spending effecs in a high-deb sae. The analysis is conduced using he model calibraed o Argenina. 19 A. The Economy in a High-Deb Sae To analyze he economy in a high-deb sae, we need o firs disurb he economy such ha is deb is much above he seady-sae level. We assume ha a sequence of small negaive TFP shocks ( 1 percen each period) hi he economy for 57 quarers saring in he seady sae ( = 80), where he deb-o-annual GDP raio is 0.24. A = 0 (defined as he iniial period of a high-deb sae analyzed here), he deb-o-annual GDP raio climbs o 0.52, and a N and a T have reurned o heir seady-sae values. 20 From = 80 o 1, he governmen underakes minimal fiscal adjusmens by seing γ = 0.04, below he esimaed γ = 0.06 for Argenina. The sae a = 0 is S 0 = { b d 0, εa 0, εg 0, k 1 N, 1} kt. Due o earlier negaive TFP shocks, k 1 N and kt 1 are abou 14 and 9 percen below heir seady-sae values. A ime 0, he ineres rae increases by abou 43 basis poins relaive o he seady-sae level. 19 In he model wih fiscal limis, he ax rae is endogenously deermined, and he AR(1) specificaion of he ax rule (17) furher expands he sae space. To increase compuaional efficiency, we rewrie he ax policy as ln τ τ = γlr ln bd, where he revised fiscal adjusmen parameer γ LR = γ/(1 ρ b τ ) is he average magniude of long-erm fiscal adjusmens. 20 From = 22 o = 1, ε a = 0, and a N and a T gradually reurn o he seady-sae level because ρ a > 0.

15 Figure 5 depics he ransiion dynamics reurning from a high-deb sae ( = 0) owards he seady sae under hree fiscal adjusmen speeds. The doed-dashed lines assume γ remains a 0.04 (or γ LR = 0.174) hroughou he horizon. The solid and dashed lines show he dynamics under wo bigger γ s o be discussed laer. The x-axis is in years and he y-axis in levels. The ineres rae (or risk premium 21 ) is repored as he annual rae in percen. For reference, he ligh doed lines are he sochasic seady sae as if here were no shocks hrough he simulaion periods. The fiscal adjusmen under γ = 0.04 represens he scenario wihou deliberae consolidaion effors. A higher deb level plus a higher ineres rae requires more deb paymen. In addiion, he ax rule (17) implies an income ax rae higher han he seady-sae level (a abou 0.26 vs. 0.23), bu mos addiional ax revenue is devoed o ineres paymens. The deb-o-oupu raio says around 0.5 for en years and declines very slowly o 0.475 weny years afer. Consequenly, he risk premium only slowly reurns o is seady sae level. Even wih lile fiscal consolidaion, he economy in a high-deb sae produces less oupu relaive o he seady sae. A lower afer-ax wage rae implies ha households have less disposable income o consume (by 8.0 percen a = 0 relaive o he seady-sae consumpion). Lower capial socks plus higher ax raes also induce households o save or inves less (by 5.6 percen a = 0). Falling consumpion increases he marginal benefi of labor, exering a posiive incenive o work more. The higher income ax rae, however, discourages work due o negaive subsiuion effecs. The ne effec is a small posiive response on labor relaive o he seady sae (by 1.4 percen a = 0). Overall, oupu in a high-deb sae is lower han he seady-sae pah (by 4.3 percen a = 0). The iniial sae we simulae hrough negaive TFP shocks is only one possible siuaion wih high governmen deb, because deb can accumulae due o oher ypes of shocks. The fiscal adjusmen channel riggered by higher deb services, however, operaes in general. Alhough addiional revenue o service deb needs no come from higher income ax raes, alernaive funding mehods, such as by lower governmen spending, would also produce lower oupu relaive o he sochasic seady-sae pah. In Secion C, we show ha governmen spending has a posiive oupu muliplier, implying negaive he oupu effecs of spending reducion. B. Fiscal Consolidaion Highly indebed governmens ofen face poliical pressure o consolidae a a fas speed. Solid and dashed lines in Figure 5 consider wo oher scenarios wih γ = 0.06 and γ = 0.08, respecively, saring from ime 0. 21 From ineres raes, risk premia can be compued as he difference beween he ineres rae and a risk free real rae, which can be proxied by he average yield of he U.S. Treasury bond roughly a 3 percen (Trevino and Yaes (2012)).

16 As expeced, he deb-o-oupu raio falls more quickly when γ is higher. By he end of year five, he deb-o-oupu falls from 0.52 o 0.38 wih γ = 0.08, and he risk premium roughly reurns o is seady-sae level wo years afer he consolidaion sars. Despie he benefis of lowering risk premia, he comparison of γ = 0.06, 0.08 o γ = 0.04 (he scenario wih lile consolidaion) indicaes ha faser fiscal consolidaions are more counerproducive in erms of oupu los in he shor and medium runs. As income ax raes are higher o reire deb sooner, hey have more negaive effecs on consumpion, invesmen, and oupu, relaive o he pahs wih γ = 0.04. In he longer run, as deb falls more quickly wih a faser consolidaion, less ax revenue is needed o service deb; he ax rae falls below he rae wih γ = 0.04 abou eigh years afer consolidaion. In conras o he responses in he firs en years, lower ax raes generae less negaive consumpion and invesmen responses relaive o he pahs wih γ = 0.04. For labor responses, a faser consolidaion generaes more posiive responses in he medium run, mainly due o he subsiuion effec from lower ax raes wih γ = 0.08. In laer years, labor becomes less posiive wih γ = 0.08 mainly due o he income effec, as oupu is higher beween he wo consolidaion pahs. On he exernal side, he higher ax rae under a faser consolidaion conracs domesic demand and depreciaes he real exchange rae more han under a slower consolidaion. As he ax rae rises highes wih γ = 0.08, he real exchange rae depreciaes mos. Thus, radable oupu falls leas for he firs seven quarers, because a more depreciaed real exchange rae improves compeiiveness of he radable secor more. As he magniude of real depreciaion declines laer, he effec of lower capial under a faser consolidaion dominaes, and he radable oupu falls more relaive o a slower consolidaion. Alhough empirical evidence is inconclusive abou he relaionship beween growh and deb (Cecchei e al. (2011), Herndon e al. (2013), and Reinhar and Rogoff (2010)), our analysis suppors ha growh is lower in a high-deb sae han in he seady sae, as a resul of lower capial socks and fiscal adjusmens required o service deb. Lowering governmen deb, however, is no wihou pains. Fiscal consolidaions have overall negaive effecs on he economy on key macro variables, despie he benefis of lower risk premia. C. Governmen Spending Effecs in Differen Saes of Deb To see how governmen indebedness maers for governmen spending effecs, we examine an exogenous increase in governmen consumpion in differen saes of deb. Before he spending increase, he high-deb sae a = 0 is simulaed by a similar mehod in Secion A, excep for γ = 0.06 (esimaed value for Argenina) hrough he enire simulaion periods. The low-deb sae is he sochasic seady sae. Given his iniial sae, a series of governmen consumpion shocks are injeced saring a = 0; governmen consumpion rises by 3.1 percen of he seady-sae GDP on average for he firs year.

17 Before solving he model wih governmen spending effecs, we firs simulae he condiional disribuion, dependen on he iniial governmen spending shock. Figure 6 shows ha when comparing o he baseline (solid line), he lower iniial capial and posiive governmen consumpion shocks shif he condiional disribuion o he lef (doed-dashed lines). Lower iniial capial implies ha capial is likely o be below seady-sae value for some ime, which reduces producion capaciy and hence he maximum curren and fuure revenues ha can be raised. Togeher wih higher governmen spending, expeced fuure governmen surplus is reduced, shifing he fiscal limi disribuion o he lef. The mean deb-o-oupu raio of he condiional disribuion is 0.57, compared o 0.61 in he baseline disribuion. Figure 7 compares he ransiion dynamics wihou spending shocks (dashed lines) o hose wih he shocks (solid lines) in a high-deb sae wih he deb-o-oupu raio a 0.5. Figure 8 conducs he same experimen in a low-deb sae wih he deb raio a 0.24. The dashed lines are he pahs wihou governmen consumpion shocks, so he differences beween he wo lines are he spending effecs. The wo figures show ha governmen spending has he same qualiaive response paerns, excep for he ineres rae. In he high-deb sae, he risk premium rises by abou 150 basis poins a he peak. In conras, he fiscal expansion does no increase he premium in a low-deb sae. Given he non-lineariy of risk premium changes, spending increases raise he premium subsanially only when an economy sufficienly approaches is fiscal limis. From he oupu responses in Figures 7 and 8, governmen consumpion is expansionary only for he firs year in boh saes. To quanify governmen spending effecs, Table 2 repors he cumulaive mulipliers for oupu, consumpion, and invesmen. The cumulaive muliplier k quarers afer an increase in governmen consumpion is defined as k i=1 βi 1 y +i 1 k i=1 βi 1 p G +i 1 g, (23) +i 1 where y and g are level changes relaive o a pah wihou governmen consumpion shocks. When compuing consumpion, invesmen, and rade balance mulipliers, y is replaced by c, i, or b (see (I.33) in Appendix I for he compuaion of rade balance). The posiive consumpion mulipliers conribue o he expansionary effecs in he shor run due o is complemenariy o governmen consumpion. Lower invesmen and rade balance, however, offse he expansionary effec, yielding he oupu muliplier much below 1. Addiional borrowing o finance he governmen consumpion elevaes he deb-o-oupu raio. The emporary decline in he firs year is due o real appreciaion and reduced liabiliies in local good unis. As a resul, he risk premium does no rise much iniially in he high-deb sae. In boh saes, he income ax raes rise in response o higher deb. Despie ha governmen deficis are fully financed by exernal borrowing, governmen spending sill crowds ou invesmen hrough he fiscal adjusmen channel. 22 As menioned, he 22 In a closed economy, a higher governmen consumpion crowds ou invesmen hrough a higher domesic ineres rae.

18 complemenariy beween governmen and privae consumpion induces households o consume more. Despie higher income ax raes, households work harder o suppor higher consumpion. The deerioraed rade balance implies ha he expansionary effec comes from higher producion in he nonradable secor. The radable secor loses compeiiveness because of he real appreciaion in he firs year. The small peak oupu mulipliers (around 0.2) and long-run negaive mulipliers are consisen wih recen empirical findings for average developing counries (see foonoe 14). Comparing across he wo saes, Table 2 shows ha a smaller oupu muliplier in he high-deb sae is mainly conribued by a smaller consumpion muliplier. Governmen consumpion is less simulaive for privae consumpion because he economic cos o increase consumpion is higher in a high-deb sae. Since he governmen has o collec more revenue o service deb, he higher income ax rae in a high-deb sae implies he afer-ax wage rae is lower. Thus, he cos of incremenal consumpion in erms of leisure sacrificed is higher, so an increase in governmen consumpion becomes less effecive in raising privae consumpion. In a high-deb sae, invesmen is less crowded ou han in he low-deb sae, mainly because of less posiive consumpion responses. 23 Our resul ha governmen spending mulipliers become smaller when an economy approaches is fiscal limis echoes he findings of some recen papers. Ilzezki e al. (2013) and Nickel and Tudyka (2013) obain smaller or even negaive mulipliers when he economy has a high deb-o-gdp raio. As fiscal adjusmens loom large in a highly indebed economy, anicipaion of he adjusmens can offse he expansionary effecs of governmen spending. Corsei e al. (2013) also conclude he higher he iniial deb level, he smaller he governmen spending mulipliers. Their resuls are driven by he posiive links beween sovereign defaul risks and funding coss of he privae secor. VI. CONCLUSION We sudy fiscal limis and fiscal policy effecs in developing counries wih exernal deb. A DSGE framework wih sovereign defaul risks is consruced o simulae fiscal limi disribuions. Simulaions for Argenina show ha expeced fuure revenue plays an imporan role in explaining he relaively low fiscal limis observed in developing counries compared o developed counries. Sae-dependen disribuions inform how fiscal limis can change when an economy is hi by various ypes of shocks. In paricular, shocks ha lead o sharp real depreciaion can suddenly raise defaul probabiliies of an economy wih large exernal deb. The wo fiscal issues analyzed are fiscal consolidaion and governmen spending effecs in differen saes of deb. Fiscal consolidaions have a negaive effec on oupu, despie of 23 Our heoreical finding differs from Magud (2008), who finds ha counercyclical governmen spending has a negaive effec oupu if he iniial sae of governmen deb is high.

19 falling risk premia. While a faser consolidaion lowers deb and risk premia more quickly, i is more counerproducive han a slower consolidaion because of higher curren axes, as well as expecing higher fuure ax raes. Increasing governmen spending in a high-deb sae pushes he economy closer o is fiscal limis, raising risk premia and defaul probabiliies. The expansionary effec of a spending increase is weaker in a high-deb sae han in a low-deb sae. Alhough he model used here embeds sovereign defaul risks, i does no incorporae he negaive coss associaed wih defaul. In addiion, our hough experimens are crafed such ha a fiscal expansion in he high-deb sae increases defaul risks bu defaul probabiliies are sill moderae. In pracice, if an economy is much closer o is fiscal limis han he deb level we simulae or he size of spending increases is bigger han wha is assumed here, expansionary fiscal acions could rigger more imminen and drasic fiscal adjusmens. Fiscal mulipliers in hese circumsances can be even smaller han wha we obain here, and he economy can expose o higher defaul risks and is poenial negaive consequences.

20 APPENDIX I. EQUILIBRIUM CONDITIONS c = [ω (c ) ν 1 ν ] + (1 ω)(g ) ν 1 ν ν 1 ν (I.1) λ = ωc 1 ν c ( 1 1) ν (I.2) φ(l ) σ = λ (1 τ )w (I.3) Q N ( ) i N = 1 + κ δ k 1 N (I.4) ( i Q T T = 1 + κ k 1 T ) δ (I.5) Q N = β E λ +1 λ [ (1 τ +1 )r N +1 κ 2 ( i N +1 k N 2 ( )( ) ] i N δ) + κ +1 i N δ +1 + Q N k N k +1(1 δ) N (I.6) Q T = β E λ +1 λ [ (1 τ +1 )r T +1 κ 2 ( i T +1 k T 2 ( )( ) ] i T δ) + κ +1 i T δ +1 + Q T k T k +1(1 δ) T (I.7) l = [ (ϕ l ) 1 χ l ( l N ) 1+χ l χ l + (1 ϕ l ) 1 χ l ( l T ) 1+χ l χ l ] χ l 1+χ l (I.8) l N ( ) w = ϕ l N χ l l w (I.9) ( ) w l T = (1 ϕ l T χ l ) l w (I.10)

21 i = i N + i T (I.11) k N = (1 δ)k N 1 + i N (I.12) k T = (1 δ)k T 1 + it (I.13) y N = a N ( k N 1 ) 1 α N ( l N ) α N (I.14) α N p N y N = w N l N (I.15) (1 α N )p N y N = r N k N 1 (I.16) (1 α T )p x y T = r T k T 1 (I.17) y T = a T ( k T 1 ) 1 α T ( l T ) α T (I.18) α T ξ s y T = w T lt (I.19) 1 = [ ϕ(p N ) 1 χ + (1 ϕ)(s ) 1 χ] 1 1 χ (I.20) p G = [ ϕ G ( p N ) (1 χ) + ( 1 ϕ G ) (s ) 1 χ ] 1 1 χ (I.21) D N [ = ϕ c + i + κ ( i N 2 k N 1 ) 2 δ k 1 N + κ ( ) i T 2 δ k T 2 k 1 T 1 ] + ϕ G (p G ) χ g (I.22) y N = (p N ) χ D N (I.23)

22 y = p N y N + ξ s y T (I.24) c + i + v 2 ( i N k N 1 ) 2 δ k 1 N + v ( ) i T 2 [ ] δ k T 2 k 1 T 1 + p G g y = s q b (1 )b 1 (I.25) q = βe [(1 +1 )] (I.26) ( τ w l + r N kn 1 + rt 1) kt + q s b = s (1 )b 1 +p G } {{ } g + z (I.27) b d ln τ τ = ρ τ ln τ 1 τ + γ ln bd b + ετ (I.28) ln g g = ρ g ln g 1 g + η g ln y 1 y + εg (I.29) ln an a N = ρ a ln an 1 a N + εa (I.30) ln at a T = ρ a ln at 1 a T + εa (I.31) r = 1 q (I.32) b = y c i v 2 ( i N k N 1 ) 2 δ k 1 N v ( ) i T 2 δ k T 2 k 1 T 1 p G g (I.33) lnξ = ρ ξ lnξ 1 + ε ξ (I.34)

23 APPENDIX II. BAYESIAN ESTIMATION The esimaion purpose is o calibrae he process of economic shocks and fiscal policy rules adoped during normal imes. The esimaion is performed on he log-linearized model assuming no defaul. The pos-defaul sample from 2003:Q1 o 2012:Q2 is used for Argenina. The mos recen economic crisis in Argenina lased from 1999 o 2002, and a sovereign defaul occurred a he end of 2001. Observables include real GDP, governmen spending, revenues, and he real exchange rae. Daa are colleced from he daabase of Emerging Markes for Lain America complied by Haver Analyics. All daa are seasonally adjused, eiher a he source or by applying U.S. Census s X12 program. Real GDP is in 1993 millions of pesos. Fiscal daa are aken from he consolidaed governmen budge. Governmen spending is he sum of public consumpion and capial expendiures. Revenues include ax revenues, conribuions o social securiy, and all sources of non-ax revenue. Capial expendiures and revenues are in curren millions of pesos and deflaed by he GDP implici price index. Real exchange rae daa are aken from he JP Morgan s rade-weighed exchange rae index, and he rade weighs are based on he counry s 2000 bilaeral rade in manufacured goods. The deflaor used is WPI for domesic manufacured goods. Excep for real exchange rae, all seasonally adjused real daa (denoed by X ) are ransformed o x by ( x = 100 ln X populaion index ). (II.1) Then, x and he real exchange rae are derended o obain percen deviaions from an underlying rend, consisen wih he log-linearized model. The populaion index is consruced such ha 2008Q1=1 for Argenina. The model has no growh; daa are derended wih a linear rend, as in Smes and Wouers (2003). The minimizaion rouine csminwel by Chrisopher Sims is used o search for he se of srucural parameers ha minimize he negaive log poserior funcion. The parameer space of search is resriced o he one in which he model has a unique raional equilibrium. The mode search was iniiaed from 20 differen iniial values, and all converged o he values repored in Table 1. APPENDIX III. SIMULATING FISCAL LIMIT DISTRIBUTIONS This appendix describes procedures in simulaing fiscal limi disribuions, defined as [ B max (S ) β i θ 1 ( T max s max +i p G +ig +i z )], (III.1) +i i=0