The Tequila Effect: Theory and Evidence from Argentina

Board of Governors of the Federal Reserve System International Finance Discussion Paper Number 552 June 1996 THE TEQUILA EFFECT: THEORY AND EVIDENCE FROM ARGENTINA Mart(cid:19)(cid:16)n Uribe Note: International Finance Discussion Papers are preliminary material circulated to stimulate discussion and critical comments. References in publications to International Finance Discussion Papers(otherthan anacknowledgementthatthewriterhadaccess tounpublished material)should be cleared with the author or authors.

ABSTRACT The Tequila E(cid:11)ect hypothesis states that the economic crisis that a(cid:11)ected several South American countries in 1995 was caused by an exogenous capital (cid:13)ight triggered by the loss of con(cid:12)dence of foreign investors after the collapse of the Mexican peso in December 1994. I analyze the Argentine experiencebeforeandaftertheMexicancrisisandarguethattheTequilaE(cid:11)ectplayedanimportant role in the 1995 crisis. I model the Tequila E(cid:11)ect in an optimizing, small, open economy, as a situation in which agents at time 0 learn that at some random future date foreign investors will pull theirassetsoutofthe country. The model captureskey featuresofthe Argentine crisisof1995: thedecline in aggregatedomesticspendingandthe out(cid:13)owofcapital thatbegan in December1994; the credit crunch and interest rate hike of March 1995; the slow return of the real interest rate to its pre-crisis level, and the protracted decline in output and investment that began in March 1995.

The Tequila E(cid:11)ect: Theory and Evidence from Argentina (cid:3) Mart(cid:19)(cid:16)n Uribe 1 Introduction InApril1991theArgentinegovernmentembarkedonanambitiousin(cid:13)ation-stabilizationprogram| the Convertibility Plan|whose main components were an exchange rate peg to the U.S. dollar and a far-reaching structural reform that included expenditure cuts, the privatization of virtually all state-owned enterprises, tax reform, and (cid:12)nancial and trade liberalization. The initial results were impressive: high in(cid:13)ation disappeared instantaneously, output grew at an average annual rate of almost8%between1991and1994,andconsumptionandinvestmentexpandedatevenhigherrates. At the same time the trade balance showed large de(cid:12)cits and the real exchange rate appreciated. This economic boom came to an abrupt end with the collapse of the Mexican Peso in December 1994 when a run on the Argentine banking system began that led to a traumatic credit crunch in March 1995 and put many (cid:12)nancial institutions out of business. The (cid:12)nancial crisis had disastrous real e(cid:11)ects: In 1995, output declined 4.5% and gross domestic investment 16%; the trade balance turned positive for the (cid:12)rst time in three years, re(cid:13)ecting a contraction in domestic spending; and the real exchange rate showed the (cid:12)rst signs of depreciation as the economy entered a de(cid:13)ationary phase. What caused this reversal of fortunes? One explanation is that a boom-recession cycle is inherent to exchange-rate based stabilization programs. Empirical studies (Kiguel and Liviatan, 1992; V(cid:19)egh, 1992)have found thatthe e(cid:11)ectsof exchange-rate based stabilization programsimplemented in di(cid:11)erent countries and at di(cid:11)erent points in time are remarkably similar|initial economic expansion, trade balance deterioration and real exchange rate appreciation is followed by recession, trade balanceimprovementandreal depreciation. A vastliterature hasbeendevotedtoformalizing 1 this explanation and di(cid:11)erent mechanisms have been advanced. (cid:3) The author is a sta(cid:11) Economist at International Finance Division of the Board of Governors of the Federal Reserve System. I would like to thank Jaime Marquez, Pierre-Richard Ag(cid:19)enor, Stephanie Schmitt-Groh(cid:19)e, and seminar participants at the Inter American Development Bank for helpful comments, and Sherrell Varner for editorial assistance. This paper represents the views of the author and should not be interpreted as re(cid:13)ecting the views of the Board of Governors of the Federal Reserve System or other members of its sta(cid:11). Address for correspondence: International FinanceDivision,Stop24,Board ofGovernorsoftheFederal Reserve System,Washington,DC20551, (202)452 3780, e-mail uribem@frb.gov. 1 Forearlyexplanationsbasedonreduced-formmodelswithstickypricesandadaptiveexpectations,seeRodr(cid:19)(cid:16)guez 1

An alternative explanation argues that business cycles in LDCs are to a large extent driven by movements in the interest rate in industrialized countries. Calvo, Leiderman and Reinhart (1993) (cid:12)nd that the international interest rate is a signi(cid:12)cant determinant of the real exchange rate and of international capital (cid:13)ows to Latin America. Similarly, Dooley, Fern(cid:19)andez-Arias and Kletzer (1996) (cid:12)nd that the interest rate in developed countries is the main factor explaining movements in the secondary market price of foreign public debt of developing countries. A third possible explanation for the reversal of fortunes is the \Tequila E(cid:11)ect" (or \Contagion E(cid:11)ect") hypothesis, which states that the Argentine crisis of 1995 was caused the reluctance of foreign investors to renew their loans to the country, resulting from the fear that it would soon follow in Mexico’s footsteps. According to this hypothesis the change of con(cid:12)dence of foreign investors was exogenous to the country and, moreover, completely unrelated to domestic economic fundamentals. In this paper, I undertake an empirical and theoretical investigation of the Tequila E(cid:11)ect hypothesis. Insection2,IexaminetheArgentineeconomicexperiencesincetheimplementationofthe Convertibility Plan in 1991, with emphasis on developments after the Mexican crisis of December 2 1994. I argue that there was a clear Tequila E(cid:11)ect in the Argentine crisis of 1995, and that certain elements of that crisis can hardly be explained by either the exchange-rate based stabilization hypothesis or the international interest rate hypothesis. In section 3, I present a model of a small, open economy that starts out as a net borrower from the rest of the world. As is well known, under the assumption that the rate of time preference equals the international real interest rate, the equilibrium in such an economy is characterized by a constant level of consumption and a positive trade balance large enough to service the (constant) net foreign debt. I refer to this scenario as the crisis-free equilibrium. I then model the Tequila E(cid:11)ect as a situation in which at some point, say period 0, domestic residents learn that at some future date foreign investors will decide to hold no domestic assets. I consider (cid:12)rst the simple case of a perfect foresight, endowment economy, and then introduce uncertainty about the time of the pull out, and capital accumulation. Under perfect foresight domestic residents learn in period 0 the exact date, say period T, at which foreign investors will pull out of the country. In this case, consumption falls in period 0 and stays constant through period T at a level such that the stock of debt in period T is 0. In period T +1, consumption jumps up to the endowment level and stays at that level thereafter. The domestic interest rate is constant and equal to the international interest rate between periods 0 and T (cid:0) 1, jumps up in period T, and returns to its initial level in period 3 T +1. (1982)andDornbusch(1982). Formodelsoftemporarystabilizationandimperfectcredibility,seeCalvo(1986,1987) andfurther developmentsby Drazen and Helpman(1987,1988),Calvoand V(cid:19)egh (1993),ReinhartandV(cid:19)egh (1995), and Mendoza and Uribe (1996), among many others. For supply-side explanations see Uribe (1993), Roldo(cid:19)s (1995) and Lahiri (1995). Rebelo and V(cid:19)egh (1995) provide a survey of this literature. 2 Calvo and Reinhart (1996) analyze cross country data on asset returns to identify contagion e(cid:11)ects in (cid:12)nancial crisis. 3 The assumption that foreign investors pull completely out of the country is made for simplicity but is by no 2

This simple setup captures some key elements of the Argentine (cid:12)nancial crisis of 1995, namely, the capital out(cid:13)ow that started right after the Mexican crisis in December 1994, the corresponding trade balance improvement, and the interest rate hike of March 1995. But it leaves out other important aspects of the data, in particular, the fact that at the time of the credit crunch in March 1995, the decline in aggregate demand that had started three months before greatly deepened, generatingasigni(cid:12)cantimprovementinthetradebalanceinthesecondquarterof1995. Tocapture this feature of the crisis I analyze, in section 4, a version of the model in which domestic agents are uncertain about the time at which foreign investors will actually stop renewing their loans to the country. In this case, the credit crunch, although not completely unexpected, comes partly as a surprise and therefore induces a contraction in aggregate domestic spending. A second aspect of the data that the basic model is unable to capture is the fact that although real interest rates came down substantially after the March 1995 peak, they stayed consistently above their pre-Mexican crisis levels through the rest of 1995 and beyond. The two versions of the model described above predict, contrary to this fact, an immediate return of the interest rate to its pre-crisis level. In section 5, I extend the model by allowing for capital accumulation. In this version of the model the slow convergence of the interest rate occurs as follows: the interest rate hike in period T causes a decline in investment, implying that the economy starts period T +1 with a lower stock of physical capital. From period T + 1 on the economy is isolated from the international capital market and therefore collapses to the standard closed-economy, neoclassical growth model, which predicts that if the initial stock of capital is below its steady-state level, the capital stock converges monotonically from below to its long-run level and the interest rate, given by the marginal product of capital, converges monotonically from above. The decline in output that this third version of the model predicts is consistent with the recession observed in 1995 and, in particular, with the fact that the recession started after the interest rate hike in March 1995. 2 Argentina Before and After the Mexican Crisis The Argentine economic policy of the 1970s and 1980s was characterized by (cid:12)scal irresponsibility and repeated attempts to stabilize prices. Because of chronic (cid:12)scal imbalances, each stabilization program resulted in higher in(cid:13)ation and greater public debt. When President Carlos Menem took o(cid:14)ce in July 1989, the situation was at a critical point. In the (cid:12)rst half of that year, prices grew at an average monthly rate of 38% ((cid:12)gure 1) and the (cid:12)scal de(cid:12)cit was above 20% of GDP (table 1). 2.1 The Convertibility Plan and Structural Reforms The two main elements of the stabilization strategy adopted by the Menem administration were a (cid:12)xed exchange rate regime and a far-reaching (cid:12)scal reform. The \Convertibility Act," passed by Congress in March 1991, established a one-to-one exchange rate between the Argentine peso means necessary. All resultscarry out under the assumption of partial pull out. 3

and the U.S. dollar, eliminated all exchange and capital controls, and required that most of the 4 monetary base be backed by international reserves. The (cid:12)scal reform had four main components: tax reform, cuts in government spending, a vast privatization program, and reductions in the domestic and foreign public debt. As a result of an aggressive reform of the value-added tax which widened the tax base and fought evasion by computerizing the tax-collection system and increasing the number of random inspections of businesses, revenues from the value-added tax increased from 4% of GDP in 1990 to 5.5% in 1991 and 8.5% in 1992. Expenditure cuts came mainly from a drastic reduction in public employment, which in 1989 accounted for 40% of the federal administration’s budget. Between April 1990 and December 1991 federal government employment was cut 23.3% (table 2). Around 70% of this reduction was achieved through a freeze on (cid:12)lling vacancies and the remainder came through a voluntary early retirement program that included severance pay. Between 1989 and 1992, 51 state-owned enterprises were privatized generating, approximately $18 billion in cash and debt-reduction schemes. Income from this source represented 0.6% of GDP in 1990, 1.6% in 1991, and 1.2% in 1992. The revenue from the sale of assets was not the only way by which privatizations contributed to the reduction of the (cid:12)scal de(cid:12)cit: Ine(cid:14)ciencies in the administration of these enterprises together with politically determined prices, had resulted in systematic losses of around 4% of GDP (Edwards, 1996). From 1989 to 1990 interest payments on the public debt (lines 4 and 6 of table 1) fell by 16% of GDP. The decline resulted from a series of controversial measures by which the government 5 defaulted on most of the domestic public debt. In addition, in 1991 the country, under the Brady Plan, reached a foreign debt reduction agreement by which the external debt|$ 61 billion at that time|wasreducedabout9%andwas(cid:12)nancedatlowerratesandlonger,mostly30-year,maturities. 2.2 The Initial Boom: 1991-1994 The Convertibility Plan succeeded in quickly stopping high in(cid:13)ation. The average monthly CPI in(cid:13)ation rate fell from 11.4% in the 12 months preceding the announcement of the plan to 2.2% in the 12 months following the announcement, 1.4% in 1992, .6% in 1993, and .4% in 1994 ((cid:12)gure 1). In spite of its steady decline, in(cid:13)ation stayed signi(cid:12)cantly above the rate of devaluation during the (cid:12)rst four years of the plan, and, as a consequence, the real exchange rate steadily appreciated. The relative price of tradablesin terms ofnontradables, asmeasured bythe CPI-adjusted exchange rate 4 The charter of the central bank establishes that its holdings of government bonds cannot exceed 20 percent of international reserves. 5 The repudiation process began on the (cid:12)rst business day of 1990 with the implementation of the \Bonex plan." Under the plan, the government eliminated the de(cid:12)cit of the central bank The de(cid:12)cit of the central bank|also referred to as quasi-(cid:12)scal de(cid:12)cit|totaled in 1989 6% of GDP, and originated in interest payments, at market rates, on required reserves on time deposits. (Reserve requirements on time deposits were around 80% by thattime.) The Bonex Plan forced depositors to exchange their time deposits for government bonds (bonex) in such terms that the public lost above 70% of the real value of their deposits. A similar strategy was applied to reduce the debt of the non-(cid:12)nancial publicsector. 4

between the Argentine peso and the U.S. dollar fell almost 25% between April 1991 and December 1994 ((cid:12)gure 2). The initial phase of the plan was also characterized by a boom in output, consumption, and investment and by a large deterioration in the trade balance. Figure 3 shows annual observations forGDPandforthe ratioofconsumption,investment,andthe tradebalancetoGDPforthe period 1980-1995. Duringthe pre-Mexican-crisis phase ofthe Convertibility Plan GDPgrewatan average 6 annual rate of 7.7%, total consumption at 8.7%, and gross domestic investment at 21.8%. At the same time, imports grew at an average annual rate of 40.5% while exports expanded at only 3.0% per year, consequently, the trade balance deteriorated from a surplus of 6.6% of GDP in 1990 to a de(cid:12)cit of 4.8% of GDP in 1994. 2.3 The 1995 Financial Crisis The rapid pace of economic expansion in the (cid:12)rst four years of the Convertibility Plan ended abruptly with the Mexican crisis of December 1994. At that time a major (cid:12)nancial crisis developed asthe public began a massive withdrawal of bank deposits|(cid:12)rstof peso-denominated deposits and then of dollar-denominated deposits ((cid:12)gure 4). Total peso- and dollar-denominated bank deposits, which had grown at an average annual rate of 57% between 1991 and 1994, declined at an annual rate of62%inthe(cid:12)rstthree monthsof1995. Toful(cid:12)lltheircustomers’demands, bankswereforced to stop making new loans ((cid:12)gure 5). As a result, total loans, which between 1991 and 1994 had grown at an average annual rate of 30%, declined in the (cid:12)rst quarter of 1995 at an annual rate of 13%. This liquidity crisis led to a rise in interest rates. The prime rate rose from an average of 10% for peso-denominated loans and 8% for dollar-denominated loans over the period June 1993- December1994 to apeak of33%forpeso-denominatedloansand 22%fordollar-denominated loans 7 in March 1995 ((cid:12)gure 6. The interest rate hike, in turn, had a devastating e(cid:11)ect on the stock market ((cid:12)gure 7). Thedecline in bankdepositswasaccompanied byastrongpreferenceforforeigncurrencywhich led to a sharp reduction in the stock of foreign reserves held by the central bank ((cid:12)gure 8). In trying to cope with this situation, the monetary authority in December of 1994 lowered reserve requirements from 43% to 33% for checkable deposits and from 3% to 1% for time deposits. On March 1, 1995, the government by decree, changed the charter of the central bank and granted the institution wide (cid:13)exibility and discretion in providing liquidity, including the ability to acquire portfoliosofbanksin(cid:12)nancialstress. Thisdrasticchangeinmonetarypolicymeantthatthecentral bank became de facto a lender of last resort. In the two weeks following the change, the central 6 The labor market did not experience a similar pattern. Although employment grew moderately between 1991 and1993,thestructuralreformsinthepublicsectorandthelargechangeinrelativepricescontributedtoanincrease in the unemployment rate from 6.2% in October 1990 to 12.2% in October 1994. 7 Thespreadbetweenthepesoandthedollarinterestrate,whichhadbeenbelow2%intheperiod1993-1994rose to around 10%in March 1995, re(cid:13)ecting the possibility that the public in the (cid:12)rst quarter of1995 feared not only a bank crisis but also an exchangeratecollapse. 5

bank engaged in a massive expansion of domestic credit aimed at sterilizing the ongoing decline in the monetary base ((cid:12)gure 9). Between March 1 and March 14, 1995, international reserves fell $2 billion and the central bank sterilized 40% of the decline; in contrast, in the (cid:12)rst two months of the year (January 2 to February 28), foreign reserves fell $2.5 billion and the central bank sterilized only 13% of the decline. The credit crunch in the (cid:12)rst quarter of 1995 marked the end of four years of economic growth following the initiation of the Convertibility Plan in April 1991. Industrial production grew 4% (4quarter rate) in the (cid:12)rst quarter of 1995 but fell drastically in the second and third quarters|4.6% and 9.5% respectively|((cid:12)gure 10). The signi(cid:12)cant improvement in the trade balance observed in 1995 ((cid:12)gure 11) re(cid:13)ects a slowdown in aggregate domestic absorption. On March 14, the government announced that it had regained access to credit through international (cid:12)nancial institutions such as the IMF ($2.4 billion), the World Bank ($1.3 billion), and the IADB ($1.5 billion), and private foreign lenders ($1 billion). It also obtained a three-year $l billion loan from domestic entrepreneurs (the \patriotic" bond). On the same day, the government announced an austerity plan that included, among other measures, an increase in the value added tax rate from 18% to 21%, imposition of a 3% import tax, the elimination of export subsidies, and expenditure cuts of $1 billion. These announcements played an important role in containing the (cid:12)nancial crisis. In the second quarter of 1995, bank deposits began to recover gradually and interestratesstarted tocome down. However,total depositsand loans didnotreach theirpre-crisis levels in 1995. Similarly, the interest rate remained above its pre-Mexican crisis level throughout 1995. As a consequence of the (cid:12)nancial crisis, in 1995 the Argentine economy sunk into a recession of major proportions with GDP falling 4.5%, and gross domestic investment 16%. 2.4 Discussion The data show a clear relation between the timing of the Mexican and Argentine crises. Could this correlation be due to a common exogenous factor? As pointed out above, empirical studies have found that the interest rate in industrialized countries is the most important external factor a(cid:11)ecting capital (cid:13)ows to Latin America (Calvo et al., 1993; Dooley et al., 1996). Figure 12 shows the U.S. intended federal funds rate for the period January, 1993 to May, 1996. The federal funds rate was constant at 3% from the beginning of the sample period until January, 1994. A tightening period started in February, 1994 when the Federal Reserve raised the intended federal funds rate 25 basis points. By June, 1994, the federal funds rate stood at 4.25% and by December at 5.5%. The last increase of that tightening period occurred in February, 1995 when the federal funds rate was increased by 50 basis points to 6%. An explanation of the Argentine crisis of 1995 based on movements in the interest rate in industrialized countries would suggest that the capital out(cid:13)ow process that initiated the crisis should have started early in 1994, when the U.S. monetary authority began raising interest rates, rather than in January, 1995, when most analysts were already anticipating the end of the tightening period. 6

Anotherexplanationmentionedabovearguesthataboom-recessioncycleisinherenttoexchangerate based stabilization programs like the Argentine Convertibility Plan. The temporariness or imperfect credibility hypothesis (Calvo, 1986; Calvo and V(cid:19)egh, 1993) says that if in(cid:13)ation acts as a tax on domestic spending, then a reduction in in(cid:13)ation perceived as temporary by economic agents will induce a substitution of current for future domestic spending, causing trade balance deterioration, real appreciation, and in(cid:13)ation. The increase in current domestic spending occurs because agents take advantage of the temporary reduction in the in(cid:13)ation tax. If at some point after the implementation of the stabilization program either the devaluation rate is permanently increased oragentslearnthatthe stabilization programis permanent,the incentives forintertemporal substitution disappear and consumption falls to thelevel consistentwith permanentincome. Could these be the forces behind the 1995 Argentine recession? Under perfect foresight, the imperfect credibility hypothesis implies that the interest rate on dollar-denominated deposits is always equal to the international interestrate (Calvo,1986; Calvoand V(cid:19)egh, 1993). Addinguncertaintymayintroduce an upward slopping path for the nominal interest rate on domestic-currency-denominated assets if the probability of abandonment of the plan increases as the moment of resolution of uncertainty is approached, but the interest rate on dollar-denominated assets is, as in the perfect foresight case, alwaysequal totheinternationalinterestrate(MendozaandUribe, 1996). Therefore,theimperfect credibility hypothesis can explain the increase in the spread between peso and dollar interest rates, but not the escalation in dollar rates that occurred the (cid:12)rst quarter of 1995. Although elements of the two alternative hypothesis discussed above could have been present in the Argentine crisis of 1995, they can hardly explain its timing nor the dramatic rise in real interest rates in the (cid:12)rst quarter of1995. The TequilaorContagionE(cid:11)ect appears to have played a signi(cid:12)cant role. It remains to investigate if a model of the Tequila E(cid:11)ect can provide a satisfactory account of the real e(cid:11)ects associated with the Argentine crisis of 1995. 3 The Basic Setup In this section, I present a simple setup of a small, open economy to analyze the Tequila E(cid:11)ect undercertainty. Thecertaintycaseisauseful(cid:12)rstapproximationtothemorerealisticcaseinvolving uncertainty about the time of the pull out, which I analyze in the next section. Consider a perfect-foresight economy populated by a large number of identical, in(cid:12)nitely lived consumers with preferences de(cid:12)ned over sequences of consumption of a perishable, internationally traded good, ct, and described by the utility function 1 t X(cid:12) U(ct) (1) t=0 where (cid:12) 2 (0;1)denotesthe subjective discount factorand U((cid:1)) denotes the period utilityfunction, assumed to be strictly increasing, strictly concave, and twice continuously di(cid:11)erentiable. 7

In each period t (cid:21) 0, the consumer is endowed with y units of the good and has access to a bond, bt, that pays the real interest rate rt in period t+1. Throughout, the consumer is assumed to enter period 0 with some debt carried over from period (cid:0)1. The consumer’s budget constraint is then given by bt = (1+rt(cid:0)1)bt(cid:0)1+y(cid:0)ct (2) given (1+rt(cid:0)1)b(cid:0)1 < 0: The consumer chooses sequences of consumption and bond holdings so as to maximize his utility function subject to (2) and to a borrowing constraint of the form bt t l ! im 1 t (cid:21) 0; (3) Qj=0(1+rj) that prevents him from engaging in Ponzi-type games. The optimal paths fct;bt+1g satisfy (2) and the following Euler and transversality conditions: 0 0 U (ct)= (cid:12)(1+rt)U (ct+1) (4) bt t l ! im 1 Q t j=0(1+rj) = 0: (5) 3.1 Crisis-free Equilibrium (cid:3) Suppose that agents have access to the international (cid:12)nancial market, where the interest rate, r , (cid:3) 8 is constant and satis(cid:12)es (cid:12)(1+r ) = 1. In this case, the equilibrium conditions are obtained by (cid:3) 9 substituting r for rt in (2){(5) 0 0 U (ct) = U (ct+1) (cid:3) bt = (1+r )bt(cid:0)1+y(cid:0)ct bt t l ! im 1 (1+r (cid:3) ) t = 0 (cid:3) given (1+r )b(cid:0)1 < 0 The (cid:12)rst equilibrium condition implies that consumption is constant over time, while the second and third imply that consumption is given by (cid:3) (cid:3) c = r b(cid:0)1+y: This consumption rule implies that the stock of debt is constant and equal to b(cid:0)1 for all t (cid:21) 0. 8 Theassumptionthattheinterestrateequalstherateoftimepreferenceismadeforsimplicity. Itavoidstheneed to deal with trends in consumption and asset holdings which are not central to the issuesdiscussed in this paper. 9 (cid:3) I am assuming that r(cid:0)1 =r . 8

3.2 The Tequila E(cid:11)ect By the \Tequila E(cid:11)ect," or \exogenous (cid:12)nancial crisis," I mean a situation in which at some point, sayperiod0, domesticagentslearnthatforeign investorswishtopull theirassetsoutofthecountry 10 by time T > 0. Formally, the Tequila E(cid:11)ect sets in when domestic agents learn, at t = 0, that for t (cid:21) T, bt must be greater than or equal to b, where b> b(cid:0)1 and bt denotes aggregate net foreign 11 asset holdings in period t. For simplicity, I assume that b= 0, that is, that foreign investors wish to hold no assets invested in the domestic economy as of period T. Because before period T the country is not subject to any borrowing constraint, the equilibrium domestic interest rate must be equal to the international interest rate for 0 (cid:20) t < T. From period T on, the domestic interest 12 rate must be greater than or equal to the international interest rate. A competitive equilibrium 1 is then a set of sequences fct;bt;rtgt=0 satisfying the consumer’s optimality conditions (2){(5) and the restrictions (cid:3) rt = r t < T (6) (cid:3) (cid:3) rt (cid:21) r , bt (cid:21) 0, (rt(cid:0)r )bt = 0 t (cid:21) T (7) Equations (4) and (6) imply that consumption is constant between periods 0 and T, at a level (cid:3)(cid:3) denoted by c . Because the initial stock of assets is negative ((1+r(cid:0)1)b(cid:0)1 < 0) and the stock (cid:3)(cid:3) (cid:3) of assets in period T must be non-negative (bT (cid:21) 0), c must be less than c , the consumption level under the crisis-free equilibrium, in order for the trade balance to be large enough not only to service the debt but also to cover some amortization each period. It can be shown that the 13 equilibrium conditions (2)-(7) have a unique solution given by (cid:3) (cid:3)(cid:3) (cid:3) r b(cid:0)1 (cid:3) ct = c (cid:17) c + (1+r (cid:3) ) T+1(cid:0)1 < c t (cid:20) T, ct = y t > T, (1+r (cid:3) ) T+1(cid:0)(1+r (cid:3) ) t+1 bt = b(cid:0)1 (1+r (cid:3) ) T+1(cid:0)1 t (cid:20) T, bt = 0 t > T, 0 (cid:3)(cid:3) (cid:3) (cid:0)1 U (c ) (cid:3) rt = r t 6= T, rT = (cid:12) U 0 (y) (cid:0)1> r . 10 Throughoutthepaperthebehaviorofforeigninvestorsistakenasgiven. Calvo(1995),usingapartialequilibrium model of optimal portfolio management, provides theoretical arguments for why it might be optimal for foreign investorsto behavein a fashion consistent with the Tequila E(cid:11)ect hypothesis. 11 Note thatthe notation does not distinguish between individual and aggregate netasset holdings. I assume that the borrowing constraint is imposed on the aggregate level of net asset holdings rather than on individual net asset holdings. Because consumers are assumed to be identical, however, this distinction does not a(cid:11)ect the equilibrium. The distinction would make a di(cid:11)erence if agentswere assumed to be heterogeneous in preferences orendowments. 12 Thedomesticinterestratemaynotbelowerthantheinternationalinterestrate,forsuchasituationwouldallow agents to make in(cid:12)nite pro(cid:12)ts by exploiting arbitrage opportunities. 13 The uniquenessofthe competitive equilibrium can be seen by noting that conditions (2)-(7)are identical to the Kuhn-Tucker conditions associated with the social planner’s problem of maximizing the utility function (1) subject to (cid:3) bt (cid:20)(1+r )bt(cid:0)1+y(cid:0)ct bt(cid:21)0 t(cid:21)T (cid:3) given r b(cid:0)1 <0. This problem,in turn, has a unique solution because the objective function is strictly concave and the budget set is convex. 9

Theequilibriumpathsofforeigncapital(cid:13)ows,theinterestrate,and consumption are illustrated in (cid:12)gure 6. Foreign capital starts (cid:13)owing out of the country at the beginning of period zero and (cid:13)ows at an increasing rate until period T, when there is an interest rate hike and the country goes into autarky. Interestingly, the jump in the interest rate occurs only in period T, though all the information about foreign investors’ loss of con(cid:12)dence is revealed in period 0. The capital (cid:13)ight is accompanied by a corresponding improvement in the trade balance in period 0 that lasts until period T. Note that the equilibrium is una(cid:11)ected by whether the decision by foreign investors to pull out in period T is permanent or transitory in the sense that whether or not agents can borrow from the rest of the world from period T +1 on is immaterial for the behavior of the variables of the model. Which features of the Argentine recession of 1995 are captured by the Tequila E(cid:11)ect model? If one thinks of t = 0 as December 1994, when the Mexican crisis occurred, and of t = T as March 1995, when the interest rate hike occurred, the Tequila E(cid:11)ect hypothesis explains three important aspects of the 1995 Argentine recession: the sudden improvement in the trade balance in the (cid:12)rst quarter of 1995; the capital (cid:13)ight that started in late December 1994 and that was re(cid:13)ected in the loss of reserves by the central bank and the decline in both peso- and dollar-denominated bank deposits; and the interest rate hike in March 1995, which marked the end of the initial phase of the crisis. However, the simple model of the Tequila E(cid:11)ect fails to explain the observed deepening of the contraction in aggregate domestic absorption at the time of the interest rate hike. Instead, the model predicts a constant trade balance between periods 0 and T and a deterioration in period T +1 as consumption jumps to the endowment level. 4 Uncertain Timing of Crisis In this section, I modify the baseline model by assuming that the time at which the domestic economy is subject to the aggregate borrowing constraint is uncertain. This modi(cid:12)cation seeks to capture the deepening of the contraction in aggregate domestic absorption and the widening of the trade balance observed in Argentina at the time of the interest rate hike in March 1995. Assumethatthereexistsanexogenousrandomvariablest thattakesthevalue1iftheaggregate 14 borrowingconstraintisinplaceand0otherwise. Thatis,ifst = 1,thenbt (cid:21) 0(butnotnecessarily theotherwayaround,becausethecountry’snetforeignassetholdingscaninprinciplebevoluntarily non-negative at any time). Also assume that the borrowing constraint is imposed for sure no later than period T and that once the aggregate borrowing constraint is imposed it remains in place forever. Formally, the random variable st is assumed to satisfy Pr(st = 1jst(cid:0)1 = 1) = 1 8t and Pr(st = 1) = 1 for t (cid:21) T. In this case, the equilibrium conditions are given by equation (2) and the following modi(cid:12)ed 14 The methodology used in this section follows Drazen and Helpman (1988) and Calvo and Drazen (1993). 10

version of conditions (4){(7) 0 0 U (ct)= (cid:12)(1+rt)EtU (ct+1); (8) bt+h h l ! im 1 Et Q h j=0(1+rt+j) = 0; (9) (cid:3) (cid:3) (cid:3) rt (cid:21) r , (rt(cid:0)r )(1(cid:0)st)= 0, (rt(cid:0)r )bt = 0, btst (cid:21) 0: (10) The second expression in (10) states that if the borrowing constraint is not in place, the domestic (cid:3) interest rate must be equal to the international interest rate r ; the last expression in (10) states that aggregate net asset holdings must be non-negative at the moment the aggregate borrowing constraint is imposed. In characterizing the dynamics arising from these equilibrium conditions, it is useful to introduce some notation to distinguish the values taken by ct, bt and rt in a pre-crisis pc period (st = 0) from those taken by ct, bt and rt in the crisis period (st = 1jst(cid:0)1 = 0). Let ct c denote the value taken by consumption in period t in the event st = 0, and ct denote the value pc c pc c taken by consumption in the event st = 1jst(cid:0)1 = 0. De(cid:12)ne bt , bt, rt , and rt in a similar fashion. Figure 14 illustrates the equilibrium dynamics of capital out(cid:13)ows, the real interest rate, and consumption. Solid lines show pre-crisis values, broken lines show values in the period of the crisis (i.e., the period in which the borrowing constraint is imposed), and chain-dotted lines show values after the occurrence of the crisis. (Note that since any period in the interval (0;T) is either a pre-, at-, or post-crisis period, each panel of the (cid:12)gure displays three di(cid:11)erent lines in this interval.) As shown in the bottom left panel of the (cid:12)gure, consumption falls not only in period 0|as in the perfect foresight case|but also at the time of the crisis. The second decline is represented by the vertical distance between thesolid and thebroken lines. Consumptionrecoversone period after the crisis. Because at this point the economy is in autarky, consumption jumps to the endowment level. Before the crisis consumption is increasing over time, re(cid:13)ecting the positive income e(cid:11)ect associated with the non-occurrence of the interest rate hike. This increase is represented by the vertical distance between thebroken and thechain-dotted lines. The domesticinterestrate isequal to the international interest rate until the moment at which the aggregate borrowing constraint is imposed, and then it jumps up. Note that the interest rate hike is more pronounced the earlier the borrowing constraint is imposed. Finally, capital out(cid:13)ows are positive and increasing over time, and they jump up when the borrowing constraint is imposed. The intuition behind these results is the following: In any period 0 < t < T (cid:0)1 in which the borrowing constraint has not yet been imposed, the probability that it will be imposed in period t+1 is less than 1. Because consumers are assumed to be net borrowers from the rest of the world, it is optimal for them to save less than if they knew that the borrowing constraint will be imposed for sure in t+1 but more than if they knew for sure that it will not be imposed. If the aggregate borrowingconstraintisimposed inperiod t+1, agents ex post undersaved and hencehavetoadjust their consumption downward. If, on the other hand, the borrowing constraint is not imposed in period t+1, agents ex post oversaved and decide to spend part of that excess savings in the current period. This explains why consumption increases over time before the interest rate hike and falls 11

in the period of the interest rate hike. The following proposition provides a formal characterization of the equilibrium. Proposition 1 (Equilibrium Under Uncertain Timing of Crisis) Let ct, bt, and rt denote consumption, net aggregate asset holdings, and the domestic real interest rate in period t respectively. Let b(cid:0)1 < 0. Let st be an exogenous random variable taking values pc in f0;1g such that Pr(st = 1) = 1 for t (cid:21) T and Pr(st = 1jst(cid:0)1 = 1) = 1 for t (cid:21) 1. Let ct c denote consumption in period t in the event st = 0 and ct consumption in period t in the event pc c pc c st = 1jst(cid:0)1 = 0. De(cid:12)ne bt ,bt rt ,rt analogously. Then the equilibrium conditions (2) and (8)-(10) imply that (cid:3) (a) c0 < c(cid:0)1 = r b(cid:0)1+y, c pc c pc (b) ct < ct(cid:0)1 for 0(cid:20) t < T and cT = cT(cid:0)1, c pc (c) bt = 0 and bt < 0 8t, pc c (d) ct and cc are strictly increasing in t, c c (cid:3) (e) rt > rt+1 > r 8t (cid:21) 0. Proof: See the appendix. 5 The Tequila E(cid:11)ect in a Production Economy The two versions of the Tequila E(cid:11)ect model analyzed thus far capture several aspects facts of the Argentine recession of 1995 but leave other important facts unexplained. For example, the models predict that the real interest rate jumps at the time at which foreign investors pull out of the domestic economy and returns immediately to its pre-crisis level. However, the data shows that after the interest rate hike in March 1995, interest rates came down slowly and remained above their pre-crisis levels throughout 1995 ((cid:12)gure 6). Also, because they consider endowment economies,the modelspresented thusfarsay nothingaboutthe behavior of outputandinvestment, though the data shows that after the credit crunch of March 1995, output and investment declined sharply, and that, as shown in (cid:12)gure 10, the slowdown in output continued through 1995. In this section, I augment the baseline model with production and capital accumulation and argue that thisvariation hasthe ability toqualitativelycapture the observed behavior ofthe real interestrate, output, and investment. Consider the same structure described in section 3 but assume that output, yt, instead of being exogenously given, is produced with physical capital, kt, which depreciates at the constant rate (cid:14) 2(0;1]. Speci(cid:12)cally, yt = F(kt) kt+1 = (1(cid:0)(cid:14))kt+it where it denotesgross investmentand F((cid:1))isaproduction function satisfyingtheInadaconditions. 12

Letting f(k)(cid:17) F(k)+(1(cid:0)(cid:14))k, the household’s stock of net foreign assets evolves according to bt = (1+rt(cid:0)1)bt(cid:0)1+f(kt)(cid:0)ct(cid:0)kt+1 (11) kt+1 (cid:21) 0 (12) (1+r(cid:0)1)b(cid:0)1 < 0 and k0 > 0 Households choose sequences of consumption, capital and bonds so as to maximize the utility function (1) subject to (3), (11), (12), and the initial conditions for bond holdings and capital, taking as given the sequence for the real interest rate. The (cid:12)rst order conditions associated with this problem are equations (4), (5), (11), (12), and the following condition stating that in each period the household invests until the next period’s marginal product of capital equals the current real interest rate 0 f (kt+1)= 1+rt: (13) 1 A perfect-foresight equilibrium is a set of sequences fct;bt;rt;kt+1gt=0 satisfying these (cid:12)rst order conditions and expressions (7) and (6). Expressions (7) and (6) re(cid:13)ect the restrictions on rt and bt imposed by the aggregate borrowing constraint in place from period T on. It can be shown that 15 these equilibrium conditions have a unique solution. Figure15showstheequilibriumpathsofthevariablesofinterest. Asintheendowmenteconomy, consumptionfallsin period0andforeigncapitalstartsto(cid:13)owoutoftheeconomy. Between periods (cid:3) 0 and T (cid:0)1 the interest rate is constant and equal to the international interest rate r , implying that consumption and investment are also constant during this period (see equations (4)and (13)). In period T the interest rate jumps up, provoking a decline in investment. Therefore the economy starts period T +1 with a lower stock of capital. Because from this point on the country is unable to borrow from the rest of the world, the dynamics toward the steady state are identical to those (cid:3) in a closed economy. Thus, the capital stock converges gradually and from below to k , and the interest rate, given by the marginal product of capital, converges monotonically and from above to (cid:3) r . The prediction of a slow return of the interest rate to its pre-crisis level is a key di(cid:11)erence with those of an endowment economy which predict an interest rate jumps in period T but an (cid:3) 16 immediate return to its long-run level r . Thus, allowing for capital accumulation helps explain 15 Asin thecaseof theendowmenteconomy,theproofofthis statement followsfrom thefact thattheequilibrium conditions(4)-(7)and(11)-(13)areidenticaltothoseassociatedwiththeplanner’sproblemofmaximizing(1)subject to (cid:3) bt (cid:20)(1+r )bt(cid:0)1+f(kt)(cid:0)ct(cid:0)kt+1 kt+1 (cid:21)0 bt(cid:21)0 t(cid:21)T which has a unique solution because the objective function is strictly concave and the budget set is convex. 16 (cid:3) As noted in section 4, the interest rate also returns immediately to r in the endowment economy model with uncertain timing ofcrisis. 13

the prolonged period of high real interest rates observed in Argentina after the Mexican crisis of December 1994. Finally, both this model with perfect-foresight and capital accumulation and the endowment-economymodel with uncertain timingof crisis predict an improvement in the trade balancedrivenbyacontractioninaggregatedomesticspendingatthemomentatwhichthecountry becomessubjecttothe borrowingconstraint. However,the natureofthisimprovementin thetrade balance is quite di(cid:11)erent in the two cases. Under uncertain timing of crisis the improvement is due entirely to a decline in consumption. In the model with perfect-foresight and capital accumulation, consumption does not fall in period T but investment does, owing to the increase in the interest rate. The following proposition provides a formal characterization of the equilibrium: 1 Proposition 2 The sequences fct;bt;rt;kt+1gt=0 satisfying the equilibrium conditions (4)-(7)) and (11)-(13) are given by (cid:3)(cid:3)(cid:3) ct = c < c(cid:0)1 for t (cid:20) T, ct = f(kt)(cid:0)s(kt) for t > T, (cid:3) (cid:3) (cid:3) kt+1 = k for t < T, kT+1 < k , kt < kt+1 = s(kt)< k for t > T, (cid:3) 0 (cid:3) rt = r for t < T, rt = f (kt+1)> r for t (cid:21) T, bt = (1+r (cid:3) ) t+1 b(cid:0)1+ (1+r (cid:3) r ) (cid:3) t+1(cid:0)1 [f(k (cid:3) )(cid:0)c (cid:3)(cid:3)(cid:3) (cid:0)k (cid:3) ] for t < T, bt = 0 for t (cid:21) T, (cid:3)(cid:3)(cid:3) where c and kT+1 solve the following two equations in two unknowns (the (cid:12)rst of which ensures that bT = 0 and the other that the Euler equation (4) holds in T), (cid:3) (cid:3) (cid:3)(cid:3)(cid:3) 0 = (1+r )bT(cid:0)1+f(k )(cid:0)c (cid:0)kT+1 (14) 0 (cid:3)(cid:3)(cid:3) 0 0 U (c ) = (cid:12)f (kT+1)U (f(kT+1)(cid:0)s(kT+1)) (15) and s((cid:1)) is the familiar closed economy neoclassical savings function, that is, for any k0 > 0 kt+1 = s(kt) solves the following di(cid:11)erence equation 0 0 0 U (f(kt)(cid:0)kt+1)= (cid:12)f (kt+1)U (f(kt+1)(cid:0)kt+2) (16) (cid:3) lim kt = k (17) t!1 Proof: See the appendix. 6 Conclusion December 1994 marked the end of four years of extraordinary economic expansion of the Argentine economy under the Convertibility Plan. According to the Tequila E(cid:11)ect (or Contagion E(cid:11)ect) hypothesis, the (cid:12)nancial and economic crises that followed had their roots in the collapse of the 14

Mexican peso which undermined foreign investors’ con(cid:12)dence about the prospects for sustained stability and growth in Argentina and other Latin American countries and led to massive capital out(cid:13)ows. Several conclusions arise from the empirical and theoretical investigation of the Tequila E(cid:11)ect that I perform in this paper. First, the empirical analysis suggests that the Tequila E(cid:11)ect played an important role in the 1995 crisis. Second, the two most popular explanations of business cycles in developing countries are shown to be unable to fully account for the developments following the Mexican crisis of 1995: the explanation based on movements in international interest rates as the main determinant of economic (cid:13)uctuations in LDCs fails to explain the timing of th crisis, because while international interest rates began rising in early 1994, international capital started to (cid:13)ow out of Argentina in early 1995. An explanation based on imperfect credibility about the exchange rate policy fails to explain the dramatic increase in interest rates on dollar-denominated loans and deposits that took place in the (cid:12)rst quarter of 1995. Finally, the predictions of a simple model of the Tequila E(cid:11)ect in which domestic residents expect that at some point in the future foreign investors will stop renewing their loans to the country are shown to be consistent with many aspects of the Argentine recession of 1995, namely, the capital out(cid:13)ows right after the Mexican crisis, the corresponding trade balance improvement, and the interest rate hike in March 1995. Adding to the basic model uncertainty about the time at which foreign investors will actually pull out helps capture the deepening of the contraction in aggregate domestic spending that occurred at the time of the credit crunch of March 1995. Allowing for capital accumulation helps explain the observed slow return of the real interest rate to its pre-crisis level and the prolonged slowdown in output and investment. 15

Appendix Proof of Proposition 1: pc pc (c) Letus(cid:12)rstshowthatbt < 08t. Supposeitisnot. Then9t < T suchthatbt (cid:21) 0. Butthen (cid:3) (cid:3) if st = 0, the sequences bt+j = bt(cid:0)1, ct+j = r bt(cid:0)1 +y, and rt+j = r , j (cid:21) 0 satisfy conditions (2) pc and (8){(10) and therefore represent an equilibrium. Hence, bt (cid:21) 0 implies bt(cid:0)1 (cid:21) 0, which c leads, by repeating this reasoning, to b(cid:0)1 (cid:21) 0, a contradiction. Let us now show that bt = 0 c pc 8t. Again, suppose it is not. Then 9t (cid:21) 0 such that bt > 0. Because bt(cid:0)1 < 0, we have that c (cid:3) pc c c c (cid:3) ct = (1 + r )bt(cid:0)1 + y (cid:0) bt < y. Also, bt > 0 implies, from (10), that rt = r , and therefore c (cid:3) c (cid:3) c equation (8) implies that if st = 1, then ct+1 = ct < y and bt+1 = (1+r )bt+y(cid:0)ct+1 > (1+r )bt. c (cid:3) j c Following this argument leads to the conclusion that if st = 1 and bt > 0, then bt+j > (1+r ) bt, which violates the transversality condition (9). c pc pc (cid:3) (b) The equality cT = cT(cid:0)1 follows from equation (8) and from the facts that rT(cid:0)1 = r and c (cid:3) pc c that cT is known in period T (cid:0) 1. From equation (2) we have that bt = (1 + r )bt(cid:0)1 + y (cid:0) ct pc (cid:3) pc pc c and bt = (1+r )bt(cid:0)1 + y (cid:0) ct . Combining these two equations and recalling that bt = 0 and pc c pc pc bt < 0 yields ct < ct for t < T. On the other hand, because (cid:12)rt = 1, equation (8) implies that 0 pc 0 pc 0 c pc c U (ct ) is a weighted average of U (ct+1) and U (ct+1); and because ct+1 > ct+1, it follows that pc pc c ct < ct+1 < ct+1. c pc (a) To see that consumption falls on impact, note that because c0 < c0 , it is su(cid:14)cient to show pc pc pc pc (cid:3) that c0 < c(cid:0)1. Suppose it is not. Then because ct > ct(cid:0)1, we have that ct (cid:21) c(cid:0)1 = r b(cid:0)1 +y, pc c pc and from equation (2) it follows that bt (cid:20) b(cid:0)1. This and the fact that cT = cT(cid:0)1 implies, again c from (2), that bT (cid:20) b(cid:0)1 < 0, which violates the borrowing constraint. pc pc c (cid:3) pc (d) I have already shown that ct+1 > ct for t (cid:21) 0. Because ct = (1+r )bt(cid:0)1+y, showing that c c pc pc pc ct+1 > ct for t (cid:21) 0 is equivalent to showing that bt > bt(cid:0)1 for t (cid:21) 0 (where b(cid:0)1 (cid:17) b(cid:0)1). Combining c pc c (cid:3) pc pc (cid:3) pc pc cT = cT(cid:0)1 and cT = (1 + r )bT(cid:0)1 + y with equation (2) yields bT(cid:0)1 = (1 + r )(bT(cid:0)2 (cid:0) bT(cid:0)1), pc pc (cid:3) pc pc which implies that bT(cid:0)1 > bT(cid:0)2 and that r bT(cid:0)2 +y > cT(cid:0)1. This last expression, together with pc pc (cid:3) pc pc equation (2), in turn implies that bT(cid:0)2 > bT(cid:0)3 and that r bT(cid:0)3+y > cT(cid:0)2. Continuing with this pc pc argument leads to bt > bt(cid:0)1 8t (cid:21) 0. 0 c c 0 (e) Statement(e) followsdirectlyfromthe Euler equation U (ct)= (cid:12)(1+rt)U (y) and the facts c that ct is less than y and is increasing over time. Proof of Proposition 2: (cid:3) (cid:3)(cid:3)(cid:3) It is su(cid:14)cient to show that rt (cid:21) r t (cid:21) T and that c < c(cid:0)1, for all other equilibrium (cid:3) conditions are satis(cid:12)ed by construction. Equation (13) implies that rt (cid:21) r t (cid:21) T if and only (cid:3) if kt+1 (cid:20) k for t (cid:21) T. In turn, it is a well-known result that the system (16){(17) implies (cid:3) (cid:3) that kt will converge monotonically to k from below if and only if kT+1 (cid:20) k . I now show that (cid:3)(cid:3)(cid:3) equations (14) and (15) imply that this condition is in fact satis(cid:12)ed. and that c < c(cid:0)1 where (cid:3) (cid:3) (cid:3) c(cid:0)1 = r b(cid:0)1+f(k )(cid:0)c(cid:0)1(cid:0)k . (cid:3)(cid:3)(cid:3) Figure6depictsthetwosetsofpairs(c ;kT+1)satisfyingequations(14)and (15)individually. (cid:3)(cid:3)(cid:3) BecausebT(cid:0)1isadecreasingfunctionofc ,equation(14)impliesthatkT+1isadecreasingfunction (cid:3)(cid:3)(cid:3) (cid:3) (cid:3) of c . This relation is shown as line LL’ on the (cid:12)gure. Also, because c(cid:0)1 = r b(cid:0)1+f(k ), we (cid:3)(cid:3)(cid:3) (cid:3) have that for c = c(cid:0)1, the value of kT+1 that satis(cid:12)es (14) is k +b(cid:0)1 (point a on the (cid:12)gure). On the other hand, because f((cid:1))(cid:0) s((cid:1)) is a strictly increasing function, it follows that the right hand side of (15) is a decreasing function of kT+1. Therefore, equation (15) implies that kT+1 (cid:3)(cid:3)(cid:3) (cid:3)(cid:3)(cid:3) is an increasing function of c (line EE’ on the (cid:12)gure). I now show that if c = c(cid:0)1, then (cid:3) (cid:3) k +b(cid:0)1 < kT+1 < k , that is, that the locus satisfying equation (15) passes through a point like b on (cid:12)gure 6. This, in turn, implies that the two loci intersect at a point like c located southwest of (cid:3) (cid:3) 0 (cid:3) the point (c(cid:0)1;k ). First, evaluate the right hand side of (15) at kT+1 = k . Because (cid:12)f (k ) = 1 16

(cid:3) (cid:3) (cid:3) (cid:3) (cid:3) 0 and f(k )(cid:0)s(k )= f(k )(cid:0)k = c(cid:0)1(cid:0)r b(cid:0)1 > c(cid:0)1,the right hand side of (15)is less than U (c(cid:0)1), (cid:3)(cid:3)(cid:3) (cid:3) and, hence, for c = c(cid:0)1, equation (15) implies that kT+1 < k . Now evaluate the right hand (cid:3) 0 (cid:3) (cid:3) (cid:3) side of (15) at kT+1 = k + b(cid:0)1. Because (cid:12)f (k + b(cid:0)1) > 1 and f(k + b(cid:0)1)(cid:0) s(k + b(cid:0)1) < (cid:3) (cid:3) (cid:3) (cid:3) (cid:3) (cid:3) f(k )+(1+r )b(cid:0)1(cid:0)(k +b(cid:0)1) = f(k )(cid:0)k +r b(cid:0)1 = c(cid:0)1 (the last inequality follows from the (cid:3) strict concavity of f((cid:1)) and from the fact that s(k) > k for k < k ), the right hand side of (15) is 0 (cid:3)(cid:3)(cid:3) (cid:3)(cid:3)(cid:3) (cid:3) greater than U (c ) and, hence, for c = c(cid:0)1, equation (15) implies that kT+1 > k +b(cid:0)1. This (cid:3) (cid:3)(cid:3)(cid:3) completes the proof that kT+1 < k and c < c(cid:0)1. 17

References Calvo, Guillermo A., \Temporary Stabilization: Predetermined Exchange Rates," Journal of Political Economy, 1986, 94, pp. 1319-29. Calvo, Guillermo A., \Balance-of-Payments Crises in a Cash-in-Advance Economy," Journal of Money, Credit and Banking, February 1987, 19, pp. 19-32. Calvo, Guillermo A., \Varieties of Capital Market Crises," mimeo, Center for International Economics, University of Maryland at College Park, 1995. Calvo, Guillermo and Allan Drazen, \Uncertain Duration of Reform: Dynamic Implications," Working Papers in International Economics # WP4, Center for International Economics, University of Maryland at College Park, 1993. Calvo, G. A., L. Leiderman and C. M. Reinhart, \Capital In(cid:13)ows and Real Exchange Rate Appreciation in Latin America: The Role of External Factors," IMF Sta(cid:11) Papers, March 1993, 40(1), pp. 108-51. Calvo, Guillermo A. and Carlos A. V(cid:19)egh, \Exchange Rate Based Stabilization Under Imperfect Credibility," inHelmutFrischandAndreasWorgotter(Eds.),,OpenEconomy Macroeconomics, London, MacMillan Press, 1993, Chapter 1, pp. 3-28. Calvo, Sara and Carmen M. Reinhart, \Capital Flows to Latin America: Is There Evidence of Contagion E(cid:11)ects?/," forthcoming in Morris Goldstain, Editor,, Private Capital Flows to Emerging Markets, Institute for International Economics, Washington, DC, 1996. Dornbusch,Rudiger,\StabilizationPolicyinDevelopingCountries: WhatLessonsHaveWeLearnt?," World Development, September 1982, special issue 10, pp. 701-8. Dooley, Michael, Eduardo Fern(cid:19)andez-Arias, and Kenneth Kletzer, \Is the Debt Crisis History? Recent Private Capital In(cid:13)ows to Developing Countries," The World Bank Economic Review, January 1996, 10 (1), pp. 27-50. Drazen,Allanand ElhananHelpman, \StabilizationwithExchangeRateManagement," Quarterly Journal of Economics, 1987, 11, pp. 835-55. Drazen, Allan and Elhanan Helpman, \Stabilization with Exchange Rate Management Under Uncertainty," in Elhanan Helpman, Assaf Razin and Efraim Sadka (Eds.), Economic E(cid:11)ects of the Government Budget,, Cambridge, The MIT Press 1988: 310-27. Edwards,Sebastian, \PublicSectorDe(cid:12)citsandMacroeconomicStabilityinDevelopingEconomies," NBER working paper #5407, January 1996. Kiguel,MiguelA.andNissanLiviatan,\TheBusinessCycleAssociatedwithExchange-Rate Based Stabilizations," The World Bank Economic Review, 1992, 6, pp. 279-305. Lahiri, Amartya, \Exchange-Rate Based Stabilization Under Real Frictions: The Role of Endogenous Labor Supply," mimeo Department of Economics UCLA, 1995. Mendoza, Enrique G. and Mart(cid:19)(cid:16)n Uribe, \The Syndrome of Exchange-Rate Based Stabilizations and the Uncertain Duration of Currency Pegs," International Finance Discussion Paper #548, Board of Governors of the Federal Reserve System, Washington DC, April 1996. Rebelo, Sergio T. and Carlos A. V(cid:19)egh, \Real E(cid:11)ects of Exchange-Rate Based Stabilization: An Analysis of Competing Theories," mimeo Department of Economics, University of Rochester, 1995. 18

Reinhart Carmen M. and Carlos A. V(cid:19)egh, \Nominal InterestRates, Consumption Booms and Lack of Credibility: A Quantitative Examination," Journal of Development Economics, April 1995, 40(2), pp. 357-78. Rodr(cid:19)(cid:16)guez, Carlos A., \The Argentine Stabilization Plan of December 20th," World Development, September 1982, special issue 10, pp. 801-11. Rold(cid:19)os, Jorge E., \Supply Side E(cid:11)ects of Disin(cid:13)ation Programs," IMF Sta(cid:11) Papers, March 1995, 42(1), pp. 158-83. Uribe, Mart(cid:19)(cid:16)n, \Exchange-Rate Based In(cid:13)ation Stabilization: The Initial Real E(cid:11)ects of Credible Plans," mimeo Department of Economics, The University of Chicago, February 1993. V(cid:19)egh, Carlos A., \Stopping High In(cid:13)ation: An Analytical Overview," IMF Sta(cid:11) Papers, September 1992, 39(3), pp. 626-95. 19

Table 1: ARGENTINA: FISCAL SURPLUS OF THE FEDERAL GOVERNMENT (1985-1992) (in percentage of GDP) 1985 1986 1987 1988 1989 1990 1991 1992 1. Primary surplus before income from sale of assets 1.9 1.5 -1.1 -1.4 0.7 1.3 1.1 2.2 2. Income from sale of assets 0.2 0.1 0.2 0.2 0.3 0.6 1.4 1.2 3. Primary Surplus of the Non-Fin ancial Public Sector (1+2) 2.1 1.6 -0.9 -1.2 1.0 1.9 2.5 3.4 4. Interest of the Non - Financial Public Sector 5.9 4.3 4.7 5.4 15.6 4.5 3.7 2.2 5. Total Surplus of the Non-Financial Public Sector (3+4) -3.8 -2.7 -5.6 -6.6 -14.6 -2.6 -1.2 1.2 6. Surplus of the Central Bank -2.8 -1.6 -0.9 -0.7 -5.9 -1.0 -0.5 -0.2 7. Total Surplus of the Public Sector (5+6) -6.6 -4.3 -6.5 -7.3 -20.5 -3.6 -1.7 1.0 8. Surplus of the Public sector before sales of assets (7-2) -6.8 -4.4 -6.7 -7.5 -20.8 -4.2 -3.1 -0.2 Source: Argentine Ministry of Economics. 20

Table 2: ARGENTINA: NUMBER OF PUBLIC EMPLOYEES Federal Administration April 1990 December 1991 Central Administration 70,968 51,316 Decentralized Adminst ration and other special accounts 213,927 176,437 Agencies not included in the Budget 49,392 28,489 Total 334,297 256,242 Source: Argentine Ministry of Economics. 21

Figure 1: ARGENTINA: INFLATION (CPI, monthly percentage rates) 40 35 30 25 20 15 10 5 80 85 90 95 Note: The two vertical lines mark the pre Mexican crisis phase of the Convertibility Plan. 22

Figure 2: ARGENTINA: REAL EXCHANGE RATE (1991=1) 4 3.5 3 2.5 2 1.5 1 80 85 90 95 Note: (1) The two vertical lines mark the pre-Mexican crisis phase of the Convertibility (cid:1) us ar plan. (2) The real exchange rate is measured as E CPI =CPI , where E is the us ar pesos-per-dollar nominal exchange rate and CPI and CPI are US and Argentine Consumer Price Indexes respectively. 23

Figure 3: ARGENTINA: AGGREGATE SUPPLY AND DEMAND GDP at market prices Consumption/GDP 90 13 85 12 11 80 10 75 9 70 1980 1985 1990 1995 1980 1985 1990 1995 % Gross Dom. Investment/GDP 25 20 15 1980 1985 1990 1995 % Trade Balance/GDP 6 4 2 0 −2 −4 1980 1985 1990 1995 % Note: The two vertical lines mark the pre Mexican crisis phase of the Convertibility Plan. GDP is in million of pesos of 1986. The 1995 (cid:12)gure for trade balance is an estimate by FIEL. 24

Figure 4: ARGENTINA: BANK DEPOSITS (monthly average, in billions of current dollars) 26 24 22 peso dep. 20 18 16 dollar dep. 14 12 93 94 95 96 Note: The vertical line marks the Mexican crisis of December 1994. Figure 5: ARGENTINA: BANK LOANS (monthly average, in billions of current dollars) 38 36 34 dollar loans 32 30 28 26 24 peso loans. 22 20 93 94 95 96 Note: The vertical line marks the Mexican crisis of December 1994. 25

Figure 6: ARGENTINA: NOMINAL LOAN INTEREST RATES (30-day prime rate, in percentage per annum) 35 30 Peso rate 25 20 15 Dollar rate 10 94 95 Note: The vertical line marks the Mexican crisis of December 1994. Figure 7: ARGENTINA: MERVAL STOCK INDEX (in current dollars) 900 800 700 600 500 400 300 200 93 94 95 96 Note: The vertical line marks the Mexican crisis of December 1994. 26

Figure 8: ARGENTINA: INTERNATIONAL RESERVES (monthly average, in billions of current dollars) 17 16 15 14 13 12 11 10 93 94 95 96 Note: The vertical line marks the Mexican crisis of December 1994. Figure 9: ARGENTINA: DOMESTIC CREDIT (daily, In millions of current dollars) 2000 1800 1600 1400 1200 1000 800 600 400 200 1/2/95 1/16 2/1 2/15 3/1 3/14 Note: Domestic Credit is de(cid:12)ned as Monetary Base + Dollar Deposits of banks in the central bank - Total International Reserves of the central bank. 27

Figure 10: ARGENTINA: INDUSTRIAL PRODUCTION (4-quarter growth rate, in percent) 8 6 4 2 0 −2 −4 −6 −8 −10 93 94 95 96 Note: The vertical line marks the Mexican crisis of December 1994. Figure 11: ARGENTINA: EXPORTS AND IMPORTS (in billions of current dollars) 2.2 imports 2 1.8 1.6 1.4 1.2 1 exports 93 94 95 96 Note: The vertical line marks the Mexican crisis of December 1994. 28

Figure 12: U.S. INTENDED FEDERAL FUNDS RATE (End of month, in percent) 6.5 6 5.5 5 4.5 4 3.5 3 93 94 95 96 Note: The vertical line marks the Mexican crisis of December 1994. 29

Figure 13: THE TEQUILA EFFECT IN A PERFECT-FORESIGHT, ENDOWMENT ECONOMY capital outflows interest rate (b−b ) t t−1 • r* 0 T t 0 T t consumption y c* c** 0 T t 30

Figure 14: THE TEQUILA EFFECT UNDER UNCERTAINTY capital outflows interest rate (b−b ) t t−1 r* 0 T t 0 T t consumption y pre−crisis values c* at−crisis values after−crisis values 0 T t 31

Figure 15: THE TEQUILA EFFECT IN A PRODUCTION ECONOMY capital outflows interest rate (b−b ) t t−1 r* 0 T t 0 T t consumption output y f(k*) c* c** 0 T t 0 T t 32

Figure 16: PERIOD-T CONSUMPTION AND INVESTMENT IN THE PRODUCTION ECONOMY L E¢ k* • b c k • T+1 k*+b •a −1 E L¢ c*** c −1 33