This item is only available as the following downloads:
! FIXED AGAINST FLOATING: AN INVESTIGATION OF THE STABILITY OF BARBADOS' FIXED EXCHANGE RATE REGIME BY TRISTAN A. ZUCKER A Thesis Submitted to the Division of Social Sciences New College of Florida I n partial fulfillment of the requirements for the degree Bachelor of Arts Under the sponsorship of Dr. Tarron Khemraj Sarasota, Florida May 2012
ii Acknowledgements I would like to thank the following people and institutions for thei r contributions to this thesis The Central Bank of Barbados, Delaney Anderson, Sir Courtney Blackman, Governor DeLisle Worrell, Professor Duff Cooper, Emily Martin, Jacquelyn Bolles, Dr. Michael Campbell, New College of Florida, Professor Patrick Van Horn, Dr. Roland Craigwell, Professo r Tarron Khemraj, and Dr. Winston Moore
iii Contents Acknowledgements . . . . . . . ii Contents . . . . . . . . iii List of Illustrations . . . . . . . v List of Tables . . . . . . . . vi Abstract . . . . . . . . v ii I ntroduction . . . . . . . . 1 1 Literature Review and Barbadian Context . . . 4 1 .1 Role and Importance of the Foreign Exchange Rate . . 4 1. 2 Foreign Exchange Rate Determination Models . . . 5 1.2.1 National Price Levels / Purchasing Power Parity . . 5 1.2.2 Interest Rate Parity Models . . . . . 7 1.2.3 Flexible Price Monetary Model . . . . 10 1.2.4 Sticky Price Monetary Model . . . . 13 1.2.5 General Equilibrium Monetary Model . . . 16 1.2.6 Portfolio Balance Model . . . . . 1 8 1.2.7 J C urve Effect . . . . . . 21 1.3 Exchange Rate Crisis Models . . . . . 22 1.3.1 First Generation Models . . . . . 24 1.3.2 Second Generation Models . . . . 29 1.3.3 Third Generation Models . . . . . 35 1. 4 Indicators o f Exchange Rate Crises . . . . 36 1.5 Barbadian Context 1991 Currency Crisis . . . 42 2 M ethodology . . . . . . . 45 2.1 Introduction . . . . . . . 45 2.2 Signal Extraction Approach . . . . . 46 2.2.1 Scope of the Model . . . . . 47 2.2.2 Definition of a Currency Crisis . . . . 48 2.2.3 Generation of Predictions . . . . . 49 2.2.4 Justification of Applying the Approach to the Thesis Question 50 2.3 Model Specifications of the Thesis Approach . . . 51 2.3.1 Indicato rs . . . . . . . 51 2.3.2 Definition of a Currency Crisis . . . . 56 2.3.3 Crisis Window Selection . . . . . 57 2.3.4 Comparison Country Selection . . . . 58 2.3.5 Currency Crisi s Selection . . . . . 59 2.3.6 Data Range Selection . . . . . 60 2.3.7 Indicator Threshold Selection . . . . 60 2.4 Modification of the Signal Extraction Approach . . . 61 2.5 Application of the Signal Extraction Approach . . . 61 3 Presentation and Analysis of Results . . . 64 3.1 Introduction . . . . . . . 64 3.2 Credit to the Public Sector . . . . . . 64
iv 3.3 Domestic Credit . . . . . . . 65 3.4 Foreign Exchange Reserves . . . . . 67 3.5 M1 . . . . . . . . . 68 3.6 M2 / Foreign Exchange Reserves . . . . . 69 3.7 Real Exchange Rate Deviation from Trend . . . 71 3.8 Real Interest Rate Differential . . . . . 72 3.9 Inflation Differential . . . . . . 73 3.10 Summary of Findings . . . . . . 75 3.11 Discussion of Findings . . . . . . 76 4 Conclusion . . . . . . . 79 4.1 Introduction . . . . . . . 79 4.2 Methodological Issues and Potential Corrections . . . 79 4.2 .1 Signal Extraction Approach . . . . 80 4.2 .2 Indicators . . . . . . . 81 4.2 .3 Crisis Window . . . . . . 83 4.2 .4 Comparison Countries . . . . . 83 4.2 .5 Currency Crise s . . . . . . 83 4.2 .6 Data Range . . . . . . . 85 4.2 .7 Thresholds . . . . . . . 86 4.2 .8 Application of the Signal Extraction Approach . . 88 4.3 Conjectures about Barbados' Stability . . . . 88 4.3 .1 Barbados' Actions . . . . . . 89 4.3 .2 Results of Actions . . . . . . 93 4.3 .3 Country Specific Factors . . . . . 94 4.3 .4 Differences in Actions Between Barbados and Comparison Countries . . . . . . . 97 4.4 Opportunities for Future Research . . . . . 98 4.4 .1 Improvements and Extensions of the Methodology . 98 4.4 .2 Further Investigation of Conjectures . . . 101 4.5 Conclusion . . . . . . . 101 Appendix . . . . . . . . 104 Bibliography . . . . . . . 112
v List of Illustrations 1.1 Sticky Price Monetary Model . . . . . . 15 1.2 Shadow Exchange Rate . . . . . . . 31
vi List of Tables 2.1 Indicator Signal Options . . . . . . . 48 3.1 Detailed Country Experiences . . . . . . 75 3.2 Summarized Country Experiences . . . . . . 76
vii FIXED AGAINST FLOATING: AN INVESTIGATION OF THE STABILITY OF BARBADOS' FIXED EXCHANGE RATE REGIME Tristan A. Zucker New College of Florida, 2012 ABSTRACT Barbados has maintained a fixed exchange rate regime since July of 1975, whereas many other countries in the Caribbean and around the world have witnessed sp ectacular regime collapses. This thesis investigates the island's exchange rate stability by analyzing movements in macroeconomic variables to compar e the economic turmoil of Barbados leading up to its most severe crisis with the experiences of three countries: Argentina, Jamaica, and Trinidad and Tobago. Barbados is found to have exp erienced more turmoil than Argenti na and Trinidad and Tobago. This th esis offers two explanations for this finding. First, limitations of the methodology might have led to a miscalculation of the comparative economic turmoil of Barbados. Second, unique characteristics of Barbados including island wide support for the peg g ed rate and strong social cohesiveness, might have enabled the island to maintain its fixed exchan ge rate by allowing it to quickly implement draconian economic adjustment measures Dr. Tarron Khemraj Division of Social Sciences
1 Introduction The island of Barbados lies far out in the Atlantic Ocean, the easternmost isla n d of the Lesser Antilles. T he British claimed the uninha bited coral outcropping in 1625 T o this day it has never experienced the turmoil of being co nquered by a foreign power unli ke many other islands in the Caribbean Likewise, it has never experienced the turmoil of a forced foreign exchange rate regime change Unique characteristics of the island, including barrier reefs and the direction of the trade winds, precluded it from invasion. Unique characteristics of the island, including a socially cohesive population in strong support of the Barbados dollar's peg to the U.S. dollar, might have also precluded it from a foreign exchange rate regime ch ange. The intent of this thesis is to examine Barbados' exchange rate stability since July of 1975, when it adopted a hard peg to the U.S. dollar. It is hypothesized that Barbados experienced little economic turmoil leading up to its 1991 currency crisis and thus did not experience sufficient pressure on its fixed exchange rate regime to result in a loss of the peg. If the hypothesis is not rejected, it is possible to conclude that the island has maintained stability despite the spectacu lar exchange rate regime failures of its neighbors in the Caribbean and around the world, because economic conditions did not deteriorate to the extent that would necessitate a regime change. If the hypothesis is indeed rejected and it is found that Barba dos has experienced economic turmoil of a sufficient amount to warrant a regime change, alternative explanations of the island's exchange rate stability are mandated. This thesis does reject the hypothesis and finds that Barbados experienced a level of press ure on its fixed exchange rate that was greater than the amount necessary to cause
2 the collap s e of Argentina s and Trinidad and Tobago 's fixed exchange rates. While the validity of this finding is dependent upon a methodology that is weak in some res pects, the finding still prompted research into the causes of Barbados stability other than the expected strong economic performance of the island The subsequent investigation highlight s th e significant policy reaction undertaken by Barbados government during its 1991 exchange rate crisi s, including island wide wage cuts, layoffs of a significant portion of public sector workers and a global credit freeze. These reactionary measures were taken with great speed and probably made it possible for Barbado s to engender enough confidence in its fixed exchange regime to prevent the peg's collapse Without the strong social cohesiveness of the island, the policy reaction probably would have faced many more political hurd les and never have occurred in time, if it even would have. The subject of this thesis is vital to both Barbados and other natio ns with fixed exchange rates. Understanding the factors that preclude exchange rate crises, whether they are taking specific measures to avoid economic turmoil or addressing turmoil through ce r tain actions, is mandatory for maintainin g a strong economic system. Ideally, t his thesis can provide a starting point for analyzing Barbados' experience and provide some preliminary lessons from Barbados' succ ess for the island itself and countries around the world. Ultimately, the severity of a country's economic turmoil might play a less important role in an exchange rate crisis than previously thought. Instead, the strength of a country 's government and its policy response could be of paramount importance. This thesis begins by providing an overview of the exchange rate and currency crisis literature, as well as an overview of the island' s most severe crisis in 1991 Chapter
3 2 outlines a currency crisis prediction approach from the literature and adaptations of the approach that ar e undertaken to creat e a framework for evaluating this thesis' hypothesis. In Chapter 3 the findings of this thesis are reported. Finally, Chapter 4 completes this thesis by offering two explanations for the rejection of the hypothesis, avenues for further research, and some concluding remarks.
4 Chapter 1 Literature Review and Barbadian Context 1. 1 Role and Importance of the Foreign Exchange Rate The most important price in an open economy that allows for the flow of goods and services in and out of the country is the relative price of the domestic currency to foreign currencies, the foreign exchange rate. Changes in the exchange rate can impact all a reas of the economy, including the expected impacts on imports and exports, as well as "prices, wages, interest rates, production levels, and employment opportunities" (Isard, 1995, p. 1). In fact, if an economy starts from a position of equilibrium, a ch ange in the exchange rate "produces instantaneous disequilibria in all markets for exportable and importable commodities and services, as well as the markets for internationally traded ass e ts" ( Riechel, 1978, p. 1). The disequilibria in export and import and internationally traded asset markets elicit a response of changes in prices in those markets that eventually impact the general price level; changes in real income and real wealth are ultimately affected by the general price level change and the entire economy is influe n ced ( Riechel, 1978, p. 1). The more open the economy and the greater the importance of imports, exports, and international assets, the more important the foreign exchange rate is; the impact of changes in the levels and composition of t hese variables induced by changes in the foreign exchange rate is proportional to the degree they comprise economic activity. As the openness of the economy increases, the importance of the exchange rate increases as well.
5 1. 2 Foreign Exchange Rate Determination Models Economists have historically focused on three variables to explain the behavior of exchange rates: national price levels, interest rates, and the balance of payments (Isard, 1995, p. 57). Modern theories of ex change rate determination have built on these historical roots and now consider a wider variety of factors, in addition to more nuanced analyses. Despite these advances, empirical testing of exchange rate determination models has not shown them to be over ly successful, but also has not completely dismissed them (Taylor, 1 995, p. 32). This section survey s various models, starting with an examination of early works, which utilized national price levels and interest rates. This examination relies on Isard's (1995) exposition of the precursors more modern theories Afterward, the flexible price monetary model, the sticky price monetary model, the general equilibrium monetary model, and the portfolio balance model will be examined. This entire overview is in debted to Taylor's (1995) excellent survey article. Hallwood and MacDonald (2000) heavily influenced the sticky price monetary model and general equilibrium monetary model sections. 1.2.1 National Price Levels / Purchasing Power Parity The underlying rationale for the determination of exchange rates through national price levels is that the price level of a country determines the purchasing power of a citizen of that country. The higher the price level, the lower the purchasing power of a given amount of currency When exchanging currencies between two countries, a citizen of a country with lower purchasing power should have to exchange more domestic currency per unit of foreign currency of a country with higher purchasing power. Otherwise, the citizen would be able to gain more purchasing power through the
6 exchange of currency and many citizens would attempt to make this exchange of currency. With the higher demand for foreign currency, the amount of domestic currency necessary to purchase one unit of foreign currency would rise until purchasing power was equalized across the two countries and there was no excess demand. Absolute Purchasing Power Parity Absolute Purchasing Power Parity (PPP) theory draws on the intuition above and asserts that the exchange rate between two currencies should equalize purchasing power and thus consist of the ratio between the price levels of the two currencies: (1.1) S = units of domestic currency per one unit of foreign currency P = domestic price level P* = foreign price level An increase in S is a depreciation of the domestic currency, as it takes more units of the domestic currency to purchase one unit of foreign currency tha n before. Alternatively, a d ecrease in S is an appreciation. This follows from absolute PPP theory, as an increase in S means the ratio of P to P* increases and domestic purchasing power declines. Thus, one unit of foreign currency wil l necessitate more units of domestic currency to maintain equivalent purchasing power across a currency exchange. Relative Purchasing Power Parity Relative PPP theory is a weaker form of A bsolute PPP theory and posits that the exchange rate is some constant proportion of price levels. This formulation recognizes that other factors may affect the numerical value of the exchange rate (e.g. transaction
7 costs) although these factors are assumed not to change, as shown by the time independence of the constant k (1.2) Taking the logarithm and then the difference of (1.2) shows that any change in one of the price levels creat e s an equiproportionate change in the exchange rate: (1.3) PPP theory merely sets out a relationship between the exchange rate and the price levels of the two relevant countries; it does not provide an indication of the direction o f causation of changes. T here may even be two way causation with the exchange rate infl uencing price levels and price levels influencing the exc hange rate Thus PPP theory cannot really be elevated to the status of a complete model of exchange rate or price levels, as both are potentially endogenous (Isard, 1995, p. 59). To conclude, emp irical studies have generally dismissed th e PPP relationship as a v iable exchange rate model in either the short or me dium run. This is because there are many other factors that have more im mediate effects and it is untenable to assume that prices adjust instantaneously. However, i n the long run, PPP theory has not entirely been rejected due to some supportive econometric studies of exchange rates in the long run (Isard, 1995, p. 59). 1.2.2 Interest Rate Parity Models It makes economic sense that a n interest rate differential can affect the exchange rate by increasing the demand for domestic currency and decreasing the demand for foreign currency when a highe r rate of return can be found domestically The interest rate parity hypothesis formalizes the above concept with both the covered and uncovered interest rate parity conditions. PPP theory is linked to the current account because of its
8 impact on trade while interest rate parity is linked to the capital account because of its impact on the fina ncial returns of investments Covered Interest Rate Parity In formulating this model, let the forward exchange rate be F : a rate of domestic currency in terms of foreign currency agreed upon by two parties at which they will exchange currencies at a spec ified date in the future. F t will denote the forward rate at time t and S t will denote the spot rate at time t or the rate at which domestic currency can be traded for foreign currency at t ime t Next let r t and r t be the domestic and foreign interest rates respectively, between t and t+ 1. Using this framework the return that a unit of foreign cur rency could ex pect to achieve in terms of domestic currency will be considered If at time t one u nit of foreign currency was exchanged for dome stic currency at the spot rate S t it would gen e rate S t (1 + r t ) u nits of domestic currency from time t to t + 1. If at time t one unit of foreign cur r ency was held in a foreign currency denominated asset, it would provide a foreign currency return of (1 + r t *) This return could then be converted to domestic curr ency at a previously agreed upon forward ra te and ultimately deliver a domestic currency retu r n of F t (1 + r t *). If thes e two methods of investment are considered equivalent in all aspects except for the currency denomination and the interest rate received, arbitrage would tend to force the following result: (1.4) If r t and r t are close to zero, the follo wing approximation is fairly valid, where f t and s t are the natural logs of F t and S t respectively.
9 (1.5) This formulation is termed covered interest parity because for the investor who uses the foreign curren cy denominated investment, the domestic currency value of the investment is assured by the previously agreed upon forward rate. The forward rate takes into account the difference in expected returns of money invested domestically and abroad and its diffe rence from the spot rate in logs is approximately the difference between domestic and foreign interest rates Uncov ered Interest Rate Parity Alternatively, the investor who used the forward rate could choose to utilize the spot exchange rate at time t+ 1, S t+ 1 to convert a foreign currency return into domestic currency. Thu s, the eventual value of the investment would be S t+ 1 (1 + r t *). The problem with this approach is that the future spot rate is uncertain and the investor must make the decision at t ime t Thus, arbitrage would tend to deliver the following formulation, based on expectations. (1.6) E t S t+ 1 = the expected t+ 1 spot rate as decided at time t Again, an approximation can be made to a rrive at (1.7) By combining the results from the covered interest parity condition and the uncovered interest parity condition, we have (1.8)
10 Ultimately, if valid, these conditions allow us to predict the effect of changes in interest rates on the expected future spot rate. If the expected future spot rate can be taken as an indicato r of the future spot rate, then ( 1 .9) where u t +1 is the prediction error. This implies that the future spot rate could be predicted by the interest rate differential, (1 10) or the forward rate, (1.11) However, several issues negate this approach, which Isard (19 95, p. 81 83) reports on. F irst expectations about the future spot rate are historically unreli able. Second interest rate differential s "explain only a small proportion of subsequent changes in the exchange rate Third the forward rate predictor equation is not a valid regression due to the fact that sample variances of the spot and forward rate are almost equivalent. Additionally, i t is important to note that both covered and uncovered interest parity rest on the assumption that the domestic and foreign investment vehicles are identical, except for the currency and the interest rate. If this is not the case, then these conditions br eak down. Reasons for failure can include different likelihoods of default between the investments and capital controls. 1.2.3 Flex ible Price Monetary Model The Flexible Price Monetary Model ( FPMM ) d efines the exchange rate as the relative price of two monies. Working from this definition, it uses the relative supply of and demand for money to attempt to model the exchange rate. As is customary, the
11 demand for money, m i s a function of real income, y the price level, p a nd the level of the nominal interest r ate, i is the income elasticity of money demand and is the interest rate semi elasticity of money demand. V ariables, except for the interest rate, are expressed in logarithmic form and asterisks denote foreign variables. The equations of equilibrium in the domestic a nd foreign money markets respectively, are ( 1 .12) and (1.13) The FPMM assumes that continuous purchasing power parity holds and utilizes this parity condition to move from the money market to the actual determination of the exchange rate. The absolute pu rchasing power parity condition again is ( 1 .14) s t = the logarithm of the exchange rate expressed as units of domestic currency necessary to purchase one unit of foreign currency This condition shows how the domestic and foreign money supplies influence the exchange rate through their impact on their respective countries' price levels. Solving the money market equations for p t and p t and then substituting them into the purchasing p ower parity condition equation (1.14) provides the fundamental flexible pr ice monetary equation: (1.15) Analyzing equation (1. 15) and recalling that an increase in the val u e of s t is a depreciation of the domestic currency leads to the following conclusions:
12 (i) An increase in the domestic money su p ply, m t will cause a depreciation of the domestic currency ceteris paribus This follows from the fact that increasing the domestic money supply will create more domestic currency per unit of foreign currency. Alternatively, an increa se in the forei gn money su p ply, m t will cause an appreciation. (ii) An increase in the domestic real i n c ome, y t will lead to an appreciation while an increase in foreign real income, y t will lead to a depreciation. The increase in real income increases the demand for mo ney, and thus increases its value relative to the foreign currency if other factors are held equal. (iii) An increase in the domestic interest rate, i t will cause a depreciation, while the opposite will occur for an increase in the foreign interest rate, i t This conclusion stems from the FPMM assumption that the decrease in domestic demand for money from an increase in the domestic interest rate overwhelms the increase in the demand for money by foreigners. However, Taylor (1995, p. 22) cautions that the domestic interest rate is endogenous to the model, and thus changes in it cannot be analyzed without allowing for changes in the other variables Taylor (1995, p. 22) notes that "open economy macroeconomics is essentially about six aggregate markets: goo ds, labor, money, foreign exchange, domestic bonds (i.e., non money assets) and foreign bonds. However, the FPMM only directly focuses on equilibrium in one of these markets, the money market. Nonetheless, the FPMM achieves a status of a market clearing general equilibrium model by its assumption of flexible prices, a freely adjustable exchange rate, the perfect substitutability of domestic
13 and foreign non money assets and Walras' law that an n market system is necessarily in equilibrium if n 1 market s are in equilibrium. Flexible prices establish equilibrium in the goods and labor markets, while a freely adjustable exchange rate establishes equilibrium in the foreign exchange market. The perfect substitutability of domestic and foreign non money ass ets essentially means that the two markets can be co llapsed into one. When the money market as described above is in equilibrium, the necessary n 1 markets are in equilibrium, which necessitates that the combined non money asset market is the n th market that must be in equilibrium as well. Despite the simplicity and elegance of the FPMM, its central assumption of continuous purchasing power parity has been dismissed. This led to the genesis of two other classes of models: the stick y price monetary model s and the equilibrium models. Additionally, the FPMM does not take into account the dual influences of an increase in an interest rate: while a higher interest rate decreases the demand for domestic money by domestic agents, it also increases the demand f or domestic money by foreign agents looking to find a higher rate of return. 1.2.4 St icky Price Monetary Model The fundamental difference between the FPMM and the Sticky Price Monetary Model ( SPMM ) is the obvious rejection of flexible prices, which in turn leads to the rejection of the continuous purchasing p ower parity condition. However, the model still assumes that purchasing power parity holds i n the long run and thus utilizes ( 1 .16) to arrive at the final exchange rate after prices have fully adjusted to their new long term values. The basic intuition behind the model will be explained using a simple example
14 and a diagram from Taylor (1995). It is important to note that the model assumes the domestic real income level and the foreign price level, real income level, and interest rate are fixed. Consider a situation in which the central bank reduces the domestic money supply. As prices are fi xed in the short run, this will cause an increase in the domestic interest rate to equilibrate money supply and money demand. This increase over the fixed foreign return will engender an appreciation of the domestic currency by way of increased capital in flows. Investors are aware that this is a temporary, artificial appreciation, and will continue to move assets into the domestic economy only so long as the expected rate of depreciation of the exchange rate is less than the excess domestic return they ar e receiving. The uncovered interest parity condition provides the short run equilibrium condition, which is when the interest rate differential is equal to the expected depreciation of the exchange rate: ( 1.17 ) The expected depreciation is evidence of the short run overshooting of the new long run exchange rate w hich Dornbusch (1976) describes. Eventually the exchange rate will depreciate to the long run purchasing power parity level as prices begin to decline, the real money supply increases, and the domestic interest rate decreases. This process can be observed in i llustration 1 .1 with the logarithm of the nominal exchange rate on the horizontal axis and the logarithm of the domestic price level on the vertical axis.
15 Illustration 1.1 : Sticky Price Monetary Model The 45 degree line represents the long run purchasing powe r parity condition when the foreign price level is held constant: as the domestic price level increases, t here is a depreciation of the nominal exchange rate of equal measure. The downward sloping l i ne s, and represents the "money market equilibrium in an open economy in which uncovered interest rate parity holds at all times" (Hallwood and MacDonald, 2000, p. 189). The negative slope of the money market line results from the assumption that money demand is po sitively related to the price level and negatively related to the interest rate. As the price level increases, the interest rate must decrease in order to maintain the money market equilibrium. The decrease in interest rate according to the uncovered int erest parity condition implies an expected appreciation in the exchange rate, which involves moving to the left along the horizontal axis. A decrease in the money supply shifts the money market line inward because at each price level, a higher interest ra te is required to clear the money market. Finally, any points not on the 45 degree line are unstable and the price level will tend to adjust, along with the exchange rate, to move toward purchasing power parity. If the economy is at a point below (above) the line, the
16 exchange rate is depreciated (appreciated) relative to the purchasing power parity level, which cause excess domestic demand (supply) and an increase (decrease) in the price le v el to bring the economy back to equilibrium. Starting with equilibrium at point A, the mo ney supply is decreased, which results in an inward shift of the money market line. Since prices ar e sticky, the new short run equilibrium is point C, which is characterized by the same price l evel and the short run equilibrium exchange rate of However, as prices decrease in the long run the economy moves down the money market line to point B, causing a depreciation of the exchange rate. Once at point B, the economy is again in long run equilibrium with the purchasing power parity condition holding and exchange rate It should be noted that there was a temporary overshooting of the equilibrium exchange rate of as described by Dornbusch. Ultimately, the effect of the decrease in the domestic money supp ly is an appreciation of the exchange rate, but not to the extent of the initial appreciation. 1.2.5 General E quilibrium Monetary Model The General Equilibrium Monetary Model ( GEMM ) is essentially an extension of the FPMM to include the impact of change s in preferences for domestic versus foreign goods and real supply shocks. To construct the model, two countries and two types of goods, domestic and foreign are used Prices are still flexible and identical representative agents "distinguish between do mestic and foreign goods in terms of well defined preferences," which are also homothetic (Taylor, 1995, p. 24). These preferences are manifest in the model by the relative price of foreign goods, the units of domestic production required to purchase one unit of foreign production. As preferences shift
17 toward the consumption of foreign production, the relative price of foreign goods will increase, as demand for them has increased. In fact, the relative price of foreig n goods is the real exchange r ate, t as described by (1.18) T he monetary equlilbria conditions outlined in (1. 12) and (1. 13) of the FPMM an be used to arrive at the equation for determining the GEMM nomin al exchange rate (for simplicity's sake, = !* and = = 0): (1.19) The similarity between (1. 19) and the FPMM is apparent and the models both exhibit similar behavior for changes in the money supply and real i ncome. Where they differ is in the real exchange rate t erm, # t which is what allows the GEMM to incorporate changes in preferences and real supply shocks. Two examples will be provided to examine how these changes affect the nominal exchange rate. Firs t, assume that there is a shift in preferences toward domestic goods, so that the demand for domestic goods increases relative to foreign goods. This will tend to cause a real appreciation, or a fall in the real exchange rate, which in turn c a uses s t to f all. It is conceivable that the increase in demand would have caused a change in relative prices, so that the price of domestic output increased. However, this is not the case as the change in the nominal e xchange rate completely absorbs the demand press ure. Second, assume that there is an increase in domestic productivity, which results in two separate impacts: a relative price effect and a money demand effect. The relative price effect is manifest in the reduction in the relative price of domestic g oods, as domestic goods are now less expensive to produce compared to foreign goods then they
18 were before. This causes both the real exchange rate, and subsequently the nominal excha nge rate, to increase ( depreciate ) However, the money demand effect operates in the opposite direction. An increase in productivity causes the transactions demand for domestic money to rise, which implies a decrea s e in s t (appreciate). The overall effect on the nominal exchan ge rate i s ambiguous, although the lower the degree of the substitutability between foreign and domestic production, the larger the impact of the relative price effect. Thus, with a small degree of substitutability, an increase in domestic productivity will probab ly lead to a depreciation of the nominal exchange rate. 1.2.6 Portfolio Balance Model In all of the models previously examined, domestic and foreign assets were assumed to be perfectly substitutable. In fact, those markets were collapsed into one ma rket in order to achieve equilibrium. The Portfolio Balance Model ( PBM ) recognizes the imperfect substitutability of these assets and can be used to analyze the impact of changes in asset preferences and allocation on the current account. To build a simp le model, let W equal the wealth of the private sector, which is composed of three different components: domestic money, M domestic bonds, B, and foreign bonds, B* In a floating exchange rate regime, the balance of payments must balance a current accoun t deficit a capital account surplus of equal and vice versa. In this model, B* is equivalent to the stock resulting from the capital account and thus the rate of accumulati o n or dis accumulation of B* is given by the current account's value. The equati on for wealth and the standard asset demand equations for its components are a s follows, with the subscripts 1 and 2 denoting the partial derivatives of the relevant argum e nts:
19 components are as follows, with the subscripts 1 and 2 denoting the partial der ivatives of the relevant arguments: As is customary, S is the exchange rate in units of domestic currency per unit of foreign currency, with S e equal to the expected rate of depreciation of the domestic currency. i and i* are the domestic and foreign rates of interest, respectively. P is the domestic price level and since the foreign price level is held constant and assumed to be one, S/P is the real exchange rate. Since the current account is composed of the trade balan ce, T and net debt service receipts, i*B* the change in the capital account, # is as given. In order t o simplify the analysis, it is assume d that expectations are static and S e = 0. To understand the workings of the model, consider a simple example in which the monetary authorities print money in order to perform an open market purchase of domestic bonds The result of the increase in the money supply is a fall in the domestic interest rate, i which causes representative agents to reduce their domestic bond holdings, B Agents will rebalance their portfolios with regard to interest bearing assets by purchasing foreign bonds, necessitating the purchase of foreign currency. This will cause a depreciation of the exchange rate, as the demand for domestic currency has decreas ed with respect to supply. If it is assume d that the economy was initially in an Zucker 21 changes in asset preferences and allocation on the current account. To build a simple model, let W equal the wealth of the private sector, which is composed of three different components: domestic money, M domestic bonds, B, and foreign bonds, B* In a floating exchange rate regime, the balance of payments must balance a current account deficit a capital account surplus of equal and vice versa. In this model, B* is equivalent to the stock resulting from the capital account and thus the rate of accumulation or dis accumulation of B* is given by the current account's value. The equation for wealth and the standard asset demand equations for its components are as follows, with the subscripts 1 and 2 denoting the partial derivatives of the relevant arguments: W M + B + SB ( 1.20 ) M = M i i + S e ( ) W M 1 < 0, M 2 < 0 ( 1.21 ) B = B i i + S e ( ) W B 1 > 0, B 2 < 0 ( 1.22 ) SB = B i i + S e ( ) W B 1 < 0, B 2 > 0 ( 1.23 ) B = T S / P ( ) + i B T 1 > 0 ( 1.24 ) As is customary, S is the exchange rate in units of domestic currency per unit of foreign currency, with S e equal to the expected rate of depreciation of the domestic currency. i and i* are the domestic and foreign rates of interest, respectively. P is the domestic price level and since the foreign price level is held constant and assumed to be one, S/P is the real exchange rate. Since the current account is composed of the trade balance, T and net debt service receipts, i*B* the change in the capital account, is as
20 equilibrium state, with a current account level of zero (the trade balance is zero and no net foreign assets are held), then the depr eciation of the exchange rate will result in a positive trade balance as the domestic economy has become more competitive domestic goods cost less in terms of foreign currency The current account is now positive, indicating that # is positive and dom estic agents are aquiring foreign assets. In their efforts to rebalance their portfolios to their desired levels of the three components, domestic agents will sell off their newly aquired foreign assets and cause an exchange rate appreciation. This in tu rn will decrease competitiveness and reduce the trade balance. Additionally, prices will start to increase as they adjust to the original increase in the money supply, which will lead to a further decrease in competitiveness and reduction in the trade bal ance. The increase in prices and appreciation of the exchange rate from its large initial depreciation will cause the trade balance to once again equal zero, as in the original equilibrium state. However, even though the trade balance is equal to zero, t he economy has not reached i t s long run equilibrium state because domestic agents still are receiving interest payments on their foreign assets of the value i*B* To reach a long run equilibrium state, the exchange rate must appreciate and the p rice level must rise until the negative trade balance exactly cancels out the interest payments of domestic agents foreign assets. The ultimate effect of the initial purchase of domestic bonds is a decrease in the domestic interest rate and a depreciation of the ex chage rate. However, as with the SPMM, there was an intitial overshooting of the exchange rate which was corrected as agents continued to rebalance their portfolios.
21 1.2.7 J C urve Effect While the J c urve effect is not an exchange rate determination theory, it does illustrate the short ru n and long ru n effects of a depreciation in t he real exchange rate on the trade balance If the J curve effect holds for an economy that experiences a real depreciation, the tra de balance will worsen in the short ru n and improve in the long run ( Ros e and Yellen, 1989, p. 56) The effect receives its name from the shape the trade balance traces out when graphed with time on the horizontal axis: the trade balance immediately decli nes, but then asymptotically increases over time, produ cing a "j" that the horizontal axis splits in half. In the short run, a country that experi e nces a depreciation will see import and export volumes remain the same as agents cannot immediately change their purchasing habits. Additionally, the price in domestic currency of exports will remain unchanged as prices are stic ky in the short run. Thus, export revenue will remain unchanged. However, the depreciation will mean that the domestic country will have to pay more per unit of foreign goods in domestic currency than it used to. At the same volume of imports in the short run, i mport expenditure will increase. With an equivalent export revenue and greater import expenditure, the short r un adverse effect on the trade balance is realized I n the long run, the lower price of exports to the foreign country will increase export volume, increasing export revenue. Domestic market agents will reduce import expenditure as the increased price c auses market agents to decrease the volume of imports demanded. Thus, in the long run, export revenue increases and import
22 expenditure decreases compared to pre depreciation amounts, improving the trade balance. 1.3 Exchange Rate Crisis Models Currency crises are characterized by two main facets: a large, persistent drop in the currency's value and resulting negative economic effects that can include "recession and unemployment; banking and business failures; inflation; reduction in investment, and losses in wealth," as well as a "heightened level of uncertainty in business and other transactions" (Breuer, 2004, p. 295). The enormity of these consequences has spawned a large literature on explaining the causes of currency crises, especially for fixed exchange rate regimes. Three generations of currency crisis models provide the main analytical framework for investigating and understanding the multitude of crises that have occurred. As is normal, each of these models arose to more effectively ex plain the crises that preceded it. When the first generation model could not properly be used to analyze crises that followed its genesis, the second generation model was developed; the same process resulted in the creation of the third generation model. These models are not monolithic, but have many variations, which cannot all be covered in a single thesis much less a portion of a thesis This section give s a broad overview of the three generations of models without tying too deeply into one formulati on. First generation currency crisis models were developed in response to various crisis situations such as Mexico (1973 82), Argentina (1978 81), and the general Latin America sovereign debt crises of the 1980s. The main cause of these currency crises were "overly expansive domestic policies" that led to the loss of a fixed exchange rate peg (Flood and Marion, 199 9 p. 1). Specifically, first generation models "show how a
23 fixed exchange rate policy combined with excessively expansionary pre crisis fund amentals push the economy into crisis, with the private sector trying to profit from the dismantling the inconsistent policies" (Flood and Marion, 199 9 p. 1). Second generation currency crisis models attempt to explain the speculative attacks in Europe (1992 93) and Mexico (1994 95) that could not be explained by the first generation models. First generation models focused on the poor, current condition of fundamentals, while these crises seemed unrelated to current economic fundamentals as portrayed in those models. The contribution of second generation models is highlighting the simple notion that government policy is not unchangeable (assumed in the basic first generation model that will be analyzed below) and the importance of private expectations' impact on the choices facing government with regard to the exchan ge rate regime. Third generation currency crisis models arose to e xplain the Asian financial crisi s (1997 98). In this crisis, macroeconomic fundamentals that the first and second genera tion models utilized were strong: "annual growth rates were high, inflation rates were low, budget deficits were low, current account deficits were management, capital inflows were strong, and political stability reigned" (Breuer, 2004, p. 301). However, the banking sector was characterized by poor domestic loans and unhedged, short term borrowing from foreign banks" (Breur, 2004, p. 301). Third generation models take into account the importance of the banking sector with regard to currency crises and ma ke explicit the interaction between the banking sector and international currency markets.
24 1.3.1 First Generation Model s This section draws heavily on Hallwood and MacDonald (2000, Chapter 14 ) and Agenor, Bhandari, and Flood (1992). The first gene ration currency crisis model was mainly develop e d by Krugman (1979), Flood and Garber (1984), and Blanco and Garber (1986). The main logic of the model is as follows. Certain factors, such as extensive domestic credit creation resulting from expansionary monetary policy, the monetization of the fiscal deficit, current account deficits, etc. will cause the persistent loss of international reserves. Eventually, reserves will fall to zero (or some other level that the central bank is not willing to see them fall below) and a country will have to abandon its fixed exchange rate regime. However, at the time of abandonment, the former fixed rate will be overvalued and thus the exchange rate will undergo an immediate depreciation and cause a capital loss for ho lders of domestic currency. Speculators will recognize this eventuality by perceiving the loss of international reserves and policies / situations that are inconsistent with the maintenance of the fixed exchange rate, as well as current poor macroeconomic fundamentals. To protect themselves from otherwise inevitable capital loss, they will mount a speculative attack in which they buy up all of the central bank's foreign reserves and thus cause the fixed exchange rate regime to collapse. This will occur b efore the date of natural depletion of foreign reserves, at the point when speculators can "confidently expect non negative return on speculat i on" ( Pes enti and Tille, 2000, p. 4). This process is illustrated below u s ing excessive domestic credit creation as the policy that is inconsistent with the maintenance of a fixed exchange rate.
25 The theoretical framework of the first gen eration currency crisis model present ed here utilizes both the monetary approach to the balance of pa yments and the FPMM of the exchange rate. The monetary approach to the balance of payments provides the insight that a fall in foreign exchange reserves is caused by domestic credit expansion that outstrips the growth of the demand for domestic money supply. This fall in fore ign exchange reserves eventually leads to the destruction of the exchange rate peg. The FPMM serves two important purposes. First, it provides a calculation of the shadow floating exchange rate, which is used by speculative agents to determine whether th e exchange rate peg is over or undervalued. Second, it substantiates the conclusion that a speculative attack on a fixed exchange rate will take place before foreign exchange reserves are depleted. The FPMM provides this substantiation because it includ es not only current values of fundamentals to arrive at the exchange rate, but also expected excess money supplies. The importance of excess money supplies will become apparent upon further exa mination of the first gener ation model The first generati on currency crisis model starts with the following assumpt i ons: (i) perfect foresight on the part of speculative agents; (ii) a small country; (iii) fixed outp u t at y ; and (iv) purchasing power parity and uncovered interest rate parity both hold. All variables, except for interest rates, are in natural log form. T he standard mone y demand function is used and is as follows. m t p t = y i t > 0 ( 1 .25 ) m t = nominal money supply
26 p t = domestic price level i t = domestic interest rate Money supply is given b y m t = D t + ( 1 ) R t ( 1.26 ) D t = domestic credit created by the central bank. R t = domestic currency value of foreign exchange reserves = initial share of domestic credit in money supply and 0 < < 1 The model posits the money supply as the domestic credit of the central bank and domestic currency value of foreign exchange reserves. This is because in the first generation model there are no private b a nks ( Agenor, Bhandari, and Flood, 1992, p. 359). The domestic rate of credit expansion is assumed to be constant and giv e n by > 0 ( 1.27 ) By letting the foreign price level be equal to one (0 in logs), PPP is give n by (1.28) The uncovered parity condition is as follows i t = i + E t s t (1.29) i* = the foreign rate of interest, which is assumed to be con s tant = the change in the foreign exchange rate Under the initial assumption of perfect foresight, the expected chang e in the exchange rate should be equal to the actual change in the foreign exchange rate and t hus E t s t = s t By letting domestic output equal 1, = 0, and substituting both (1.28) and (1.29) into ( 1 .25 ) a new formulation of money demand is arriv e d at
27 i t = i + E t s t (1.30) Similarly, (1.30) can be substituted into (1. 2 6 ) to arrive at an equation of foreign exch ange reserves. Here, a fixed exchange rate is ass u med (1.31) The rate at which foreign exchange reserves are reduced is found by subs tituting (1.27) into ( 1 .31 ) R t = / (1.32) # = (1 ) / is negative because agents will constantly be shifting from domestic to foreign currency, as they their demand for domestic currency is not rising at the same rate that domestic credit is expanding. Additionally, the conclusion that arises from this depl etion of foreign exchange reserves is that the domestic monetary authorities will eventually have to abandon the current fixed exchange rate. In this model, abandonment takes the form of switching to a cleanly floating regime. To determine exactly when t he fixed exc h ange rate regime will collapse, FPMM is used to arrive at an equation for the shadow exchange rate s t = ( D 0 + ) + t (1.33) = the shadow exchange rate, which is the foreign exchange rate that would prevail if the rate was allowed to float and reserves were z ero ( Agenor, Bhandari, and Flood, 1992, p. 361). D 0 = domestic credit at t = 0
28 T he depreciation rate of the shadow exchange rate is given by $ t Again, if there was no domestic credit expansion, and was equal to 0, the shadow exchange rate would not depreciate. Finally, an equation for the time at which the fixed e xchange rate regime collapses, when the shadow exchange rate just rises above the fixed rate can be formu l ated (1.34) t c = the time of collapse R 0 = initial reserves Combined with the rest of the model, this equation provides a few insights about the timeframe o f collapse. The time until collapse will decrease (i) the sm a ller is, or the greater the portion of the initial money stock that is composed of domestic credit (ii) the smaller R 0 is, or the smaller the initial foreign exchange reserves of the central bank (iii) the larger is, or the greater the rate of domestic credit expansion (iv) the larger is, or the greater the semi interest elasticity of the demand for money. This last insight derives from the fact that currency depreciation creates expe ctations of inflation, which drives up the nominal interest rate and reduces money demand. The intuition behind the model is that speculators realize excessive domestic credit expansion and a fixed exchange rate cannot coexist indefinitely because the for eign exchange reserves of the central bank will be depleted and it will no longer be able to
29 maintain the fixed rate. Speculators desire to avoid the losses associated with the change from a fixed to a flexible regime with a depreciated and depreciating r ate and thus will convert domestic currency to foreign currency (in an amount equal to the entire foreign reserves of the bank) in a speculative attack before the natural time of depletion. The time that speculators will choose to make this massive conver sion is giv e n by t c the time when the shadow exchange rate (the rate that would exist post speculative attack) depreciates just past the fixed exchange rate. This will cause the currency crisis to occur before it would otherwise had with a natural rundow n of reserves from a fixed domestic money demand and excessive money supply because the shadow exchange rate equals the fixed exchange rate when there is still a positive level of reserves. Had speculators drawn reserves down before they did, when the s hadow rate was below the fixed rate, they would have experience an instantaneous capital loss because their foreign currency in terms of domestic currency would be worth less than the original amount of domestic currency. If they were to draw down reserve s after the shadow exchange rate depreciated above the fixed exchange rate, they would experience instantaneous capital gain for the opposite reason. With perfect foresight, the total drawdown of reserves comes right at the point where the shadow exchange rate just surpasses the fixed rate each of the other situations is incompatible with a perfect foresight equilib r ium ( Agenor, Bhandari, and Flood, 1992, p. 361). To see why, consider if the shadow exchange rate was a few hundredths of a currency unit a bove the fixed rate. Then, the first speculator to draw down central bank reserves would make all of the instantaneous profit. Thus, all of the speculators push the attack time back to just when they will not realize any losses. 1.3.2 Second Generation Models
30 In contrast to first generation models in which government and private action follow predetermined paths, second generation models introduce the possibility of government policy reactions to private actions as well as government r esponses to an "explicit trade off between fixed exchange rate policy and other objectives" (Flood and Marion, 1999 p. 7). T wo versions of models that fall under the second generation umbrella are presented below. The first examine s how mul t iple equilib ria (a hallmark of second generation models) can develop when government policy depends on whether or not a speculative attack is attempted. In contrast to first generation models in which poor fundamentals and the policies that caused them pushed the ec onomy into a specu lative attack situation, second generation models can include "attack conditional policy changes" which can pull the economy into a full speculative attack situation and the resulting loss of a fixed exchange rate regim e. The second mode l examine s how a shift in market expectations can actually fulfill those expectations the self fulfilling prophecy aspect of second generation models. Three characteristics of second generation models become apparent when reviewing the examples below. First, a switch to a floating regime is not necessarily according to fundamentals as it is in the first generation crisis models. Second, multiple equlilbria are possible. Third n o explanation is provided within the model as to how actions are coordinat ed among speculators or how expectations change. The first version of the second generation model remains in the familiar territory of the domestic credit exp ansion scenario that was examined above in the first generation model and is expertly explained by Flood and Marion (1999). If there is no speculative attack, domestic credit grows at some rate 0 Here it is assume d that 0 = 0 so that the
31 fixed exchange rate could be maintained indefinitely for certain initial levels of domestic credit. If ther e is a speculative attack on the fixed exchange rate regime, domestic credit grows at an increased rate of 1 the attack conditional policy ch a nge ( Obstfeld, 1986, p. 76). T he different scenarios that can arise in this second generation model example c an be examined in i llustration 1.2 of the fixed exchange rate and the two shadow exchange rates resulting from the two domestic credit growth rates, 0 (the lower line) and 1 Domestic credit ( d ) is on the horizontal axis and the nominal exchange rate ( s ) is on the vertical axis. Illustration 1.2 : Shadow Exchange Rate Case 1: d < d B The economy will initially be on the lower shadow rate l ine, If speculators were to initiate an attack, the shadow rate would jump to the higher shadow rate line, as a result of the government's attack conditional policy response. However, all points to the left of d B on the higher shado w rate line fall below the fixed exchange rate and thus speculators would realize a net capital loss if they were to attack. Since the shadow rate is the exchange rate that would prevail if the government did not
32 possess any international reserves and the currency was flexible (the result of a successful speculative attack), any domestic currency converted in the attack would immediately lose value in domestic terms at the appreciated exchange rate. Ultimately, the economy can remain in this equilibrium with a fixed exchange rate indefinitely, as s peculators will not attack and suffer losses. Case 2: d = d B Now the economy is at point C o n the shadow rate line. Speculators could successfully attack and shift the shadow rate from point C to point B. However, they would not realize any immediate gain (nor avoid any loss) from this maneuver. Therefore, equilibrium can occur at point B or point C; speculators are indifferent between the two points as neither represents a profit opp ortunity. Case 3: d B < d < d A In this case, the economy is on the lower shadow rate line between points C and A. A successful speculative attack would shift the shadow rate to the higher shadow rate and guarantee a profit. However, a speculative attack is not assured and multiple equ ilibria might be possible. This is the result of an economy that is made up of small and uncoordinated speculators or speculators that face costs in mounting an attack. If all of these agents do not believe than an attack will occur, they will not indivi dually undertake an attack because they would incur needless costs and the economy will remain on the lower line indefinitely. If speculators are convinced that a full speculative attack will occur, than all will engage in the attack, because to do otherw ise would ensure the loss of a profit opportunity. This attack will shift the shadow rate to the higher line and the fixed exchange rate will collapse as the domestic credit growth rate increases and
33 international reserves are fully depleted. The case in which mul t iple equilibria are not possible for d B < d < d A is when some large, well financed trader has the ability to initiate an attack by itself. Otherwise, a coordination failure of small speculators will continue to keep the economy in the no attack equilibrium; "the economy can maintain the fixed exchange rate indefinitely unless something coordinates expectations and actions to cause an attack" (Flood and Marion, 1998, p. 16). The second generation model by itself does not offer any insight into wh at will cause an attack in Case 3 and coordinate the actions of the many small speculators, but Morris and Shin (1 9 95), Banerjee (1992), Bikhchandani, Hirshleifer, and Welsh (19 92) and Calvo and Mendoza (1996 ) provide some potential explanations. Case 4 : d $ d A In this trivial case, the fixed exchange rate will surely be attacked, as this assured profit case replicates the first generation currency crises condition. In the second generation model above, speculators' actions in attacking the exchange rate pulled the economy into a full fledged attack and loss of the fixed exchange rate regime because the government made an attack conditional policy change that moved the economy to a new shadow exchange rate schedule. In Case 3 all that was needed was a l arge market mover or the coordination of many smaller market players to bring about the collapse of the fixed exchange rate regime Other second generation models focus on the trade off the government faces between maintaining the fixed exchange rate and ac hieving other important economic goals, such as full employment. When the costs of maintaining the peg are outweighed by the benefits of a regime switch, the government will no longer take actions to maintain it. In
34 this version of a second generation mo del, it is changes in market expectations that change the trade offs the government faces and encourage the government to abandon the p eg. Saxena (2004, p. 328 ) and Pesenti and Tille (200 0 p. 5 6) both influenced this section. The model begins with two basic tenets: 1. The government could realize benefits from abandoning the fixed exchange regime, including inflating away the domestic currency denominated debt burden or lowering unemployment by employing expansionary monetary policies. 2. The government rea lizes benefits from maintaining the fixed exchange regime, including the traditional arguments of increased international trade and investment, as a credibility anchor in the face of previous high levels of inflation, and even as a source of national pride Initially the perceived bene fits of maintaining the peg outweigh the costs of abandoning the peg and thus the government takes actions to maintain it. However, if the cost of maintenance rises, the government might no longer be so inclined and the mark et will witness the collapse of the fixed exchange rate regime. The key insight of second generation models is that changes in expectations can increase (or decrease) the impact of tenet one. The costs of maintaining the fixed exchange rate regime will ri se if people expect that it will be abandoned. Anticipation of an increased possibility of devaluation or collapse by foreign investors who make loans denomina ted in domestic currency will induce foreign investors to r aise the interest rate they char ge. This in turn will decrease domestic investment and reduce domestic aggregate demand, leading to increased unemployment. This situation would increase the opportunity cost of maintaining the
35 fixed exchange rate regime as the government could "abandon the fi xed rate to boost aggregate demand and employment" from an increase in exp o rts ( Pesenti and Tille, 2000 p. 6). If demand and employment are depressed enough, the benefits of abandoning the fixed rate will outweigh the costs and expectations will have led to the collapse of the fixed exchange rate a self fulfilling prophecy. If it is assume d as second generation models do, that market expectations are not necessarily driven by fundamental s then abandonment expectations can just be market whims. Thus, this model ha s two equilbria, with the only difference the expectations of economic agents. 1.3.3 Third Generation Models Although the initial third generation currency crisis models arrived in the late 1990s, the literature concerning them is still f airly heterogeneous. Nonetheless, there are some main parallels between the models, as expound e d by Krznar (2004). These models focus on the banking sector, particularly with regard to foreign over indebtedness, moral hazard and asymmetric information re sulting from government and international financial institution guarantees of banks, and lack of effective regulation. These factors lead to a banking system crisis, which in turn leads to a currency crisis by the analysis of the first generation model: g over nment guarantees result in the central bank acting as lender of last resort and increasing domestic credit. A "herd effect" is also present some models in which a few seeking refuge from domestic financial problems in foreign currencies result in a mass exodus to safety. Finally, third generation models often consider a contagion effect, in which a currency crisis spreads from country to country. This spread occurs through linkages involving reduced competitiveness resulting from the currency depreciation of another country; fina ncial interdependence in which foreign creditors of
36 the country experiencing the initial crisis are forced to recall loans from a country not undergoing a crisis; and capital outflows precipitated by the view that countries similar to the one experiencing a crisis are likely to experience a crisis as well. 1.4 Indicators of Exchange Rate Crises Identifying a set of ind icators that can help predict currency crises is of g reat interest to varied parties. F inancial market participants desire indicators be cause "they want to make money, policymakers because they wish to avoid the crisis, and academics because they have a long history of fascination with fi nancial cri s es" ( Kaminsky Lizondo and Reinhart, 1998, p. 1 2) [henceforth KLR] This thesis is also interested in identifying a relevant set of indicators for Barbados, which will enable the testing of the hypothesis that Barbados' indicator behavior has been be n ign in contrast to the comparison countries indicator behavior s which is why it has not los t its peg or devalued. KLR (1998) utilize a signal extraction approach to identify good indicators on the basis of their ability to accurately signal a currency c risis. This section first present s a comprehensive list of indicators tested in the twenty e ight papers that KLR (1998) reviewed. Next it describe s the method they used to determine the literature's view of the most successful indicators, as well as their general analysis of the findings of th e indicator literature. Then it outline s the indicators that KLR tested with their signal approach, the methodology behind the signal approach, and finally the indicators that they found to be good predictors of a currency crisis. The main indicators that were found in the literature review can be grouped in ten general categories and are as follows ( KLR, 1998):
37 1. Capital account: international reserves, capital flows, short term capital flows, foreign direct inve stment, and differential between domestic and foreign interest rates. 2. Debt profile: public fo r eign debt, total foreign debt, short term debt, share of debt classified by type of creditor and by interest structure, debt service, and foreign aid. 3. Current account: real exchange rate, current account balance, trade balance, exports, imports, terms of trade, price of exports, savings, and investment. 4. International variables: foreign real GDP growth, interest rates, and price level. 5. Financial liberalization: credit growth, change in the money multiplier, real interest rates, and spread between bank lending and deposit interest rates. 6. Other financial variables: central bank credit to the banking system, gap between money demand and supply, money growth, bond yields, domestic inflation, "shadow" exchange rate, parallel market exchange rate premium central exchange rate parity, position of the exchange rate within the official band, and M2/international reserves. 7. Real sector: real GDP growth, output, output gap, employment/unemployment, wages, and changes in stock prices. 8. Fiscal variables: fiscal d eficit, government consumption, and credit to the public sector. 9. Institutional/structural factors: openness, trade concentration, dummies for multiple exchange rates, exchange controls, duration of the fixed exchange rate
38 periods, financial liberalization, banking crises, past foreign exchange market crises, and past foreign exchange market events. 10. Political variables: dummies for elations, incumbent electoral victory or loss, change of government, legal executive transfer, illegal executive transfer, left wing government, and new finance minister; also, degree of political instability (qualitative variable based on judgmen t ). KLR' s (1998) main consideration in identifying good indicators was that the indicator was found to be statistically significant whe n used to predict a currency crisis in the paper that examined it Additionally, indicators were only selected from studies that systematically compared indicator behavior in pre crisis times with its behavior in tranquil times or non crisis countries (a control group). These restrictions significantly narrowed down the list of indicators that had any success. The following general conclusions were reached: 1. Significant indicators were not constrained to either the real or financial sectors, nor were they just domestic or just foreign in nature: good indicators are of a broad variety. Additionally, this conclusion is in line with the evidence that a multitude of economic and occasionally political problems precede currency crises. 2. According to the liter ature the best indicators are international reserves, the real exchange rate, credit growth, credit to the public sector and domestic inflation. Additionally, the trade balance, export performance, money growth, M2/international reserves, real GDP growth and the fiscal deficit are found to be important as well.
39 3. Strong conclusions about the other indicators reviewed cannot necessarily be made, as they were not included in multiple studies and thus do not have enough relevant evidence. However, some ins titutional, political, foreign, and financial variables seem promising in their ability to predict a currency crisis. 4. Neither the current account balance, nor external debt profile variables had good predictive power. 5. Market variables, such as interest rate differentials and exchange rate expectations were not effective predictors of currency crises. KLR (1998, p. 17) tested the following indicators in their signal extraction approach: (1) international reserves (in U.S. dollars); (2) imports (in U.S. dollars); (3) exports (in U.S. dollars); (4) the terms of trade (defined as the unit value of exports over the unit value of imports); (5) deviations of the real exchange rate from trend (in percentage terms); (6) the differential between foreign (U.S. or German) and domestic real interest rates on deposits (monthly rates, deflated using consumer prices and me asured in percentage points); (7) "excess" real M1 balances; (8) the money multiplier (of M2); (9) the ratio of domestic credit to GDP; (10) the real interest rate on deposits (monthly rates, deflated using consumer prices and measured in percentage points ); (11) the ratio of (nominal) lending o deposit interest rates; (12) the stock of commercial bank deposits (in nominal terms); (13) the ratio of broad money (converted into foreign currency) to gross international reserves; (14) an index of output; and (1 5) an index of equity prices (measured in U.S. dollars).
40 These indicators were chosen based on unexplained theoretical considerations (assumed to be related to conclusions resulting from the review of the literature) and whether or not data were availabl e on a monthly basis. The actual signal approach used by KLR (1998) looks at the ability of indicators to provide a signal of an impending crisis by rising above some threshold. The approach defines a crisis period as one in which an exchange market press ure index is higher than its mean by three standard deviations or more. The exchange market pressure index is calculated by using a weighted average of the negative of monthly percentage changes in gross international reserves and the monthly percentage c hanges in the exchange rate, with weights assigned in order to give both components the same conditional variance. The higher the index is, the greater the pressure to exchange domestic currency for foreign currency. The threshold that the indicator's v alue must rise above (or fall below depending on the indicator) to give a signal was set by choosing the value that would put 10 20% of observations of the variable above (or below) the chosen value. The actual percent of the observations above (or below) the threshold value was chosen to minimize the ratio of false to good signals. A good signal for an indicator is defined as the indicator rising above (or falling below) the threshold value within the twenty four months leading up to an actual crisis. I n this way, a signal would provide predictive power of a currency crisis. After utilizing their signal approach, KLR (1998, p. 18 23) evaluated the tested indicators using two main measures: the adjusted noise to signal ratio (which was the ratio of false signals to good signals, with false signals divided by the number of months in which they could have been correctly issued and good signals divided by the number
41 of months in which they could have been correctly issued) and the "number of months in advance of the crisis when the first signal occurs." The lower the adjusted noise to signal ratio, the more probable it is that the indicator correctly signals an impending crisis. Another important measure of the quality of an indicator is its persistence, the number of times a signal was issued in the pre crisis period as compared to more tranquil times. However, this val ue is just the reciprocal of the adjusted noise to signal ratio and thus is already taken into account in the evaluation. Ultimately, KLR (1998, p. 1) find that the best indicators utilizing the signal approach are "exports, deviations of the real exchang e rate from trend, the ratio of broad money to gross international reserves, output, and equity prices." In a different attempt to predict currency cr i ses, Burkart and Coudert (2002) analyzed fifteen emerging economies and constructed a global model of c urrency crisis indicators, as well as regional models for both Latin America and Asia. The fifteen countries considered were Argentina, Bolivia, Brazil, Chile, Colombia, Mexico, Peru, Indonesia, Malaysia, the Philippines, Thailand, Turkey, Hungary, Poland and South Africa In Burkart and Coudert's (2002) best model, a global score of the likelihood of a currency crisis is found as a function of the following indicators: (+) reserves / M2, (+) reserves / total debt, ( ) short term debt / total debt, ( ) d eviation of the real effective exchange rate from its long term value, ( ) region contagion indicator, and ( ) inflation (the signs indicate the effect on the global score when there is an increase in the indicator and overall, a lower score means a higher probability of a crisis within the next ye a r). Burkart and Coudert's (2002) model differs from KLR 's (1998) not only in the variables found to be most significant as predicators of a currency crisis, but in that Burkart and
42 Coudert's model provides a pro bability of a currency crisis resulting from a combination of all the indicators. K LR (1998, p. 24) suggest a possible extension of their work to be combining "the information on the various indicators to estimate the probability of a crisis condition on simultaneous signals from any subset of indicato r s." Burkart and Coudert (2002) have not quite accomplished this extension, as they only use threshold values in order to identify relevant indicators rather than to actually use thresholds in their model to calculate the probability of a currency crisis, but nonetheless have constructed a model with a 79.6% successful prediction ratio. 1.5 Barbadian Context The 1991 Currency Crisis The main factors leading to the pressure on the fixed exchange rate in 1991 include a reduction in foreign exchange earnings coupled with an increase in imports, domestic credit expansion resulting from expansionary fiscal policy, and an increase in external debt service. Since the 1970s, Barbados has been transitioning fro m an agriculture based economy to a tourism and financial services based economy and by 1991 the island was already significantly dependent on tourism. In fact, by the mid 1980s tourism earnings were equal to almost 30% of GDP and more than 50% of total export earnings (Howard, 1989). It is therefore not surprising that the negative impact of the Gulf War and the general international economic slowdown in the period leading up to 1991 on Barbadian tourists arrivals resulted in a substantial decrease in f oreign exchange earnings. The Central Bank of Barbados reported that gross spending of tourists grew on average 13.1% between 1986 and 1990 while it fell by around 10% in 1991 ( Research and Economic Analysis Department of the Central Bank of Barbados 199 2, p. 13) [henceforth READ]
43 The other main foreign exchange earning sectors, agriculture and manufacturing, were also in recession, leading to a further deterioration of the current account of the balance of payments (Dalrymple, 1995, p. 294). Imports, on the other hand, continued to grow leading up to the 1991 crisis period. A feedback mechanism of the anticipation of devaluation was a substantial increase in imports in the first quarter of 1991: capital goods by 40.8%, consumer items by 19.2%, and r aw materials by 13% ( READ 1992, p. 14). Overall, imports increased 23% in the first quarter of 1991, following an increase of 10% in the first quarter of 1990 (Dalrymple, 1995, p. 290). Barbados witnessed an election year in 1991 and the rapid domestic credit expansion can mainly be attributed to this, as was the case in 1981 currency troubles The roots of the large deficit in this period leading up to the crisis were increases in wages, transfers, and capital expenditure that all would support the incu mbent party's bid for reelection, especially evident in 1989 and 1990. For fiscal year 1990/91, the deficit to GDP ratio rose to 8.2%, which the government chose to monetize in the tradition of the classic first generation currency crisis model ( READ 199 2, p. 10). As the model posited, the increase in the money supply without a concurrent increase in domestic money demand caused a rundown of reserves; by July 1991, foreign exchange reserves had declined to about two weeks of import cover, which was largel y an impact of this monetization. Barbados is currently the third most developed country in the Western Hemisphere, following Canada and the United States. This is partially due to a stated policy of improving infrastructure and other areas necessary for development through
44 acquiring foreign debt, which began in the 1970s (Dalrymple, 1995, p. 291). From 1972 to 1986, the ratio of external debt to GDP increased nearly fourfold and debt service increased more than sixfold (Boamah, 1988). Additionally, the composition of the debt changed from development financing to consumption financing, which also involved a change from "concessionary project loans to short term high interest loans negotiated on the international capital markets" (Dalrymple, 1995, p. 291 ). In 1991, Barbados faced debt service of 15.6% of exports, as compared to 4.8% of exports in 1982 ( READ 1992, p. 15). The consequence s of th is high level of debt service were worsened in the early 1990s by the country's inability to roll over maturing loans due to the decline in international finance for emerg ing market countries (Worrell et al., 2003, p. 4) This decline was due to investors' risk aversion caused by several currency crises in emerging markets in the early 1990s. Barbados' borrowing was further hampered because of foreign creditors increased concern for its poor economic status in this period (Haynes, 1997, p. 95). Thus, when Yen Bond B (Yen 4.3 billion), issued in 1986 came due in 1991, precious foreign exchange reserves had to be used to pay its va lue as it could not be refinanced (Williams, 2001, p. 153).
45 Chapter 2 Methodology 2. 1 Introduction The purpose of this thesis is to investigate why Barbados has been able to maintain a fixed exchange rate while many other countries have experienced exchange rate crises that have led to devaluations or abandonment of their pegged regimes. According to t heory curren cy crises are not random events but are often preceded by adverse changes in the economic situation of the country that experiences the crisis. For example section 1.5 Barbadian Context 1991 Currency Crisis showed that leading up to the 1991 crisis, Barb ados realized negative movements of imports, government expenditure and various other important macroeconomic variables for t he island Section 1. 2 Foreign Exchange Rate Dete rmination Models and s ection 1. 3 Exchange Rate Cr isis Model s corroborate and elaborate on the view that certain important variables are the main determinants of the stability (or instability) of a country's foreign exchange rate. Thus, it can be hypothesized that differences in the behaviors of certain macroecono mic variables of Barbados from those of the comparison countries which did experience a crisis caused exchange rate regime change could be the reason Barbados has been able to maintain i ts regime. In particular, th is thesis argues that if the important ec onomic variables of the island underwent less adverse change leading up to times of negative pressure on the country's exchange rate as compared to countries that lost their fixed exchange rate regimes then Barbados has been able to maintain its fixed exc hange rate because there was not as much pressure to devalue.
46 This thesis will utilize a methodology that compares differences in behavior of eight important economic variables between Barbados and three comparison countries that did lose their fixed exchange rate regimes: Argentina, Jamaica, and Trinidad and Tobago. The remainder of this chapter outline s the analytical framework that th is thesis' methodology builds off of in section 2.2 Signal Extraction Approach, adapts the components of the approach to the situation at hand in section 2.3 Model Specifications of the Thesis Ap proach, and modifies the signal extraction approach and its application to the hypothesis in section 2.4 M odification of the Signal Extraction Approach and section 2.5 Application of the Signal Extraction Approach 2. 2 Signal Extraction Approach In their seminal paper, Leading Indicators of Currency Crises KLR (1998) outline a method for predicting curre ncy crises that takes advantage of a broad variety of economic variables that have been found to indicate crises. The objecti v e of KLR's prediction model is to provide policy makers with an early warning system that gives them enough time to make policy c orrecti ons to prevent a currency crisis from occurring. The intuition behind their signal extraction approach is that important variables will behave abnormally leading up to a country's currency crisis. This abnormal behavior can be characterized by an indicator's value passing above some threshold, thus signaling that a crisis might occur in the near future. By aggregating the signals given off by the set of indicators, policy makers can gain some idea of the likelihood of a crisis in the future. If o ne seems imminent, they can take appropriate action to preclude it. A more detailed
47 explanation of the approach follows, which draws heavily on Edison's (2003) expansi o n of KLR's model by the inclusion of more variables and more countries for calibration 2.2.1 Scope of the Model The scope of the model is determined by the choice of indicator variables and variable transformation, the countries covered, the time period covered, the crisis prediction window, and the frequency of the da t a The indicator variables are the most important part of the model, as they provide the data actually used to predict the crises. KLR (1998) used the variables outline d in section 1.4 Indicators of Exchange Rate Crises which were selected because of t heoretical and empirical considerations, as well as the availability of monthly data. Edison (2003) selected the fourteen variables that KLR (1998) used, along with seven new variables: US output, G 7 output, US interest rates, oil prices, the level of M2 /foreign exchange reserves, the change in short term debt/foreign exchange reserves, and the level of short term debt/foreign exchange reserves. The indicator value used in the model is the twelve month percentage change of the underlying variable for all variables except the deviation of the real exchange rate from trend, the excess' of real M1 balances, and the interest rate variables. The deviation of the real exchange rate from trend is the difference between the real exchange rate and the trend, whi le the excess' of real M1 balances is the difference between the supply and demand of real M1. Interest rate indicators were calculated as the twelve month level change in the underlying variable. For a country to be included in Edison's model, it must have experienced a currency crisis between January 1970 and April 1995, the time period covered. Edison
4 8 examined twenty eight developed and emerging economies, adding Korea, Portugal, South Africa, Greece, India, Pakistan, Sri Lanka, and Singapo r e to KLR' s original selection of countries from Latin America: Argentina, Bolivia, Brazil, Colombia, Chile, Mexico, Peru, Uruguay, and Venezuela; Asia: Indonesia, Malaysia, the Philippines, and Thailand; and other regions: Denmark, Finland, Israel, Norway, Spain, S weden and Turkey. Finally, Edison defined the crisis prediction window, the length of time examined in the model, to be the twenty four months leading up to a crisis. This window of examination conforms to the idea of an early warning system, in that is s hould be able to give meaningful signals well in advance of a crisis so that the authorities can begin to take proper steps. 2.2.2 Definition of a Currency Crisis A crisis is defined as occurring in any given month if the exchange market pressure index varies from its mean for the s ample period by greater than 2.5 standard deviations. The exchange market pressure index is calculated by using a weighted average of the negative of monthly percentage changes in gross international reserves and the monthly percentage change in the exchange rate, with weights assigned to give both components equal variance. The pressure to exchange domestic currency for foreign currency increases as the index increases. 2.2.3 Generation of Predictions An indicator provide s a signal of an impending crisis if it exceeds a threshold value, and no signal if the indicator stays within the threshold. A threshold is numerically defined as the value corresponding with some percentile of the all of the
49 values the indicator ta kes i n the sample. For example, if the 10 th percentile of the foreign exchange reserves indi c ator was a 15% decline in reserves, then a signal would be issued for any month during which reserves declined by 15% or more. The indicator threshold percentiles are calculated on an indicator by indicator basis to "maximize the signaling performance of each indicator" (Edison, 2003, p. 24). The calculation process can be understood if we let: A = the number of times an indicator issues a signal and a crisis occurs w ithin the crisis window (a good signal) B = the number of times an indicator issues a signal and a crisis does not follow within the crisis window (a bad signal / noise) C = an indicator that does not issue a signal but a crisis occurs within the crisis wi ndow (a missed signal) D = an indicator that does not issue a signal and a crisis does not follow within the crisis window (a good silent signal) These four options can be represented in the following table (Edison, 2003, p. 24). Table 2.1 : Indicator Signal Options Crisis within 24 months No crisis within 24 months Signal issued A B No signal issued C D To maximize signal performance, a balance must be struck between having too many bad signals and the risk of missing too many crises if an indicator does not signal. If a threshold is set at a more extreme level (a lower percentile), the risk of observing many bad signals is lowered because the indicator's value must be much more sever e to pass the threshold and signal. If a threshold i s set at a less extreme level (a higher
50 percentile), the risk of missing predicting a crisis is lowered. A balance between these two options is achieved by choosing the threshold for each indicator that minimizes the no ise to signal ratio, [B/(B+D)]/[A/(A +C)] across all of the countries in the model. Noise, B/(B+D), is the number of signaled crises when an actual crisis didn't occur over the number of months with no crisis within twenty four months. The signal component is the number of correct signals a s a percentage of times there should have been a signal (i.e. a crisis occurred). Naturally, not all indicator noise to signal minimizing threshold percentiles will be the same, although they generally fall between the 10 th and 20 th percentiles. The advantage of the threshold approach using percentiles is that the actual threshold value is tailored to the specific range of an indicator's values for each country. Therefore, Sri Lanka and Argentina can both have numerical threshold values for each indi cator, calculated using the same percentile, that take into account the range of indicator values for the respective countries. For example, if Argentina on average has foreign exchange reserves that fluctuate more widely than the reserves of Sri Lanka, t he threshold value for its international reserves indicator will be a larger decline than that of Sri Lanka's. If it was n o t and some absolute threshold number (e.g. a 10% change in reserves) was used across all countries, Argentina's international reserv es indicator might send crisis signals more often than was warranted and Sri Lanka's might send crisis signals less often than warranted. 2.2.4 Justification of the Applying the Approach to the Thesis Question While the signal extraction approach is cur rently used as an early warning system for predicting a currency crisis in a single country, it can also be modified for use as a
51 backward looking tool to examine the behavior of the indicators leading up to a crisis. The general approach provides the fra mework for analyzing whether or not a macroeconomic variable is behaving abnormally (a signal) and to what extent (multiple signals during the crisis window). Furthermore, the framework extends itself to inter country analysis in that the number of signal s that precede a crisis (either for each indicator or in aggregate) can be compared to make preliminary conclusions about the differences in the severity of experiences. These comparisons are justifiable in that a signal statistically has the same meaning for two countries because the signal is tied to a percentile of the indicator values of the respective countries. 2. 3 Model Specifications of the Thesis Approach To adapt the signal extraction approach to the current research question, certain changes must be made to it, in addition to cho o sing the indicators to analyze. Specification of the indicators, definition of a currency crisis, the choice of a crisis window, the choice of comparison countrie s, the choice of currency crises to examine, the data range, and the choice of thresholds is made below. 2.3.1 Indicators An extensive theoretical literature exists on the important variables that determine exchange rates, the main conclusions of which were surveyed in s ection 1 2 Foreign Exchang e Rate Determination Models. This literature has informed theories of exchange rate regime collapse, s uch as the three generations of exchange rate crisis models examined in section 1.3 Exchange Rate Crisis Models. Ultimately, researchers have attempte d to apply the knowledge contained in both bodies of literature to predict exchange rate crises and in doing so have used three main empirical methods: a structural
52 approach, an approach that uses discrete variable techniques and panel data, and a signal e xtraction approach (Flood, Marion and Yepez, 2010, p. 3 4). KLR (1998, p. 9 10) reviewed a large body of empirical work conforming to these different methods and found that the main variables that have been used can be classified into ten general categor ies: capital account, debt profile, current account, international, financial liberalization, other financial, real sector, fiscal, institutional/structural factors, and political. These are reported in full in section 1.4 Indicators of Exchange Rate Cr ises. Ultimately, this thesis will consider the following variables, which are sourced from the literature re v iew, KLR's ten categories, and Edison (2003) and meet the constraints of the methodology. For example, only variables that are reported in monthl y frequency and are available for Barbados and all of the compari s on countries during the examined time period can be considered. Justifications for the choice of each of the variables are found below, and stem from the empirical results of KLR (1998, p. 10 13 and p. 44 45) and Edison (2003), as well as theoretical considerations found in the literature review. The format provided below is indicator (critical region, threshold, transformation). Indicators for which Edison (2003, p. 8) did not provide crit ical regions that minimize the noise to signal ratio are indicated with an asterisk and given the most conservative (in terms of maximum noise reduction) critical region of .1. 1. Foreign exchange reserves (.1, lower, 12 month percentage change) Ultimatel y, it is the inadequacy of a Central Bank's foreign exchange reserves to meet market agents' demands that causes the collapse of a fixed exchange rate regime. When the Central Bank runs out of reserves, the supply of foreign exchange cannot meet
53 the deman d for foreign exchange at the pegged rate and thus the domestic currency depreciates or is devalued, destroying the peg. Additionally, a loss of foreign exchange reserves shows that market participants and not confident in the exchange rate regime and are trying to convert into a currency that will maintain value. Thus, the level of foreign exchange reserves is a very important indicator of the health of peg, both as an indicator of the authorities' ability to maintain the chosen regime and market percept ion of regime stability. Foreign exchange reserves for each country are simply total reserves minus gold (International Financial Statistics (IFS) line 1L.DZF). 2. Real exchange rate deviation from trend (.1, lower, base indicator) The real exchange rate for a country is the cost of a basket of goods in the country relative to another country. If the domestic real exchange rate increases, the country has experienced a real depreciation, as it takes more domestic output to purchase a n equal amount of foreign output. On the other hand, if the domestic real exchange rate decreases, the country has experienced a real appreciation, as it takes less domestic output to purchase an equal amount of foreign output. An appreciation of the rea l exchange rate causes domestic exports to be more expensive for foreign countries, causing them to decrease consumption. At the same time, domestic imports become less expensive, caus ing the country to increase purchases of imports. This simultaneous de crease in foreign exchange earnings and increase in foreign exchange outflows can lead to a crisis, and thus a real appreciation is an indicator of an exchange rate crisis. The real exchange rate is composed of the foreign exchange rate in country currency per U.S. dollar (IFS line AE.ZF), the U.S. consumer price index (IFS line
54 11164ZF), and the country consumer price index (IFS line 64ZF). The real exchange is calculated as the foreign exchange rate multiplied by the U.S. consumer price index divided b y the country consumer price i n dex ( Kipici and Kesriyeli 1997) The Hodrick Prescott filter ( % = 14,400) is used to calculate the trend from this data and the final indicator is the real exchange rate trend subtracted from the real exchange rate for a giv en month. 3. Domestic credit (.2*, upper, 12 month percentage change) Substantial buildup of domestic credit is at the heart of the first generation currency crisis model. Excessive domestic credit expansion can cause the supply of domestic money to outs trip demand for domestic money, causing market agents to prefer to hold foreign money at the current exchange rate. Foreign exchange reserves dwindle as more and more agents exchange their domestic money until the Central Bank can no longer maintain the p eg and the regime collapses. An increase in the level of domestic credit is often a precursor to an exchange rate regime switch. Domestic credit is calculated as total domestic credit (IFS line 32ZF). The critical region of .2 is starred because .2 is t he region for domestic credit divided by GDP in Edison (2003), not just domestic credit. 4. Credit to the public sector (.1*, upper, 12 month percentage change) KLR (1998, p. 12) found that credit to the public sector received "ample support" as a usef ul indicator in the currency crisis literature. This is not surprising, as a country with a large amount of credit often finances it with foreign sources and thus makes payments in foreign currency. Additionally, both foreign and domestically financed cr edit can increase the domestic money supply without a concurrent increase in demand.
55 Credit to the public sector is calculated as net claims on central government (IFS line 32AN.ZF). 5. M1 (.1*, upper, 12 month percentage change) Excessive g rowth in M1 is indicative of loose monetary policy, and can lead to an outsize increase in the money supply and thus an increased demand for foreign exchange reserves. M1 is money (IFS line 34...ZF). 6. M2 / foreign exchange reserves (.1, upper, 12 month percentage change) An increase in M2 relative to reserves often is seen before a crisis as expansionary monetary policy and / or a decline in reserves both put pressure on the peg. M2 is money plus quasi money (IFS line 35LZF), while foreign exchange rese rves are total reserves minus gold (IFS line 1L.DZF). 7. Domestic inflation differential (.1*, upper, 12 month level change) Most of the foreign exchange rate models covered in the literature review incorporate an inflation differential, because the inf lation of a currency is the rate at which purchasing power is lost. If domestic inflation outpaces foreign inflation, the demand for foreign currency will increase relative to domestic currency in an effort to preserve the value of assets. The domestic in flation differential is calculated from the country consumer price index (IFS line 64...ZF) and the U.S. consumer price index (IFS line 11164...ZF). Inflation is calculated on a monthly basis for the country and the U.S. and then the U.S. rate is subtract ed from the country rate. 8. Real interest rate differential (.1, lower, 12 month level change)
56 Whenever real U.S. interest rates rise relative to real interest rates of a country, there is usually a capital outflow from that country as investors can g ain a better return on their assets. This causes a loss of foreign exchange reserves. The domestic nominal interest rate is the country's lending interest rate (IFS li n e 60P ..ZF) and the foreign nominal interest rate is the U.S.'s lending interest rate (I FS 11160PZF). To convert the nominal interest rates to real terms the country consumer price index (IFS line 64...ZF) and the U.S. consumer price index (IFS line 11164...ZF) are used. First, annual nominal interest rates are used to construct a measur e of monthly nominal interest rates Second monthly real interest rates are calculated by subtracting the monthly inflation rate, as calculated using the respective consumer price indexes. Third the domestic monthly real interest rate is subtracted from the U.S. monthly real interest rate to create the real interest rate differential. Unfortunately, variables like external debt measures and current / financial / capital account are not measured at monthly frequency and cannot justifiably be interpolated to monthly frequency. However, KLR (1998, p. 12) found that neither external debt measures nor the current account received much support in the literature as useful indicators of currency crises. 2.3.2 Definition of a Currency Crisis Edison and KLR bo th utilize deviations in an exchange market pressure index to define currency crisis periods. As stated above, this index is the weighted average of the percentage change in the exchange rate and the percentage change in f oreign exchange reserves. This m ethod of defining currency crises was rejected for the compa r ison countrie s as the present research is not concerned with speculative attacks that involve
57 significant reserve loss but do not result in a regime switch; instead, it is concerned with periods when an actual devaluation took place. Thus, this thesis adopts another conventional measure of exchange rate crises formalized by Frankel and Rose (1996), reported in Edison (2003), and modified from log changes to percentage changes as in Berg and Patti llo (1999). This study denotes a month as containing a currency crisis if there is a 25% or larger depreciation of the nominal exchange rate compared to the previous month occurs, given that this depreciation "exceeds the previous year's change by a mar gin of at least 10%" (Edison, 2003, p. 23). Requiring that the monthly depreciation exceeds the amount of depreciation over the previous year by at least 10% eliminates depicting a month as containing a currency crisis when in fact it is just part of a lo ng, downward trend in the nominal exchange rate. Barbados has never experienced a forced devaluation, which is the motivating factor of this thesis, and thus cannot use significant depreciation to signify a troubled currency time. Instead, the second component of the exchange market pressure index, a significant loss of foreign exchange reserves (b oth monthly and yearly), is used as the primary indicator of pressure to devalue. 2.3.3 Crisis Window Selection This thesis examine s the twenty four months prior to the month of the currency crisis in each of t he countries. However, it also examine s the month of the currency crisis and the twenty four months after the currency crisis for each of the indicators. This expanded period o f analysis allows for a better understanding of the factors leading up to the crisis and how these factors changed during and after the crisis.
58 2.3.4 Comparison Country Selection The choice of comparison countries is one of the defining aspects of this thesis' methodology. Ideally, a comparison country would be exactly similar to Barbados in all respects, except for the fact that it experienced exchange rate depreciation or devaluation and different economic circumstances leading up to that regime chang e True counterfactual s as described above cannot be found in the real world, and thus use must be made of the countries and data that are available. Ultimately, Argentina, Jamaica, and Trinidad and Tobago were chosen as the comparators; all of these cou ntries have experienced exchange rate crises resulting in devaluation, depreciation, and/ or abandonment of their pegs to the U.S. dollar. Barbados is a geographically small island nation in the Caribbean that has a British colonial history and a relati vely small population of just fewer than 300,000 persons. Jamaica and Trinidad and Tobago are also both small, island nations located in the Caribbean that share Barbados' British colonial history and have populations of about 2.8 million and 1.3 million persons, respectively. All three countries are members of the Caribbean Community (CARICOM) and are fairly dependent on tourism. Argentina is the only non Caribbean and non English speaking comparison country Its statistics are fairly different from th ose of the three other countries in the study, but is still located in Latin America, a region often compared to the Caribbean. Its presence provides a country with a fairly spectacular exchange rate crisis and a country that also has better data availabi lity than the other comparator s. Altogether, the three co mparison countries provide a broad range of experiences to compare with Barbados experience
59 2.3.5 Currency Crisis Selection This thesis ut i lizes one crisis from each of the comparator s to compare to one crisis of Barbados. Because o f the univariate comparison method undertaken, including more crises for either Barbados or the comparison countries would result in an unmanageable number of indicator / comparison country crisis / Barbado s crisis combinations. In a previous study undertaken by th is researcher a method of aggregating data across multiple crises in the comparison countries history was used to reduce the number of comparisons necessary, but this suffered from various analy sis issues. Among them were how to interpret averaged indicator values and how to account for a smoothing of indicator values, because signaling behavior did not always occur in the same month or months leading up to a crisis. Although Barbados has expe rienced significant pressure on its foreign exchange rate to point of requesting IMF assistance in three different instances, the 1991 crisis is considered to be the most severe and thus posed the greatest risk for depreciation or devaluation (Worrell et a l., 2003, p. 3). This crisis is examined in this thesis, because understanding why Barbados was able to maintain stability throughout this most extreme period will provide a better framework for understanding less severe crises. Another consideration is which month of the 1991 crisis is considered the crisis month. As stated in the definition of a currency crisis above, the main indicator is a large loss of foreign exchange reserves. August 1991 was chosen as the crisis month because, of all the months in 1991, the island experienced the largest loss of exchange reserves on both a monthly and yearly basis in that month. Additionally, the International Monetary Fund began negotiations for assistance with Barbados in August of 1991
60 For the Caribbean com parator s, Jamaica and Trinidad and Tobago, the month during which the largest month over month depreciation occurred was selected as the currency crisis month. For Jamaica, this is November 1983, when the island experienced a 75% depreciation following th e adoption of an exchange rate auction system. For Trinidad and Tobago, this is December 1985, when the country experi e nced a 39% devaluation. Worrell et al (1998) support the selection of these crises in their examination of Caribbean exchange rate cri ses by providing analysis of the turmoil before the crises and the large impact of the crises Argentina's 2002 currency crisis is widely considered a s the country's worst one in terms of impact. This perception and the long period of tranquility leading up to the January depreciation of 40% influenced its selection as the examined currency crisis. 2.3.6 Data Range Selection This thesis will use data from June 197 6, when Trinidad and Tobago was the last country to adopt a fixed exchange rate, to December 2007, the latest date that comparable data is available from accessible sources. Data from this period will be used to calculate the means and standard deviations of each indicator for each country. 2.3.7 Indicator Threshold Selection This thesis use s the noise minimizing threshold percentiles that Edison (2003) calculated for the indicators that are shared between her study and the current study For indica tors not included in her study, the tenth percentile is used, which is on the noise minimizing side of the typical indicator threshold value continuum.
61 2 .4 Modification of the Signal Extraction Approach KLR and Edison's methodolog ies treat all signals and all absences of signals the same. An indicator exceeding a threshold by 5% is considered equivalent to an indicator exceeding a threshold by 200 %, as a signal is issued in either case, even though these differences could result in d rastically different pressures on an exchange rate regime. Additionally, an indicator that is 5% below a threshold is considered the same as an indicator that is 200% below, as a signal is not issued in either case. Thus, their methodologies mask importa nt differences in indicator behavior and country experiences This thesis improves their methodologies by further transforming the indicators in two steps to take into account the actual severity of indicator behavior First, for each month of the data set, the difference between the original indicator and the threshold value for that indicator is taken (or the difference between the threshold value and the original indicator, so that a signal yields a positive value). Second, this difference is then di vided by the threshold value to arrive at the final indicator value for each month: the percentage above or below the threshold. Now, instead of just making an inter country comparison based on whether or not an indicator signaled for a given month, it is possible to compare to what extent a country's indicator fell below, approached, or exceeded a threshold. Furthermore, a graphical analysis that yields a truly comparable picture of a country's experience with regard to an indicator is now possible 2. 5 Application of the Signal Extraction Approach The research hypothesis this thesis aims to test is that Barbados experienced less economic turmoil leading up to its 1991 crisis than the comparison countries experienced
62 leading up to their crises. If the hypothesis is not rejected, the conclusion is that Barbados' comparative lack of pressure on its exchange rate regime enabled the island to maintain its peg to the U.S. dollar. If the hypothesis is rejected, the conclusion is that certain factors enab led Barbados to maintain its fixed exchange rate regime despite experiencing turmoil substantial enough to cause the loss of a peg in the comparison countries. Thus, the method of a nalysis focus es on comparing each country's indicator performance with Bar bados indicator performance both in terms of individual indicators and the aggregate performance of all of a country's i ndicators to determine comparative economic turmoil The comparison methodology this thesis employs is essentially graphical in natu re. The horizontal axis is the month of the indicator data point. The crisis month is in the center of the horizontal axis, with the twenty four months before the crisis to the left and the twenty four after the crisis to the right. A vertical line at m onth twenty five highlight s when the crisis o ccurred. The vertical axis measure s the percentage above or below the threshold for the indicator, with a dark horizontal line across the graph indicating the threshold at the 0% point. Ligh ter horizontal lines are drawn at each 100% interval above and below the threshold. Adverse performance that is indicative of a currency crisis is demonstrated by signaling behavior exhibiting one or more of the following characteristics: A. Above threshold behavior: indic ator values close to or above the threshold. B. Prol onged above threshold behavior: indicator values displ aying the above threshold behavior indicator behavior for multiple months, with months closer to the crisis date receiving heavier weighting.
63 C. Rapid thres hold behavior: indicator values that rapidly move toward and above the threshold, with months exhibiting this behavior that are closer to the crisis receiving heavier weighting. The opp o site behavior of A C indicat e less adverse change for an indicator. Indicator values that are more than 100% below the threshold actually signify a beneficial change in the underlying variable. Each indicator of each country is analyzed using the above method. Then, each indicator's performance is ch aracterized as neutr al, mild, or severe for each country and presented in t able 3.1 Finally the numbe r of neutral, mild, and severe indicators for each country ar e tabulated in t able 3.2. This table provides the basis for country comparison, in that the country with the most severe and mild indicators had the worst economic experience leading up to its crisis, according to this thesis' methodology Analysis of indicator behavior subseque nt to the crisis is not as critical to the present research, but can provide an idea of government / wider economic response to the exchange rate crisis and how the economy recovered (or didn't recover) following the crisis.
64 Chapter 3 Presentation and Analysis of Results 3.1 Introduction An analysis of the data described in the previous chapter, according to the methodology in the previous chapter, follows The aim of the analysis is to determine whether Barbados' economic experie nc e before its 1991 currency crisis was more or less adverse than the experiences of the comparison countries during their respective crises. The three main indicators of economic trauma according to this thesis' methodology are an indicator passing above the threshold, an indicator staying above the threshold for an extended length of time, and a rapid movement to a position above the threshold. The graphs in the a ppendix were used to analyze indicator experience. The percentage above or below the threshold is on the vertical axis and the month that the indicator is reported for is on the horizontal axis. The "B" refers to the number of months before the crisis, "Crisis" refers to the month of the crisis, and "A" refers to the number of months after the crisis. If the indicator is more than 100% below the threshold, there is an advantageous change in the indicator. 3.2 Credit to the Public Sector Barbados Around 14 months before Barbados' crisis, a buildup in credit to the public sector began to occur. The indicator continued to approach the threshold, coming within about 50% of threshold for about four months. While there was no alarming increase visible Barbados' government was already over extended and thus this increase in the amount of credit, though small in percentage terms, was large in dollar terms. After the crisis, Barbados' indicator diverged from the threshold, although credit was still increasing.
65 Argentina Leading up to th e crisis, Argentina did not experience any adverse or positive changes in credit to the public sector. After the crisis, there was a slight build up in credit for about 12 months, although at no point was Argentina closer than 90% below the threshold. Jam aica Jamaica's credit to the public sector was neutral over the entire time period, with just a slight, constant increase in credit. Trinidad and Tobago Up until 12 months after the crisis, Trinidad and Tobago was neutral with regard to credit to the publi c sector growth. However, between 12 and 24 months after the crisis, there were several sharp drops in the amount of credit from a year before. In month 20, credit to the public sector surpassed the threshold by about 50%. Summary Overall, Barbados faire d the worst with regard to credit to the public sector leading up to the crisis month. It registered many months of increases in credit (any data point above 100%), especially in the 12 months before the crisis. However, none of these months had values that were closer to the threshold than 50% below it 3.3 Domestic Credit Barbados Twenty three months before the crisis, Barbados' domestic credit surpasse d the t hreshold. It consistently grew over the entire period, declining to about 60% below the threshold in month 11 before gradually moving back toward the threshold until the crisis
66 month. Afterward, the indicator drops significantly, although the level of domestic credit does not truly decrease at any point. Argentina The level of dom estic credit remains unchanged for Argentina up until the month of the crisis, when the indicator starts to increase until it reaches about 50% below the threshold in the eighth month after the crisis From that point on the indicator declines again until the level of domestic credit does not c hange, starting in month 16 after the crisis. It does not seem as if a build up of domestic credit was the cause of Argentina's crisis. Ja m aica Jamaica's level of domestic credit remains unchanged over the entire f orty nine month p eriod Trinidad and Tobago For all months except 24 20 before the crisis, Trinidad and Tobago's level of domestic credit remains unchanged. Summary Barbados and Argentina are the only countries that showed any significant movement in the level of domestic credit. Argentina witnessed an increase in domestic credit after the crisis month, with the indicator reaching a maximum of 50% below the threshold. On the other hand, Barbados surpassed the threshold once and hovered near the threshold especially in the period leading up the crisis. This indicates that there was significant domestic credit build up leading up to the crisis.
67 3.4 Foreign Exchange Reserves Barbados Barbados started the examination period at the threshold with regard t o foreign exchange reserves before plunging down to about 175% below the threshold (signaling an increase in foreign exchange reserves) and then shooting up above the threshold (a maximum of 75% above for one month) for the last five months before the cris is. Above threshold values of the ind icator continued until month 6 after the crisis when Barbados began to rapidly gain back foreign exchange reserves. It continued to gain foreign exchange reserves for the rest of the months in the twenty four month pe riod after the crisis. Argentina Argentina's foreign exchange reserves were fairly stable until about month 12 before the crisis, when the indicator began to gradually move closer to t he threshold. From month 12 before to month 18 after the crisis, Argent ina was losing foreign exchange reserves. However, the indicator never passed above 40% below the threshold. Jamaica The island had highly erratic indicator behavior, alternating between large increases in foreign exchange reserves and large decreases in foreign exchange reserves with the indicator reaching the threshold at three separate instances before the crisis. In the eight months before the crisis, the indicator was always 60% below the threshold or closer. Overall, Jamaica saw significant pressur e from its foreign exchange reserve losses. Trinidad and Tobago
68 Trinidad and Tobago's indicator was always within 50% of the threshold leading up to the crisis, except for the month right before the crisis. This shows a continual loss of reserves and significant pressure on the exchange rate. After the crisis, reserve loss accelerated, probably do to uncertaint y about the value of the Trinidad and Tobago dollar. Summary Barbados was the only country with above threshold behavior with regard to foreign exchange reserves and clearly had the worst experience leading up to its crisis. However, it also had periods of strong increases in foreign exchange reserves while other countries like Argentin a and Trinidad and Tobago on ly saw erosion of reserves. Jam aica was similar to Barbados in that it had periods of growth and decline in reserves, although the declines were not bad compared to Barbados. 3.5 M1 Barbados The island saw decreases in M 1 until about month 8 before the crisis. From that point on, M1 was always increasing, with two significant spikes at about 50% below and 10% below the threshold before the crisis. The 10% spike was at month three before the crisis. Argentina Argentina did not experience any decline or growth in M1 leading up to or af ter its crisis. Jamaica
69 At 17 and 15 months before the crisis, Jamaica experienced two indicator spikes about 10% above the threshold. From that point on, the indicator gradually declined as the crisis approached. However, M1 was always increasing and in creased dramatically (still below the threshold) again right after the crisis before oscillating in the region between the 100% line and the threshold. There was one more spike at month 19 after the crisis. Trinidad and Tobago Trinidad and Tobago experi enced slight decreases in M1 for the entire period leading up to the crisis and the vast majority of the months after the crisis. Summary Jamaica had the most harmful experience with the M1 variable, although Barbados was not substantially different becaus e of the great pressure it experienced closer to the crisis. Neither Argentina nor Trinidad and Tobago showed much movement in M1. 3.6 M2 / Foreign Exchange Reserves Barbados The island's experience with M2 / foreign exchange reserves is similar to its experience with the foreign exchange reserves indicator, although this indicator shows more adverse changes. Again, there is above and near threshold behavior at the beginning of the observation period, before a decline in the indicator and a slight decli ne in the underlying va riable. Starting in month 11 before the crisis, there is a rapid build up in M2 / foreign exchange reserves with month 2 before the crisis actually registering more than 100% above the threshold. Similar to the foreign exchange re serves indicator,
70 there is a sharp decline in the indicator following the crisis and over the entire after crisis period. Argentina Unlike Argentina's foreign exchange reserves indicator, which shows some adverse behavior leading up to and after the crisis its M2 / foreign exchange reserves indicator is flat at 100% below the threshold over the entire observation period. Jamaica Over almost the entire observation period leading up to the crisis, Jamaica witnesses an increase in M2 relative to foreign exchange reserves. Nine of the twelve months before the crisis are above the threshold, indicating significant negative pressure on the foreign exchange rate. Trinidad and Tobago Over the twenty four months leading up to the crisis, Trinidad and Tobago is increasing M2 / foreign exchange reserves. The indicator consistently stays within 60% of the threshold except for two months before the crisis when it declines to about 75% below the threshold. Interestingly, there is a large spike abov e the threshold around month 14 after the crisis and much activity around the threshold up until the end of the observation period. Summary While Barbados was the only country to spike above 100% above the threshold with regard to M2 / foreign exchange reserves, Jamaica h ad more consistent above threshold behavior. In sum, both seem to have had an equal experience with the
71 indicator. Argentina and Trinidad and Tobago did not show any important movements in the indicator. 3.7 Real Exchange Rate Deviation from Trend Ba rbados Barbados' indicator was above the threshold for months 24 22 before the crisis but then the country gained competitiveness leading up to the crisis month. After the crisis, competitiveness decreased with several spikes above the threshold, includi ng one that was 75% above the threshold. Argentina Argentina had a gradual loss of competitiveness in the two years leading up to the crisis, with the indicator surpassing the thr eshold around month 7 before the crisis and continuing to register above the threshold until the devaluation. Jamaica Jamaica followed the same gradual loss of competiveness as Argentina, but with a maximum distance of 115% above the threshold. After the devaluation, it too immediately gained competitiveness and continued to do so for the rest of the observation period. Trinidad and Tobago Trinidad and Tobago was like the other two comparator s in experiencing a loss of competitiveness leading up to the crisis. However, it spiked the highest above the threshold, at about 130%. The country also experienced the greatest subsequent increase in competitiveness following its devaluation. Summary
72 Barbados' experience with its real exchange rate pales in comparison to the three comparator s. All of them witnessed extreme losses of competitiveness leading up to their crisis months. 3.8 Real Interest Rate Differential Barbados Barbados' real inter est rate differential indicator rose above the threshold once about nine months before the crisis and touched the threshold two time s after that and before the crisis. In the nine months leading up to the crisis, there was generally adverse behavior in th e indicator. After the crisis, there was an enormous spike in the indicator of about 150% above the threshold, as well as another spike of about 75% above the threshold. These were probably due to fears about the stability of Barbados' economy and the ne cessary interest rates that had to be charged to justify lend ing money. Argentina The country's real interest rate differential had some ad verse changes in the 24 12 months before the crisis. However, from that point on, there was no visible pressure on the foreign exchange rate until the month of the crisis (the month before the crisis is a missing data point). About twenty months after the crisis, there was a larg e spike in the differential, and the indicator was located at about 200% above the threshold for several months. Jamaic a Jamaica's indicator did not pass above the threshold leading up to its crisis. However, there was an increase in the gap between Jamai ca's interest rate and that of the
73 U.S, as evidenced by indicator behavior above 100%. After the crisis, there was one spike above the threshold at about 50% above the threshold. Trinidad and Tobago The islands had a roller coaster experience leading up to and following their crisis. There were four points where the indicator almost surpassed the threshold, coupled with many significant closings of the gap in the interest rates. After the crisis, there were seven spikes at or above the threshold. Summar y Barbados was the only country with any movement above the threshold, and its experience appears to be slightly worse than that of Trinidad and Tobago. Neither Argentina nor Jamaica approached those countries' severity of experience, although they did ha ve several increases in their interest rates compared to the U.S. 3.9 Inflation Differential Barbados The island had one spike above the threshold for the inflation di fferential, with two months of very close values leading up to the crisis. The majorit y of the pre crisis period was spent in between the threshold and 100% below the threshold, indicating an adverse change in the underlying variable. After the crisis, Barbados witnessed a large spike of about 150% above the threshold and smaller spikes th at either reached or just passed the threshold. Otherwise, its indicator oscillated around the 100% below the threshold line. Argentina
74 Argentina had a very neutral experience with regard to its inflation differential, as its indicator did not stray far from 100% below the threshold. Jamaica Like Barbados, Jamaica spent most of its pre crisis time in between the threshold and 100% below the threshold. However, it did not have the advantageous reversal months that Barbados experienced for a few months lea ding up to the crisis. After the crisis, there was one spike of about 50% above the threshold and generally mild performance on average otherwise. Trinidad and Tobago Trinidad and Tobago experienced large oscillations around the 100% line and the indicat or almost surpassed the threshold three separate times. In the months directly preceding the crisis, the indicator was more neutral and actually showed beneficial change the month of the crisis. Summary Overall, Barbados and Trinidad and Tobago appeared t o have the most problematic experiences leading up to their crises. Barbados' seemed to have a problematic experience because of threshold behavior and one month above the threshold very close to the crisis and Trinidad and Tobago because of the long span of adverse behavior. Jamaica had some adverse indicator behavior leading up to the crisis, while Argentina was very docile.
75 3.10 Summary of Findings Before evaluating whether Barbados experienced less economic pressure leading up to its crisis, two tables will be presented that summarize the results of the above analysis. Table 3.1 condenses the experience of the different countries for each of the indicators with characterizations of severe, mild, or neutral. This is done with a focus on the time p eriod preceding the crisis. Table 3.1 : Detailed Country Experiences Barbados Argentina Jamaica Trinidad Credit to the Public Sector Neutral / Mild Neutral Neutral Neutral Domestic Credit Mild Neutral Neutral Neutral Foreign Exchange Reserves Mild / Severe Mild Mild / Severe Mild M1 Mild Neutral Mild /Severe Neutral M2 / Foreign Exchange Reserves Severe Neutral Severe Mild Real Exchange Rate Deviation from Trend Neutral Severe Severe Severe Real Interest Rate Differential Mild / Severe Neutral Mild Mild / Severe Inflation Differential Mild / Severe Neutral Mild Mild Table 3.2 below provides a summary of the number of each characterization each country received.
76 Table 3.2 : Summarize d Country Experiences Barbados Argentina Jamaica Trinidad Severe 2.5 1 3 1.5 Mild 4 1 3 3.5 Neutral 1.5 6 2 3 3. 11 Discussion of Findings From the aggregates presented above, Jamaica appears to be the only country that suffered more adverse performance in its indicators than Barbados. Trinidad and Tobago followed Barbados, with one less severe characterization and Argentina did not seem to have experienced any particularly drastic economic turmoil. This finding leads to a preliminary rejection of the research hypothesis and conclusion that Barbados' exchange rate regime stability does not stem from mild economic conditions leading up to its crisis. The hypothesis receives only a preliminary rejection because of the methodological problems that are presented in the concluding chapter, which might undermine the validity of the study. Barbados performed most severely on foreign exchange reserv es, M2 / foreign exchange reserves, the real interest rate differential, and the inflation differential. The indicators for domestic credit and M1 were mild. Together, these findings indicate that Barbados suffered from a first generation currency crisis The money supply expanded and residents increased demand for foreign currency was untenable with the supply of foreign currency. From the examination of the 1991 crisis in the second chapter, it is clear that foreign debt also played a role in the run down of foreign exchange reserves that ultimately brought the country to a near loss of the fixed exchange rate. Argentina's only severely performing indicator was the real exchange rate. For an export dependent country such as Argentina, the loss of ex port competitiveness can
77 and did have a major impact, especially with the revaluation of the Brazilian real and international revaluation of the U.S dollar. Like Barbados, Argentina also suffered from a build up of foreign debt that it ultimately had to d efault on in a twin banking and currency crisis. The anticipation of this default and capital flight caused the mild characterization of the foreign exchange reserves indicator. Jamaica was dealt a dual blow leading up to its currency crisis: M2 / foreig n exchange reserves and the real exchange rate both moved severely adversely. Thus, Jamaica had a growth of broad money that outpaced its foreign exchange reserves and caused demand for foreign currencies to outpace supply and a loss of foreign exchange e arning power with the decrease in competitiveness of its exports. It is no wonder that Jamaica suf f ered a devaluation because of the perfect storm that affected it. Trinidad and Tobago also suffered from a severe drop in competitiveness due to the adve rse change in its real exchange rate. It is a very export dependent economy and thus could not accommodate the effects on its exports. Additionally, it also suffered from an increase in its real interest rate as compared to the U.S., a proxy for market p articipants' perception of increased risk in the economy as compared to the U.S. Foreign exchange reserves, M2 / foreign exchange reserves, and the inflation differential also showed mild behavior leading up to the exchange rate crisis. The signal extrac tion approached that was modified to investigate compare the currency crisis experiences of Barbados and the comparator s does not typically register a large number of signals leading up to a crisis. For the countries chosen, there was a surprising amount of notification of the cri ses, leading to a preliminary conclusion that the approach is a useful way to compare experiences. With the comparison, it has been
78 seen that Barbados did not escape from problematic movements in indicators of currency crises. T his rejection of this thesis' hypothesis leads to an investigation of why Barbados was able to maintain stability that is addressed in the concluding chapter. Chapter 4 Conclusion will examine methodological issues and unique characteristics of the isla nd that might explain the rejection of the hypothesis as well suggest avenues for future research and provide concluding remarks.
79 Chapter 4 Conclusion 4.1 Introduction Chapter 4 presents some methodological issues that might explain why the findings presented in Chapter 3 reject the hypothesis that Barbados has maintained its fixed exchange rate because of mild economic conditions leading up to its exchange rate crisis. The intuition is that the methodology used might have some flaws that invalidate the conclusions. Another explanation of this thesis' findings is that unique characteristics of Barbados have enabled it to ma intain exchange rate stability despite volatile economic conditions. These characteristics are examined in section 4.3 Conjectures about Barbados' Stability. Section 4. 4 Opportunities for Further Research provides ways to improve the methodology and applications other than the thesis question, as well as pr omising avenues for investigation of the un ique characteristics of Barbados and how they compare to the comparator s Section 4.5 Conclusion completes this thesis by providing an overview of the findings and hi ghlighting the contributions this thesis mak es to the literature. 4. 2 Methodological Issues and Potential Corrections This section provides an in depth analysis of the different aspects of the methodology that might have led to an erroneous conclusion regarding the severity of Barbados' experience leading up to its 1991 crisis. Additionally, it provides possible actions to address these issues.
80 4. 2 .1 Signal Extraction Approach There are two main problems with the signal extraction approach that was modified for use in this thesis. The first problem is that the entire approach is predicated on the assumption that drast ic changes in underlying indicators are indicative of an impending currency crisis. The second problem, the low predictive power of the signal extraction approach, is probably partially a result of the first. An indicator in the signal extraction approac h will only issue a signal (pass above the threshold) if the twelve month percentage change in the indicator (other than interest rate, inflation, and real exchange rate indicators) is an extreme value. On the other hand, a slow but steady adverse change in an indicator over the two years leading up to a crisis would be ignored. The problem is that both scenarios could result in an indicator, for example foreign exchange reserves, reaching a critical level. However, the signal extraction approach will no t pick up on the gradual decline in foreign exchange reserves and thus not indicate the dire consequences of such movement. With regard to the current thesis question, the comparison countrie s might have more dire economic conditions than Barbados, but be cause these were developed over time, they are not considered. The fact that the signal extraction approach ignores slow but steady adverse changes might explain why the currency crisis prediction literature regards the approach as somewhat unreliable. B erg and Pattillo (1999) examined the ability of several different currency crisis prediction models' ability to predict the Asian financial crisis of 1997. They found that the signal extraction model was the only one that "achieved a measure of success" ( Berg and Pattillo, 1999, p. 127). However, this must be understood
81 against their finding that "most crises are still missed and most alarms are false" (Berg and Pattillo, 1999, p. 131). The inability of the approach to accurately predict crises calls int o question its ability to explain crises as well. Given that signals often produced false alarms and that the model also did not predict many crises that happened, one cannot put too much weight into an analysis of the severity of the economic situation l eading up to a currency crisis by the number of and severity of signals issued. Furthermore, the traditional signal extraction approached used a composite indicator that took into account all of the various indicators. Since this thesis does a comparison of individual indicators, there is even less support for conclusions drawn from their isolated signals. The lack of a numerical aggregation method is somewhat mitigated by the summing of the neutral, mild, and severe characterizations for each country in the previous chapter. 4. 2 .2 Indicators Two issues with the indicators that make up the analysis in this thesis are present: important indicators have been excluded and it is not clear how to weigh the importance of the various indicato rs for Barbados and the comparison countrie s. This thesis used monthly data so that a high level of detail could be captured when examining the experience of the various countries leading up to the crisis. However, this precluded the inclusion of many important indicat ors. It was felt that it would methodologically unsound to interpolate annual values t o monthly values and thus indicators only available in annual frequency have been left out including GDP, unemployment, balance of payment variables, and governance mea sures Debt statistics are crucial for understanding currency crises, and these too have been exclud ed, because data could not be acquire d For Barbados, service exports are a main earner of foreign exchange
82 because of the central nature of the tourism i ndustry to the Barbadian economy. However, data for this indicator could not be found and thus the indicator was excluded Another issue with excluded indicators is that Barbados and the comparison countrie s do not operate in vacuums and are impacted by the world around them. The fact that their crises did not occur during the same month means that non country specific factors could have affected each of the countries, factors that were not measured by the indicators that this study used. The central im portance of service exports to Barbados highlights the other crucial problem with the analysis. It is unclear how to weight the importance of the various indicators when the countries examined ar e fairly heterogeneous. For example, while service export levels are critical to Barbados, they might not be of any importance to Argentina. Without a clear method of determining weights or incorporating them into the methodology, all indicators were treated as equally important. The problem with this method is that a country might have received a severe characterization of an indicator that is of no importance, and yet was treated as evidence of economic turmoil. The signal extraction approach does offer one method of weighting indicators, namely assigning we ights based on the ability of an indicator to predict a currency crisis, aggregated over the entire sample of countries. However, these weights are universally applied for all countries and thus each country will receive the same weighting for a given ind icator. The impact of equally weighted indicators is that the severity of a country's experience leading up to a crisis could be extremely skewed, in that several indicators which are unimportant for the country at hand are important for the other countri es in the calibration model. This leads to the same result as the example above.
83 4. 2 .3 Crisis Window The crisis window of twenty four months before a crisis was adopted by the signal extraction approach a priori ( KLR 1998, p. 17). There is no empiric al reason why this is the correct period of observation and thus a different time span might yield more accurate conclusions. 4.2 .4 Comparison Countries A critical component of this thesis methodology is the choice of comparison countrie s. Ideally, thes e countries would be exactly similar in make up to Barbados, except would have suffered a currency crisis. In the real world, this is not possible and thus proxies for true counterfactuals were found in Argentina, Jamaica, and Trinidad and Tobago. These countries have heterogeneous experiences that make it difficult to validly compare their indicator performance. If Barbados suffers a currency crisis because of an increase in the money supply and Argentina experiences a currency crisis because of capital flight, it is very difficult to determine the relative severity of their crises. 4. 2 .5 Currency Crises For Barbados, the most severe currency crisis was chosen as the crisis to investigate because it offered the greatest chance of devaluation or depreci ation; if it could be understood why Barbados did not have a regime change when it was most likely to, then explaining the other currency crises ought to be much easier. However, the choice of a currency crisis for the other countries is a much more compl icated matter, as each country had several crises in which a devaluation or depreciation occurred. It is not clear whether the most severe or the least severe crisis ought to be the one examined for each comparison country The least severe crisis would probably offer the closest
84 comparison to Barbados, as the tipping point between devaluation / depreciation and exchange rate stability probably lies somewhere in between the comparator s experience and Barbados' experience. On the other hand, if Barbados were to experience a similar amount of economic turmoil as the comparison countrie s during their most severe crises, it would be much easier to reject the hypothesis. This is because even if Barbados were to have a slightly less traumatic time leading up to its crisis, the comparators would be very extreme crises and it would not be hard to hypothesize that Barbados could have sufficiently qualified for one of their less extreme crises. The choice of using the most extreme crises for each comparison country mandate s a criterion or set of criteria for determining the most extreme crisis for each country. Two main methods offer themselves: the crisis that resulted in the largest drop in the value of the currency or the crisis that resulted in the most economic turmoil The former method was used for Jamaica and Trinidad and Tobago, as determining relative economic turmoil raises many more questions of quantification The latter method was used for Argentina because the 2002 crisis was very clearly the most severe and exhibited a large (though not the largest) drop in the currency's value. While other crises had larger drops in the currency's va lue, they were not as impactful. The preceding discussion of the decision of crises brings several points to ligh t. First is the assumption that if Barbados' experience leading up to its crisis is only slightly less severe than that of the comparison countrie s' worst crises, it probably would have witnessed a crisis except for some outside factors like the social structure of t he island. The problem with applying this assumption is that it is not clear how the point at which Barbados' experience becomes too mild compared to the comparator s to warrant the
85 assumption that Barbados would have experienced a devaluation or depreciat ion. Luckily, aggregating Barbados' indicators placed it in the middle of the comparison countrie s in terms of severity and so this first point does not affect this thesis' conclusions. Second, a mistake could have been made in selecting the most severe crisis for each of the comparison countrie s. This could be because the largest drop in a currency's value is not always indicative of the worst crisis, as is shown in the case of Argentina. Thus, the crises selected for Jamaica and Trinidad and Tobago mi ght ought to be different, which cannot be clear without further research. If in fact the wrong "worst" crisis was selected for any of the comparison countrie s, this thesis methodology is not critically shaken Even if the least severe crisis was select ed, and Barbados did have more economic turmoil leading up to its crisis than the comparison countrie s, then one could preliminary reject the hypothesis because Barbados passed the economic turmoil threshold of the other countries Third, the selection of different currency crises could drastically change the severity of the behavior of the different indicators for the different countries. Thus, without further investigation, it can not be determined how robust the methodology is. 4. 2 .6 Data Range By sel ecting June 1976, when Trinidad and Tobago adopted a fixed exchange rate to the U.S. dollar, as the date to start the calibration of the thresholds, this thesis made the implicit assumption that it is best to use data from when a country is under a fixed e xchange rate. All of the comparison countrie s experienced months outside of a fixed exchange rate during the period between June 1976 and December 2007 and so violated
86 the mandate of the assumption. Therefore, the data that is fed into each indicator's t hreshold calculation might render comparisons between the comparator s and Barbados biased. This is because Barbados' indicator thresholds are calculated under only a fixed exchange rate while the comparison countrie s include data from floating rate periods. Systematic diffe rences in variable values between floating rate and fixed rate periods could skew the thresho l ds. A potentially skewed indicator can be seen by considering the real exchange rate deviation from tren d indicator. Under a floating exchange rate, the exchange rate component of the real exchange rate will deviate much more than under a fixed rate. This will necessarily cause the threshold to be a higher value, decreasing likelihood that a country w ould reach the threshold under the "milder" fixed rate period. The ultimate impact is that Barbados will appear to have experienced more economic turmoil comparatively than is warranted. While December 2007 was selected because it was the last date for wh ich data was in the correct format for analysis, the most recent date could also have been selected Preliminary examination of the data does not indicate that changing the ending or beginning date would greatly affect the thresholds, but is still another point that weakens the robustness of the methodology. 4. 2 .7 Thresholds There are three main issues with the use of indicator thresholds in this thesis. First, outlier values have an undue affect on the thresholds because the elimination of the largest five values of an indicator often will significantly change the standard deviation and the average of the indicator. Significant change of the standard deviation and the average significantly changes the threshold, thus impacting the number of signals the
87 indicator will issue during the heightened activity of a crisis. If a few outliers have caused the threshold to be very high, even abnormal behavior during a crisis will not be recorded and thus a country will not appear to have economic turmoil leading up to a crisis. Second, a country that has had multiple crises will, all else equal, have fewer signals for each crisis. This results from the calculation of the threshold as the top 10 20% of the most extreme values of an indicator. For example, if th ere are 250 months in the calibration period, then 25 months will be in the top 10% of the most extreme cases. If there is only one crisis, then all 24 months preceding the crisis and the month of the crisis can be one of the 25 extreme months. However, if there are two crises, then those 25 extreme months will be split between the 50 examination months of the country 's two crises Adding more crises will only decrease the number of extreme months available for each crisis of a country Thus, a country like Barbados, which has had three periods of cur r ency trouble will have more extreme months per crisis than a country like Argentina, which has had five crises. This result leads to biased comparisons of Barbados and the comparator s. Third, the economic structure of a country can change over time, leading to a greater or lower tolerance of indicator movement before it succumbs to a currency crisis (Craigwell, 2011) The standard deviation and average of each indicator is time invariant and does not take this possibility into account. A country which on has a very low variation in foreign exchange reserves at the beginning of the calibration period and a very high variation in foreign exchange reserves at the end of the calibration period will have a medi um standard variation that combines both periods. Therefore, the beginning
88 of the calibration period will appear to be more problematic than the end of the calibration period, even if the country's economic structure has changed in such a way to allow for greater variations without adverse impacts. 4. 2 .8 Application of the Signal Extraction Approach While it is believed that the adaptation of the sig nal extraction approach for this thesis has yielded richer insight into the experiences of Barbados and the comparison countrie s than the signal extraction approach could do so, two issues still remain: one is closely related to the modification and one is more general. The first issue is that the adaptation is visual and fairly subjective, while the tradi tional signal extraction approach relies on counting the number of signals that each indicator emits. It is hard to determine what the difference between a severe characterization and a mild characterization is and then to uniformly apply this distinction across four countries and eight indicators for each country. The second issue is that Argentina's 2002 crisis, which is almost unarguably the worst experienced by any of the countries studied, does not appear t o be that severe using the methodology. T his could be an indication that the application of the methodology is faulty, that some important indicators were not included, or that country specific factors caused the fall out of the crisis to be much worse than the indicator behavior leading up to th e crisis herald ed 4.3 Conjectures about Barbados' Stability If the hypothesis that Barbados has avoided an exchange rate regime change because the economic turmoil it has experienced leading up to its 1991 crisis is not as severe as that of comparison c ountries can justifiably be rejected, notwithstanding the
89 methodological issues explored above, then there must be country specific factors that have enabled Barbados' stability. There are two ways these country specific factors could be manifest. First, they could have precluded Barbados from experiencing a regime change passively: these factors would have kept Barbados from reaching the crisis point. However, Barbados did reach the crisis point and in September 1991 had only US$7 million in foreign exc hange reserves, as compared to US$125 300 million in the 1980s (Blackman, 2003, p. 5). Second, these country specific factors could have enabled Barbados to take appropriate actions to recover from the crisis and maintain the regime. This thesis argues that the second type of manifestation can explain the rejection of the hypothesis and stability of Barbados' foreign exchange rate if Barbados' experience was as severe as has been found. T his section seeks to paint a picture of why Barbad os' fixed exchange rate regime has survived. 4. 3 .1 Barbados' Actions In August of 1991, Barbados entered into negotiations with the International Monetary Fund regarding a Stand by Arrangement and access to the IMF's Compensatory and Contingency Financi ng Facility ( READ, n.d. pg. 14). Assistance under the Stand by Arrangement was contingent on measures that aimed to "reduce demand for foreign exchange below the available supply," despite the expected output contraction, higher unemployment, and inflati on in the short run that these measures would bring ( READ, n.d., pg. 14). Actions included expenditure cuts, higher taxes, and higher charges for public services, as well as increased securities requirements for commercial banks and greater commercial ban k flexibility in setting loan rates ( READ,
90 n.d pg. 14). The main path to a decrease in demand for foreign exchange was envisioned through three intermediate steps ( READ, n.d., pg. 14): a. cutting the fiscal deficit from an estimated 5.7% of GDP to 1% in fis cal year 1991/92; b. improving the financial efficiency of public sector enterprises; and c. reducing private sector credit so as to lower spending on imports. The broad measures outlined above are discussed in more detail below. Ultimately, Barbados hoped to employ a real (implicit) devaluation by reducing import expenditure directly, rather than undergo a nominal (explicit) devaluation by reducing the value of the Barbados dollar a gainst the U.S. dollar and circuitously reducing import expenditure ( Amo Ya r t le y, 2011 ). When the first signs of an exchange rate crisis appeared, the government began a slight adjustment process with a budget for fiscal year 1991/92 that involved the introduction of "a stabilization tax on personal incomes, raising consumption taxes by 2 percentage points, raising employer's contribution to the severance payments scheme from 0.25% to 1% of wages and committing to scale back capital ex penditure and keep transfers [to government owned enterprises] in check" (Haynes, 1997, p. 97). However, it quickly became clear that these steps were not drastic enough to instill confidence in the parity and so Barbados commenced discussions with the In ternational Monetary Fund and committed to much more drastic actions. These actions included fiscal measures like increased stabilization / other taxes, reductions in government payroll, and improved financial relations with government owned corporations. The Barbadian government also influenced private sector consumption by calling for reduced private sector wages
91 and a reduction in credit. Monetary policy was also tightened, although not to as great an effect as the other measures. Fiscal M easures T axes a. increase of the stabilization tax from "1.5% to 4% for individuals earning in excess of BDS$15,000" (Haynes, 1997, p. 98) b. increase of "ceilings for contributions to payroll levies and the National Insurance Fund" (Haynes, 1997, p. 98) c. increase of "con tribution rates and income ceilings for unemployment insurance" (Haynes, 1997, p. 98) d. increase of "tariffs and user fees on public transportation, housing, water, postal rates, and natural gas" and general consumption taxes (Haynes, 1997, p. 98) e. imposition of a petroleum tax and luxury goods surcharge (Haynes, 1997, p. 98) Fiscal Measures O ther a. imposition of an 8% wage cut for public sector work er s for 18 months starting in October 1991 (Worre l l et al., 1998, p. 145) b. release of about 10% (3,000) governmen t employees from payroll Interestingly, there was an emphasis on ensuring that there was at least one breadwinner lef t in each household when making severance decisions (Worrell et al., 1998, p. 145; Blackman, 2 003 ) c. closing or sale of public commercial e nterprise that "created fiscal pressures for transfers or loan guarantees P roceeds from the sales in the telecommunications and manufacturing sectors were used to realize foreign exchange earnings at particularly dire times in late 1991 (Haynes, 1997, p 98)
92 d. reduction of capital expenditure by BDS$83.6 million in 1991, as compared to $240 million in 1990 ( READ, 1992 p. 12) e. reduction of government transfers to commercial enterprises by BD S$8.6 million (READ, 1992 p. 12) Monetary M ea s ures a. raising of the discount rate ( READ, 1992, p. 3) b. increase of the proportion of deposits commercial banks had to hold in government securities ( READ, 1992 p. 3) c. removal of the ceiling on the average commercial bank lending rate ( READ, 1992 p. 3) d. decree of a global credit standstill in December 1991 that was designed to immediately decrease expenditure in the private sector to provide time for fiscal measures to take effect (Worrell et al., 2003). The ceiling on the credit commercial banks were allowed to e xtend was removed in May 1993 ( Trent and Wood 2005, p. xxii) Other M ea s ures a. expansion of wage restraint activities to the private sector by calling for a freeze of national basic wages for two year; however, productivity bonuses and profit sharing agreeme nts were still acceptable (Haynes, 1997, p. 102). This comprehensive measure was accomplished through a "tripartite accord on wages, prices, and productivity between government and representative organizations of workers and employers," the members of whic h came to be known as Social Partners (Worrell et al, 200 3 ).
93 b. in addition to borrowing from the International Monetary Fund, the Central Bank of Barbados borrowed from regional central banks and the Royal Bank of Scotland ( READ, 1992 p. 13) c. national disc ussions on the state of the economy and used moral suasion to convince regional firms to keep their foreign exchange in Barbados ( DeLisle, 2011). 4. 3 .2 Results of Actions Barbados fully recovered from its balance of payments difficulties in 1991. It did not have to devalue, as the IMF had initially suggested during negotiations and quickly returned to an adequate level of reserves. The fiscal measures and wage restraint policies caused significant import contraction (imports fell from 48% of GDP to 40% of GDP between 1991 and 1992), leading to current account surpluses, which were further s trengthened in 1993 when exports began to recover (Worrell et al., 2003 p. 4 ). The deep a djustment program not only decreased the fiscal deficit from 8% of GDP for FY 1990/91 to 2% for FY 199 1 /92, but was so effective that further measures were unnecessary. In fact, its impact in the last quarter of 1991 was a crucial component of the growth in foreign exchange reserves of $23.1 million ( READ, 1992 p. 13). Once the specter of devaluation or a loss of the peg was dismissed, the government turned to the problem of domestic recession ; f urther economic adjustment was only made through governmen t expenditure restraint (Haynes, 1997, p. 99). Despite Barbados' recovery, it should be noted that the country was not able to execute structural changes outlined in the Stand by Agreement with the IMF and thus was not able to fully utilize the financial resources contained therein (Haynes, 1997, p. 86).
94 4.3 .3 Country S pecific Factors According to the current Governor of the Central Bank of Barbados, DeLisle Worrell (2011), Barbados was able to maintain its fixed exchange rate despite the 1991 crisis b ecause of the draconian correction measures that were taken and the speed with which they were taken. The measures not only directly improved Barbados' position, but also signaled to market agents that Barbados' government was fully committed to maintaini ng the parity, thus avoiding a self fulfilling crisis. This section will examine the features of Barbados that enabled it to undertake the measures that it did in the timeframe that it did. One of the most basic features of Barbadian society is the suppor t for the exchange rate parity with the United States dollar. Neither of the two political parties, the Democ r atic Labour Party nor the Barbados Labour Party would consider abolishing the peg for fear of extreme voter hostility The belief in the mainte nance of the exchange rate stems mainly from two sources: the general perception that it offers an effective way to "contain inflation and provide a stable environment for business decisions" and from Barbadians witnessing "the widening gap between their i mproving living standards and the stagnation or decline in the quality of life in neighboring countries whose currencies were devalued" (Fashoyin 2001 ; Worrell et al, 1998, p. 145). Barbadians did not want to experience the same economic fallout from a d evaluation or loss of the exchange rate peg that Guyana and Jamaica realized, and thus were willing to make sacrifices to preclude this possibility. Another basic feature of the island is the high degree of social cohesiveness (Blackman, 2003, p. 9). It is a small, relatively homogenous island of only about 300,000
95 persons whom are very well known to one another. In fact, policy makers usually attended one of the three elite high schools on the island and grew up together. Furthermore, the D emoc r atic Labour Party is a splinter of the Barbados Labour Party. Both parties are "cut from the same social democratic cloth" and politicians have been known to move from one party to another (Blackman, 2003, p. 9). All of these components sum to a nat ion that is very interconnected, both politically and personally. Thus, when a decision needs to be made, it is likely that Barbadians will support one another and stand by the general consensus of what is best for the island even if they individually di sagree with a measure. T here were demonstrations in the street a gainst the 8% wage cut at first. However, once it beca me apparent that no alternatives to the adjustment policies were being put forth, especially by the most powerful non government group, the trade unions, Barbadians realized that they had to accept the polices to maintain the exchange rate they desired (Worrell et al., 1998, p. 145). Crowe (1999, p. 2) concludes that the combination of the island's strong support for the peg and its soci al cohesiveness were what ultimately enabled the difficult adjustment policies to be sold to the public. Worrell's other condition of success was the speed with which these adjustment policies were implemented. By acting quickly, Barbados was able to prev ent a parallel foreign exchange market from dev eloping (Worrell, 2011 ). This speed can also probably be attributed to the strong support for the peg and the island's cohesiveness. Once the country was in agreement that the peg was to be preserved, despit e the temporary pain of adjustment measures, plans for defending the exchange rate regime were immediately put in place. Worr ell (2011 ) also maintains that policy makers and technocrats during the
96 1991 crisis were very competent and in basic agreement abo ut what had to be done, which further supported a sufficient and quick response to the exchange rate crisis. No matter how quickly Barbados put adjustment measures in place, a recovery of foreign exchange reserves would not happen instantaneously. In int ermediate months, market agents could still have caused the collapse of the peg if th ey lost confidence in the island's ability to engineer a recovery. Thus, another crucial component of Barbados' maintenance of the peg was the country's ability to instil l confidence in its ab i lity weather the crisis. Fortunately, Barbados was operating from a strong position of credibility in 1991. In 1990, the island was ranked first in terms of credit worthiness in the entire Caribbean and Latin American regio n by Institutional Investo r (Dalrymple, 1995, p. 296). Also, unlike many other countries in the region, Barbados had never experienced a forced devaluation or exchange rate regime change, despite periods of trouble in 1977 and 1982. Furthermore, the Inte rnational Monetary Fund's programs seemed to be assured and thus it seemed as if Barbados would receive all of the foreign exchange it needed in the short run to weather the storm. In general, the country had a reputation of good economic management and e conomic agents could rationalize this trouble period as a deviation from an otherwise sup erior record (Dalrymple, 1995, p. 296). Two other very important aspects that im proved confidence in Barbados were the extent of the country's response and what mark et agents would experience with a devaluation. The fiscal components of the recovery plan (not to mention the other components) were so draconian that observers could clearly see that the government was fully committed to maintaining the peg. Perception of this commitment bolstered market
97 agents who were already leanin g toward support of making sacrifices for the regime, as they would have suffered extensively with a devaluation and thus were willing to do what was necessary to ensure the survival of t he peg (Worrell, 2011 ). Despite foreign exchange controls, a run on the Central Bank could have occurred in effect if regional companies lost confidence and moved funds abroad and precipitated the development of a parallel foreign exchange market (Worrell, 2 011 ). This run did not occur. 4. 3 .4 Differences in Actions Between Barbados and Comparison Countries This section provides a very cursory look at the differences between Barbados and the comparison countrie s that might have given Barbados the unique ability to survive its 1991 crisis without devaluing or depreciating According to Worrell (2011 ), the main difference between Barbados and the comparison countrie s was the speed with which it responded to its exchange rate crisis. As noted above, this i s probably due to the cohesiveness of the country, the strong support for the parity, and the competence of those in power. The other countries waited until parallel foreign exchange markets had developed and the market had lost confidence in the fixed exc hange rate before introducing any measures. The measures they did introduce were too little and too late, as bringing stability back to their regimes after a parallel market was already well developed would have be en next to impossible. While citizens of the comparison countrie s would probably have chosen to undergo austerity measures to maintain their pegs, they were never give n the chance (Worrell, 2011 ). Worrell's sentiment conflicts with the view that Barbados is unique in its strong support for the parity.
98 Slow reactions on the parts of the comparison countrie s are hypothesized to be partially due to a lack of cohesiveness in their societies. Trinidad and Tobago, the smallest of the countries, has a population that is four times the size of th at of Barbados. As the size of a country increases, its cohesiveness usually decreases. Furthermore, there was much less income inequality in Barbados than any of the comparison countrie s, which is another contributor to cohesiveness. This is particular ly apparent in Argentina. Finally, ethnic conflicts in Trinidad and Tobago and tribe based politics in Jamaica stand in stark contrast to the homogenous Barbadian social structure ( Blackman, 2003 p. 10). Another factor that contributed to the demise of the comparison country regimes (excluding Trinidad and Tobago) is that their governments had lost credibility because of previous exchange rate crises that resulted in devaluations or regime changes. 4.4 Opportunities for Future Research Opportunities for future research involve improving or extending the methodology and conducting a more rigorous assessment of the differences between Barbados and the comparison countrie s to determine if there is merit t o the conclusion that the island maintained exchange rate stability because of its uniqueness. 4.4 .1 Improvements and Extensions of the Methodology Indicators An obvious, somewhat difficult, way to improve the methodology is to include more indicators. Well selected indicators, such as debt variables, can be used to gain a more nuanced understanding of a country's experience leading up to a crisis. For variables available at the quarterly level, the approach used in this thesis could probably
99 be used. Variables only available at the annual level might have to be perman ently excluded or only qualitatively examined. Indicator Transformations One of the underlying assumptions of the signal extraction approach is that large movements in the 12 month perce ntage change in a variable can be indicative of a coming currency crisis. A researcher should also investigate incorporating a measure of slow, but adverse change in an indicator leading up to a crisis; this might be done by examining an increasing deviat ion from a long term moving average of the indicator. Comparison Countrie s and Crises The scope of this thesis did not allow for the examination of more than one crisis per country. With more time, multiple crises from each comparison country and from B arbados could be examined to identify patterns and if the conclusions from this thesis' analysis are robust to different crises. Furthermore, alternative comparison countries could be included in the study. Craigwell (2011 ) has suggested that Barbados' experience could also be compared to that of similar countries that have successfully defend ed their pegs, like Belize and T he Bahamas. By understanding the similarities in their experiences, a future researcher might be able to gain insight into why the y all have been able to maintain their exchange rate pegs. Thresholds A future researcher should consider how to correct for the impact of outliers on the variation and averages of the different indicators so that thresholds are not skewed. A method for controlling for the effects of multiple crises would also be helpful. Another
100 modification to the calculation of thresholds is to use rolling averages and standard deviations of the thresholds. This modification would hopefully correct for changes in th e economic structure of the countries under examination during the period this is being studied Data Range To make the comparison between Barb ados and the comparator s as fair as possible, it might be useful to only use data from when the comparator s were operating under fixed exchange rates. Otherwise, the variance of the indicators will probably be inflated due to higher variation during floating time periods. This higher variance will reduce the number of signals during the crisis period and unfai rly reduce the perceived severity of the crisis. Application of the Signal Extraction Approach An important aspect of future research is to determine the import ance of each included variable in precipitating each country's crisis. This knowledge could th en be used to weight each of the variables according to their influence on a country and thus come to a fuller understanding of the economic turmoil leading up to a country's exchange rate crisis. Also, instead of the subjective visual approach used in th is thesis, a more numerical method of analysis could be undertaken. This method would most likely look at producing a weighted average of each indicator's values above and below the threshold, weighted by how close the value's month was to the crisis mont h. These weighted averages could then be compared on a side by side basis, rather than relying of a human's visual interpretation.
101 The weighting of an indicator due to its importance, in conjunction with a numerical measure of severity, could be used to c onstruct a composite indicator for each country. This would be done by summing the numerical measures of each indicator for a country, weighted by their importance. Improvement of the Signal E xtraction Approach The use of the percentage above and below t he threshold, as opposed to a signal or no signal, as the value for each indicator for each month might enhance the signal extraction approach. This would enable that approach to glean more information from the data and would correct the all or nothing na ture of the signal / no signal aspect of the approach. 4. 4 .2 Further Investigation of Conjectures The main thrust o f this thesis was to compare Barbados' pre crisis experience with that of the three comparison countrie s. However, given that the hypothe sis was rejected, conjectures about why Barbados was able to maintain its fixed exchange rate were offered. Further research should focus on investigating these conjectures in more detail by examining the country specific factors of Barbados and how they interacted with the actions that the island took to avoid losing its peg. Then, a much deeper investigation of the comparison countrie s should be undertaken to determine what the truly unique factors of Barbados are and to judge whether or not these facto rs can explain why Barbados has been able to maintain its fixed exchange rate stability. 4.5 Conclusion This thesis set out to investigate whether Barbados' maintenance of its fixed exchange rate could be explained by a lack of economic turmoil leading up to its 1991
102 crisis, as compared to the pre crisis time periods of three comparators : Argentina, Jamaica, and Trinidad and Tobago. It was found that Barbados suffered more adverse behavior in its economic indicators than Argentina and Trinidad and Tobago, but less than Jamaica. This leads to the preliminary conclusion that Barbados' exchange rate stability cann ot be fully explained by a mild experience leading up to its 1991 crisis. However, the methodology used to arrive at this conclusion, a modification of the signal extraction approach from the currency crisis prediction literature, might not be completely valid. Thus, this thesis explored two different avenues to explain rejection of the research hypothesis First, issues with the methodology were investigated, including the applicability of the signal extraction approach, excluded indicators, and the ca lculation methods of the thresholds used. Potentially, these issues could be serious enough to invalidate the conclusion that Barbados had a more adverse experience leading up to its crisis. If so, then it would not be surprisi ng that Barbados did not lo se its peg as there was not enough pressure on the fixed exchange rate regime. However, if the methodology was sound enough and the conclusion warranted, unique characteristics of Barbados could explain the island's exchange rate stability. These unique characteristics and the actions that they enabled the government and pr ivate sector to take in the defense of the parity were also investigated. Characteristics include a high level of support for the parity with the U.S. dollar and a high level of social cohesiveness both of which enabled a real devaluation in lieu of a no minal devaluation Both of these avenues, issues with the methodology and the response of Barbados to its 1991 crisis, suggest rich areas of further research that can
103 improve the methodology and provide a more detailed analysis of the uniqueness of Barbados. Ultimately, this thesis makes two major contributions to the literature. First, it develops an adaptation of the signal extraction approach that draws more information from currency crisis indicators by measuring the distance ab ove and below a threshold value of the indicator. This adaptation can be inserted back into the currency prediction field and potential create more accurate predictions by drawing insights from how close to or far above an indicator s value is from a thre shold, as opposed to a binary signal or no signal approach. Second, this thesis presents a puzzle and makes a first atte mpt to answer it : w hy has Barbados been able to maintain its fixed exchange rate even though it experience d more economic turmoil leadi ng up to its 1991 exchange rate crisis than Argentina and Trinidad and Tobago ? The first attempt at an answer suggests that Barbados' unique characteristics of strong support for the exchange rate parity, social cohesiveness, draconian response, and the s peed of its response all worked together to enable the island to overcome its challenges. If further research supports the preliminary conclusion that Barbados policy reaction w as the differentiating factor that precluded a loss of its fixed exchange rat e, then Barbados can serve as a model for other countries confronting exchange rate crises P erhaps its lessons will help prevent the devastating effects of such crises in countries far beyond the easternmost island of the Lesser Antilles
104 Appendix The graphs that follow were used to analyze the experience of Barbados and the three comparison countries. This analysis is contained in Chapter 3.
112 Bibliogr a phy Agenor, P., Bhandari, J.S. & Flood, R.P. 1992, "Speculative Attacks and Models of Balance of Payments Crises", International Monetary Fund Staff Papers, vol. 39, no 2. Amo Yartley, C. 2011, Personal Communication Bridgetown, Bar ba d os. Arize, A.C., Osang, T. & Slottje D.J. 2000, "Exchange Rate Volatility and Foreign Trade: Evidence from Thirteen LDC's", Journal of Business and Economic Statistics, vol. 18, no. 1, pp. 10. Banerjee, A. 1992, "A Simple Model of Herd Behavior", Quarterly Journal of Economics, vol. 107, p p. 797. Berg, A. & Pattillo, C. 1999, "Are Currency Crises Predictable? A Test", International Monetary Fund Staff Papers, vol. 46, no. 2, pp. 1 07. Bikhchandani, S., Hirshleifer, D. & Welch, I. 1992, "A Theory of Fads, Fashion, Custom, and Cultural Change as Informational Cascades", Journal of Political E c onomy pp. 992. Bird, G.R. 2007, An Introduction to International Macroeconomics, 3rd edn, Palgrave Macmillan, New York Blackma n, C. 2003, The Barbados Experiment with the Productivity based Social Compact Economics and Politics Article edn, Productivity Seminar of the Central Bank of The Bahamas, Nassau, The Bahamas. Blanco, H. & Garber, P.M. 1986, "Recurrent Devaluation and Speculative Attacks on the Mexican Peso", Journal of Political Economy, vol. 94, no. 1, pp. 149. Breuer, J.B. 2004, "An Exegesis on Currency and Banking Crises", Journal of Economic Surveys, vol. 18, no. 3, pp. 2 93. Burkart, O. & Coudert, V. 2002, "Leadi ng Indicators of Currency Crises for Emerging Countries", Emerging Markets Review, vol. 3, pp. 1 07. Calvo, G.A. & Mendoza, E. 1996, "Mexico's Balance of Payments Crisis: A Chronicle of a Death Foretold", Journal of International Economi cs, vol. 41, no. 3/ 4, pp. 235. Craigwell, R. 2011, Personal Communication Bridgetown, Barba d os. Crowe, C. 2000, "Exchange Rate Crises and Capital Market Imperfections in Small Open Economies", Caribbean Development Bank Staff Working Papers, vol. Working Paper No. 2/00
113 Crowe, C. 1999, "A Theoretical and Empirical Analysis of the Barbados Balance of Payments Situation", Central Bank of Barbados Working P apers pp. 29. Dalrymple, K. 1995, "Balance of Payments Crisis and Policy Options: The Case of Barbados", Central B ank of Barbados Working P aper s Dornbusch, R. 1976, "Expectations and Exchange Rate Dynamics", Journal of Political Economy, vol. 84, no. 6, pp. 1161. Edison, H.J. 2003, "Do Indicators of Financial Crises Work? An Evaluation of an Early Warning System" International Journal of Finance and Economics, vol. 8, pp. 11. Flood, R.P. & Garber, P.M. 1984, "Collapsing Exchange Rate Regimes: Some Linear Examples", Journal of International Economics, vol. 17, pp. 1. Flood, R.P. & Marion, N.P. 1999, "Perspective s on the Recent Currency Crisis Literature", International Journal of Finance and Economics, vol. 4, pp. 1. Frankel, J.A. 2003 "Experience of and Lessons from Exchange Rate Regimes in Emerging Economies", National Bureau of Economic Research Working Pape rs, vol. Working Paper 10032. Frankel, J.A. 2003, "A Proposed Monetary Regime for Small Commodity Exporters: Peg the Export Price ("PEP")", International Finance, vol. 6, no. 1, pp. 61. Frankel, J.A. & Rose, A.K. 1996, "Currency Crashes in Emerging Marke ts: An Empirical Treatment", International Financial Discussion Paper No. 534, Board of Governors of the Federal Reserve S ystem Hallwood, C.P. & MacDonald, R. 2000, International Money and Finance, 3rd edn, Blackwell Publishing, United Kingdom. Haynes C. 1997, "Lessons from Barbados' Experience with the International Monetary Fund" in Central Banking in Barbados: Reflections and Challe n ges ed s H. Codrington, R. Craigwell & C. Haynes, 1st edn, Central Bank of Barbados, pp. 83. Isard, P. 1995, Exch ange Rate Economics, 1st edn, Cambridge University Press, Cambri d ge. Kaminsky, G., Lizondo, S. & Reinhart, C.M. 1998, "Leading Indicators of Currency Crises", International Monetary Fund Staff Papers, vol. 45, no 1. Kipici, A.N. & Kesriyeli, M. 1997, "T he Real Exchange Rate Definitions and Calculations", Central Bank of the Republic of Turkey Working P aper s
114 Krugman, P. 1979, "A Model of Balance of Payments Crises", Journal of Money, Credit and Banking, vol. 11, pp. 3 11. Krznar, I. 2004, "Currency Cr isis: Theory and Practice with Application to Croatia", Croatian National Bank Working Papers, vol. WP 12. Montiel, P.J. 2003, Macroeconomics in Emerging Markets, Cambridge University Press, United Kingdom. Morris, S. & Shin, H.S. 1995, "Informational Events that Trigger Currency Attacks", Federal Reserve Bank of Philadelphia Working Papers, vol. Working Paper No. 95 24. Obstfeld, M. 1986, "Rational and Self Fulfilling Balance of Payments Crises", The American Economic Review, vol. 76, no. 1, pp. 72. Obstfeld, M. & Rogoff, K. 1995, "The Mirage of Fixed Exchange Rates", National Bureau of Economic R esearch Working Paper Series Research and Economic Analysis Department n.d., "The International Monetary Fund Programme A Summary", Central Bank of Barbados Publication Research and Economic Analysis Department 1992, 1991 Annual Report of the Central Bank of Barbados Central Bank of Barbados, Bridgetown, Barba d os. Riechel, K. 1978, Economic Effects of Exchange Rate Changes, 1st edn, Lexington Books (D.C. Heath and Company), Lexington, Massachusetts. Rose, A K. & Yellen, J.L. 1989, "Is there a J curve?", Journal of Monetary Economics, vol. 24, pp. 53. Saxena, S.C. 2004, "The Changing Nature of Currency Crises", Journal of Economic Surveys, vol. 18, no. 3, pp. 321. Taylor, M.P. 1995, "The Economics of Exchange Rates", Journal of Economic Literature, vol. 33, pp. 13. Worrell, D. 2011, Personal Communication Bridgetown, Barbados. Worrell D., Codrington, H., Craigwell, R. & Greenidge, K. 2003, "Econ omic Resilience with an Exchange Rate Peg: The Barbados Experience, 1985 2000", International Monetary Fund Working P apers Worrell, D., Marshall, D. & Smith, N. 1998, "The Political Economy of Exchange Rate Policy in the Caribbean", Central Bank of Barbados Working P apers