See discussions, stats, and author profiles for this publication at: https://www.researchgate.net/publication/244959731 A New Breed of Exchange Rate Bands: Chile, Israel and Mexico Article in Economic Policy · October 1994 DOI: 10.2307/1344497 CITATIONS READS 55 26 5 authors, including: Elhanan Helpman Leonardo Leiderman 241 PUBLICATIONS 39,671 CITATIONS 88 PUBLICATIONS 2,686 CITATIONS Harvard University SEE PROFILE Tel Aviv University SEE PROFILE George Alogoskoufis Tufts University 71 PUBLICATIONS 1,282 CITATIONS SEE PROFILE All content following this page was uploaded by George Alogoskoufis on 10 July 2014. The user has requested enhancement of the downloaded file. All in-text references underlined in blue are added to the original document and are linked to publications on ResearchGate, letting you access and read them immediately. A New Breed of Exchange Rate Bands: Chile, Israel and Mexico Elhanan Helpman; Leonardo Leiderman; Gil Bufman; George Alogoskoufis; Vittorio Grilli Economic Policy, Vol. 9, No. 19. (Oct., 1994), pp. 259-306. Stable URL: http://links.jstor.org/sici?sici=0266-4658%28199410%299%3A19%3C259%3AANBOER%3E2.0.CO%3B2-S Economic Policy is currently published by Centre for Economic Policy Research, Center for Economic Studies, and Maison des Sciences de l'Homme. Your use of the JSTOR archive indicates your acceptance of JSTOR's Terms and Conditions of Use, available at http://www.jstor.org/about/terms.html. JSTOR's Terms and Conditions of Use provides, in part, that unless you have obtained prior permission, you may not download an entire issue of a journal or multiple copies of articles, and you may use content in the JSTOR archive only for your personal, non-commercial use. Please contact the publisher regarding any further use of this work. Publisher contact information may be obtained at http://www.jstor.org/journals/cepr.html. Each copy of any part of a JSTOR transmission must contain the same copyright notice that appears on the screen or printed page of such transmission. The JSTOR Archive is a trusted digital repository providing for long-term preservation and access to leading academic journals and scholarly literature from around the world. The Archive is supported by libraries, scholarly societies, publishers, and foundations. It is an initiative of JSTOR, a not-for-profit organization with a mission to help the scholarly community take advantage of advances in technology. For more information regarding JSTOR, please contact support@jstor.org. http://www.jstor.org Thu Oct 11 10:26:09 2007 - A new breed of exchange rate bands: Chile, Israel and Mexico Elhanan Helpman, Leonardo Leiderman and Gil Bufman Tel Aviv University, Canadian Institute for Advanced Research and CEPR, Tel Aviv University and CEPR, and Bank of Israel 1. Introduction A common feature of the aftermath of several important heterodox stabilizations was the gradual shift towards a regime of increased flexibility in nominal exchange rates. While the first phase of heterodox stabilization typically relies on the role of a fixed exchange rate in breaking inertia and producing rapid disinflation, it is commonly followed by a loss of competitiveness as reflected in cumulative real exchange rate appreciation. Experience indicates that sooner or later policy-makers begin perceiving this erosion in competitiveness as a serious risk to the ultimate success of the whole programme. When this happens, and once it has become clear that fiscal and monetary fundamentals have been adjusted and major disinflation is being achieved, countries tend to enter a second phase in which they relax the fixity of the exchange rate. It is in this spirit that Chile, Israel and Mexico adopted exchange rate bands a few years after the inception of exchange rate-based disinflations. In doing so, these countries have not given up the role of the exchange rate as an important nominal anchor. Instead, bands are viewed as introducing enough flexibility for the nominal exchange rate to respond to external and internal shocks without reneging on the medium and long-term signals provided by the policy embodied in the setting of the This paper is a substantial revision of the version presented at the Economic Policy Panel. We benefited from insightful and useful comments on earlier drafts and suggestions by Vittorio Grilli and Charles Wyplosz. In addition, we are grateful to George Alogoskoufis and Economic Policy Panel participants for their comments. For data and insightful discussions on Chile and Mexico we are grateful to Agustin Carstens, Daniel Oks, Moises Schwartz, Rodrigo Vergara and numerous individuals at the Banco de Mexico. Economic Policy October 1994 Printed in Great Britain New exchange rate bands Elhanan Helpman, Leonardo Leiderman and Gil Bufman - Summary Follounng injhztzon stabilzzation programmes where a fixed exchange rate was used as an anchor, a number of countries have gradually shifted towards a regime of increased jlexibility. In particular, Chile, Israel and Mexico have for some years adopted a regime of crawling exchange rate bands. This article documents and analyses their unique experience. Of immediate concern is the question whether the shzftji-om fixed to crawling bands is perceived as the signal of a tilt towards policy relaxation. On the contra?, it could be seen as a way of restoring external competitiveness after a period of real appreciation. The evidence collected in this article supports the latter view. As a way of moving towards a sustainable equilibrium, the shift to crawling bands is found to have had a stabilizing influence on financial markets, on inflation expectations, and on the passthrough from exchange rates to przces. New exchange rate bands 261 band's central parity. Furthermore, the shift from a fixed exchange rate to an exchange rate band opens room for greater autonomy of domestic monetary policy. Adopting an exchange rate band forces the authorities to take a stand on key operational issues. First, the central parity can be fixed or it can crawl. In countries with rates of inflation above those of their trade partners (e.g. Chile, Israel and Mexico), a permanently fixed central parity is not feasible. Therefore for such countries the real choice is between infrequent realignments (as in the EMS) and a crawl. As evident in recent European developments, infrequent realignments may generate substantial uncertainty, produce interest rate volatility, and damage the sustainability of the band. On the other hand, a major difficulty with crawling bands is that the frequent realignments of the central exchange rate may elicit inflationary expectations precisely as a serious attempt at disinflation is taking place. Second, if a crawling central parity is adopted the authorities have to choose a criterion for determining the crawl's pace. It can be based on a nominal objective (e.g. an inflation target) or on a real target (e.g. the real exchange rate). Third, it is necessary to specify the width for the exchange rate band. This involves an assessment of the variance of potential shocks as well as the degree to which the authorities can commit to narrow limits of exchange rate flexibility. Fourth, the frequency and size of intramarginal interventions has to be determined. For example, the authorities may choose to react to a drop in the profitability of exports, or to a downturn in economic activity, by accelerating the rate of depreciation within the band. Similarly, they may slow down the rate of depreciation if and when the rate of inflation is expected to increase. The exchange rate bands adopted by previously high-inflation countries are substantially different from the bands observed in Europe. In addition to their market-clearing roles, nominal exchange rates are used in these countries as key anchors of inflation expectations. Typically, these bands feature a crawling central parity, a unilateral commitment by the country's authorities to intervene to support the band, and a much larger width than in Europe before the recent crisis. For example, Chile's current band features a crawling central parity and fluctuation limits of 10% around this rate. At the beginning of each month the authorities announce the rate of crawl of the central parity for that month based on the difference between the rate of inflation in the previous month and a forecast of foreign inflation. In Israel, the authorities announce their inflation target for a given year, and then pre-announce a daily nominal depreciation for that year mostly based on the difference between the inflation target and a forecast of external inflation over the same time period. At the present time, the exchange rate is allowed to fluctuate 262 Elhanan Helpman, Leonardo L e i d m a n and Gzl Bufman within 5% of the central parity. Mexico's band is quite unique: it features a gradual crawl of the band's upper limit while the lower limit is maintained fixed. Consequently, the band's width rises over time, and it is expected to reach 7% by the end of 1994. This kind of exchange rate bands is appealing to countries with persistent inflation differentials. In fact, bands could be an appealing exchange rate system for other economies in Latin America and for several former centrally-planned economies in Eastern Europe. Thus in February 1994 Colombia has replaced its crawling-peg arrangement with a crawling band. Most of the available evidence about exchange rate bands has been based on the experience of Western European economies that operated their bands under a multilateral commitment to maintain and defend them. In contrast, this paper documents and analyses the experience with exchange rate bands in Chile, Israel and Mexico. In particular, it will be useful to rely on Israel's experience to assess whether a shift from fixed- to crawling-central-parity bands was associated with a stabilizing or destabilizing influence on financial markets, on inflationary expectations, and on the passthrough from exchange rates to prices. We review in the next section the main theoretical considerations that guided our empirical work. Based on that discussion, we elaborate in Section 3 on the case of Israel. Section 4 employs various statistical indicators to compare behaviour under fixed- vs. crawling-bands. Evidence for Chile and Mexico is provided in Section 5. Last, Section 6 draws the main lessons from the comparison of the three countries and discusses some of the practical policy implications of the analysis. 2. Theoretical considerations 2.1 Implications of target zones Recent analyses of target zones provide a useful benchmark for the empirical analysis of exchange rate fluctuations within the announced exchange rate bands. Krugman (1991) has shown how the mere existence of set limits on exchange rate movements affects the behaviour of the rate within its band. Under a number of conditions, to be criticised shortly, Krugman found that the relation between the exchange rate and its fundamentals - assumed to be fully described by one composite indicator - exhibits an S shape, with tangency at the band'shmits (Svensson, 1992, offers a useful survey of the literature). This finding has several important implications. First, the exchange rate should spend most of its time close to the edges of the band (its distribution is bimodal and U-shaped). Second, the expected rate of depreciation is highest at the band's lower New exchange rate bands 263 bound (when the exchange rate is strongest) and declines as the exchange rate depreciates, rising towards the band's upper limit (the distribution exhibits mean-reversion). Third, this means that the exchange rate is expected to depreciate in the lower part of the band: uncovered interest parity then implies that the differential between the domestic and foreign interest rate is highest when the exchange rate is at the band's lower limit and declines as the exchange rate rises toward the upper limit. Fourth, both the exchange rate and the interest rate differential are expected to exhibit more variability in the middle of the band than at its extremities where interventions are expected to stabilize the exchange rate (see Svensson, 1992, and references therein for more detail on these results). These results crucially depend on two unsatisfactory assumptions: (i) that the market fully believes that the monetary authorities can and will defend the bands so that the exchange rate will never break through its set limits; (ii) that the central banks only intervene at the margin. When more realistic forms of intervention are allowed, Krugman's results are radically modified. For example with 'leaning against the wind' interventions, the exchange rate is now expected to spend most of its time near the centre of the band (its distribution now exhibits an inverted-U shape, as shown by Lewis, 1991). Thus the behaviour of the exchange rate within the band may be strongly affected by the specific intervention method that maintains the target zone. Regardless of the specifics of interventions, as long as the bands are credible both the interest rate differential and expected depreciation rise as the exchange rate moves toward the most appreciated value within the band because of the probability that it will soon decline. This result is overturned in the presence of a realignment risk, when the exchange rate band is not fully credible. For example, consider the case when markets believe that the authorities may well devalue rather than defend the central parity when the exchange rate approaches the ceiling of the band. Bertola and Caballero (1992) have shown that the relation between the exchange rate and its fundamentals then assumes an invertedS shape. This means that the exchange rate spends most of its time towards the middle of the band, and that the interest differential now increases when the exchange rate moves toward its ceiling because the perceived probability of a depreciation rises, instead of declining as it does when the target zone is believed to be cast in iron. The realistic possibility that the band will not be defended leads us to express the expected rate of depreciation as the sum of two components: the expected rate of realignment - the change of the central parity - and the expected rate of currency depreciation within the band - movements within the band. Each of these components has a different implication for 264 Elhanan Helpman, Leonardo La'derman and Gil Bufman the relation between the interest rate differential and the position of the exchange rate within the band: a devaluation risk implies that the interest differential rises as the exchange rate depreciates within the band while credible bands means that the interest differential declines. This is why Svensson (1992) has argued that the interest rate differential alone could be a misleading indicator of realignment expectations.' 2.2 Imperfect capital mobility Key aspects of the preceding analysis (and in what follows), such as the relation between the interest rate differential and the exchange rate, rely on the assumption of uncovered interest rate parity and a time-invariant foreign exchange risk premium. Although for most recent periods this assumption is a reasonably good approximation in the cases of Chile, Israel and Mexico, there have been episodes that could be interpreted as deviations from this parity relation. The reason is that capital mobility has been less-than-perfect, a consequence of the non-negligible information and transaction costs as well as of the remaining regulations that characterise financial markets in these countries. For example, it has not been uncommon to observe episodes in which, in an attempt to stimulate employment in the short run, an expansionary monetary policy resulted in lower domestic nominal interest rates while at the same time foreign exchange market intervention was performed to avoid a depreciation of the exchange rate. The less-than-perfect degree of capital mobility in the countries under study, which increases the short-run efficacy of monetary and exchange rate policy, severs the predicted links between interest rates and exchange rates. This effect, however, is temporary because such actions typically give rise to increased capital outflows and to a reduction in the stock of international reserves at the domestic central bank. In this context an examination of the behaviour of capital flows and central bank international reserves can provide useful information on the sustainability and credibility of an existing exchange rate band. In particular, it can be argued that those configurations of interest rates and exchange rates that result in sizeable and persistent capital flows in only one direction are likely to weaken the sustainability of the band and to give rise to pressure for change either in the band's parameters or in monetary policy. u 1 As shown by Helpman et aL (1993), most of these theoretical implications are also valid for cases in which continuous preannounced realignments of the central parity are part of the existing band regime, as in Chile, Israel and Mexico. In this case, it is important to redefine variables so that now exchange rates are expressed in terms of their deviation from the central parity, and fundamentals in terms of their deviation from their underlying trend. New exchange rak bands 265 2.3 Policy trade-offs: inflation and the real exchange rate The adoption of bands in Chile, Israel, and Mexico was perceived as a way to resolve the fundamental short-run policy trade-off between the level of the real exchange rate and the level and variability of the nominal exchange rate (or rate of inflation) .2 On the one hand, the authorities in these countries have used exchange rate realignments to compensate for persistent inflation differentials, and have expressed their concern to preserve and improve the profitability of exports and the current account position by having a competitive real exchange rate. On the other hand, it has been recognised that frequent exchange rate deprecations could weaken the role of the exchange rate as a nominal anchor and could elicit inflationary expectations. In fact, other things equal, an accommodative exchange rate policy may have a destabilizing impact on the rate of inflation and other nominal variables in economies with a history of pronounced inflation inertia and with widespread indexation of nominal wages. (See, e.g., Dornbusch, 1982; Bruno, 1993; and Leiderman, 1993). The crawling exchange rate bands regime is a simple and verifiable system for the policy-maker to make a clear commitment about a key nominal anchor such as the path of the central nominal exchange rate, while allowing for some degree of exchange rate flexibility needed to shield exports and the current account from the impact of adverse shocks. As shown by Cukierman et aL(1994), under such a regime, other things equal, expectations of realignments increase as the exchange rate moves toward the upper limit of the band. The reason is that the policy-maker's choice is governed by the trade-off between credibility and flexibility: when the exchange rate has been for a while at its limit under the pressure of a high inflation rate, the cost of maintaining credibility with an increasingly overvalued parity is bound to become excessive, an assessment well understood by the markets. The upshot is that the behaviour of real (and not only nominal) exchange rates and of the rate of inflation can play an important role in determining the sustainability of an existing band. Thus if, for example, the parameters chosen for a given band and the underlying shocks result in an acceleration of inflation and an overvaluation of the domestic currency, pressures for a realignment and for a change in the band's parameters are likely to rise. These features which may well seriously 2 This is borne out, for example, by the statement of the president of the Banco Central de Chile in Zahler (1992). That real exchange rate behaviour is an important consideration in assessing the performance of bands in small industrial countries is transparent in the various papers collected in Argy and De Grauwe (1990). 266 Elhanan Helpman, Leonardo La'deman and Gil Bufman damage the credibility of an existing band are not revealed in an analysis that focuses on nominal exchange and interest rates alone. 3. Israel Israel's exchange rate policy has followed three main phases after the stabilization programme of July 1985. First, the NIS (New Israeli Shekel) was pegged to the US dollar. This fixed exchange rate policy was a major building block of the disinflation programme. In August 1986 the dollar peg was replaced by a peg to a basket of currencies. Second, following a sequence of devaluations of the NIS/Basket exchange rate in 1987, 1988 and early January 1989, the government adopted an exchange rate band on January 3, 1989. The band consisted of a fured NIS/Basket central parity with a 3% fluctuation zone around this parity. The band's width was enlarged to 5% in March 1990. Third, in December 1991, after an upward adjustment of the central parity, the authorities relaxed the fixity of the central parity and announced an upward crawl of the exchange rate band. In addition, an official inflation target was announced for the first time. The rate of crawl (or depreciation) of the central parity rate was set at 9% per year, to reflect the difference between Israel's inflation target (14.5%) and a forecast of foreign inflation for 1992. The announced rate of crawl for the central parity was reduced to an annual rate of 8% per year starting from November 1992 and to 6% per year starting from July 1993, and the announced inflation targets for 1993 and 1994 were 10 and 8%, respectively. We begin this section by examining the behaviour of exchange and interest rates under the various bands. In particular, we seek to determine whether the evidence conforms with the implications of the analysis presented in the previous section. We then turn to movements in foreign exchange reserves, a measure of intervention by the authorities. Lastly, we present econometric evidence on realignment expectations and a discussion of real exchange rate and inflation developments, both of which are relevant for assessing the sustainability of a given set of band policy parameters. 3.1 Exchange rates Figure 1 shows the daily NIS/Basket exchange rate from mid-1985 to the end of 1993 and Table 1 provides basic data on the various bands. Each exchange rate band is identified in Figure 1 by three lines: its central parity and its upper and lower limits. A number of features stand out: (i) the nominal exchange rate followed an upward trend, a combination of depreciation within bands and adjustments of the central parity; (ii) the New exchange rate bands 3.5 .- ISRAEL +a 5 3.0 + a, 5 2.5 m I) $ 2.0 a $ 1.5 500r CHILE r MEXICO 3,300 z 1 31250 $3,200 = 1 Pacto renewal 0ct.-20-92 F i e 1. Exchange rate bands (daily data) Notes: Israel: For the period July 1985-July 1986 the exchange rate shown is that of the NIS against the U S dollar. From August 1986 onwards the exchange rate shown is that of the NIS against the Israeli currency basket. Chile: For the period January 1986June 1992 the exchange rate shown is that of the Peso against the U S dollar. From July 1992 onwards the exchange rate shown is that of the Peso against the Chilean currency basket. Date of last observation: Israel-28 January 1994; Chile-28 June 1993; Mexico-30 June 1993. central parity was realigned no less than seven times in five years; and (iii) some upward adjustments of the central parity took place when the exchange rate was not close to its upper limit (March 1990 and 1991). The exchange rate spent too much time around the middle of the bands and too little time at their upper parts to fit the simple target zone interpretation (this visual impression is formally confirmed by statistical tests in Helpman et al. 1993). This suggests either that the authorities have been conducting mean-reverting interventions inside the bands and/or that the bands were not particularly credible. Furthermore, also in contrast to the predictions from the basic target zone interpretation, the exchange rate was on average more volatile near the limits of the band than near the central parity and no significant differences were found in the average rate of change of the exchange rate across the various regions of the band. Table 1. Israel: band characteristics, exchange rate, interest rates, prices and foreign currency reserves N Q, 00 Crawling Bands BAND 1 BAND 2 BAND 3 BAND 4 BAND 5 BAND6 BAND7 BAND8 BAND6,7,8 Period Band charactnistics and thz exchangz rate; with daily data: Mid-band rate (NIS/basket) Band width +/- (in %) Crawl of mid-band rate, annualized (in %) Average exchange rate Average deviation of exchange rate from mid-band (in %) Standard deviation from mid-band (in %) Intart rak-s using weekly data, in percent per numth tnnrc: Average domestic interest rate (monetary auction rate) Standard deviation of domestic interest rate Average foreign interest rate (LIBID) (1) Average interest differential (Israel-Foreign) Annualized average interest dierential Standard deviation of interest rate dierential h 1.95 3 2.07 3 2.19 5 2.41 5 2.55 5 - - - - 5 5 5 5 - - - - - 1.92 2.07 2.20 2.29 2.52 9 2.69 8 2.99 6 3.14 8 2.89 0.19 0.34 -1.41 -4.79 -1.15 -1.45 -0.41 -1.79 -1.15 s R i33 % & z -3 1.06 0.69 2.29 0.32 2.31 1.71 1.24 0.54 1.48 6 8 $ F 3 R 9 h ki 4a 3 Elhanan Helpman, Leonardo Leidman and Gil Bufman 270 3.2 Interest rates The shortest reliable free market interest rate available in Israel is determined in monetary auctions, where the Bank of Israel auctions off credit funds to financial institutions. The auctions took place once a week in the past, and were expanded to daily auctions at the end of 1990. For this reason we now work with weekly observations. Table 1 provides some basic interest rate data and Figure 2 presents a plot of the interest rate differential for the various bands. The figure exhibits two important characteristics: (i) most of the time the interest rate differential follows a cyclical pattern; and (ii) most of these cycles are associated with shifts across bands: the interest rate differential generally rises before, ISRAEL 89.32 90.42 91.52 Week number 93.09 CHILE I MEXICO I I Jan-86Jun-87 Nov-88Mar-90Sep-91Feb-93 Sep-86 Feb-88 Jul-89 Dec-90 May-92 Dally data Week number Figure 2. Interest rate differentials (annualized terms) Notes. Israel: The differential shown is calculated using the Bank of Israel's monetary auction rate and a currency basket weighted composite of foreign interest rates. Chile: For the period January 1986-June 1982 the differential is calculated using short-term Peso deposit rates and US. T-bill rates with similar maturity periods. From July 1992 onwards the differential shown is calculated using a currency basket weighted composite of foreign interest rates. Mexico: The differential shown is calculated using 28 day CETES rates and US T-bill rates for similar maturity periods. Date of last observation: Israel-Week ending 31 December 1993; Chile-28 June 1993; Mexico-Week ending 17 July 1993. 271 New exchange rate bands and declines after a realignment. The evidence is compatible with the notion that the market's anticipation of realignments (i.e., devaluations) played an important role in the observed volatility of interest rates. To check the prediction of a negative relation between the interest rate differential and the position of the exchange rate within the band we turn to Figure 3. If anything, there is evidence of a weak positive association between the interest rate differential and the deviation of the exchange Deviationof E.R. from midband (%) Deviation of E.R. from midband (%) Israel: bands 6-8 (crawling bands), correlation coefficient = 0.012 Israel: bands 1-5, correlation coefficient = 0.304 75 -g-.- - 50- 17.5 i4\' + m .. C 6 E v 25- w 4- f? C ln 0- .. + -1.0 -0.5 B e 6 C C -25 -15 -10 -5 -0 5 10 Deviat~onof E.R. from midband (%) Chile: 1986-1993.06, correlation coefficient=0.078 5.0 -1.5 -0.0 0.5 1.0 1.5 Dev~ationof E.R. from midband (%) 1-1993.06, correlation Mexico: 1991.I coefficient= 0.092 Figure 3. Interest rate differentials and the deviation o f the exchange rate f r o m mid-band Notes: The interest rate differentials are shown in annualized terms. For Mexico the midband rate is calculated as the average of the upper and lower bounds of the band. For further information regarding the data shown here, see notes to Figures 1 and 2. 272 Elhanan Helpman, Leonardo Leiderman and Gil Bufman rate from the central parity. Formal statistical analysis of movements in the interest rate differential within bands (Helpman et al., 1993~)further reveals that the distribution of the interest rate differential across various band regions does not conform with the basic target-zone pattern. This could well reflect the impact of monetary and exchange rate policy under imperfect capital mobility, along the lines of the discussion in Section 2.2. For example, the combination of a low domestic interest rate (aimed at stimulating employment) and increased expectations of devaluation gave rise to a sizeable outflow of capital in late 1992 which was associated with a loss of more than $1 billion of official reserves. 3.3 Foreign exchange reserves Figure 4 provides a plot of the monthly net sales of foreign currency by the Bank of Israel to the public. In Israel, the public sector obtains substantial resources in foreign currency, especially in the form of foreign aid, and the central bank ends up on average selling foreign exchange. The evidence in Figure 4 indicates that most realignments were preceded by high and generally rising net sales of foreign currency by the central bank, and were followed by foreign-currency central bank purchases from the public. The notion that in an economy such as Israel movements in foreign currency sales can serve as a useful indicator of a band's credibility and sustainability - beyond other indicators such as the exchange rate and interest rate differential - can be illustrated with developments in 1992. During the second half of 1992 the authorities kept a relatively low domestic interest rate with the aim of accommodating an apparent slowdown in inflation and to possibly stimulate economic activity in the short run. At the same time, institutional measures were taken to further liberalize the capital account (for example final clarification of the status of taxation of income from mutual funds invested abroad). Together with rising expectations of a stronger US dollar these developments were associated with a net capital outflow of about $1 billion over a short time period. While these circumstances gave rise to considerable doubts about the sustainability of the band's parameters at that time, they did not affect the interest and exchange rates because of heavy intervention by the authorities. 3 Most of the evidence is qualitatively similar to that discussed by Svensson (1992) for the EMS. New exchange rate bands 273 F i e 4. Purchases of foreign currency. Purchases (-)/sales (+) by the public Note The series shown is the net contribution of Israel's private sector to changes in Israel's foreign currency reserves. Source: The Bank of Israel. 3.4 The real exchange rate and inflation The lack of a stable competitive real exchange rate during the second half of the 1980s, shown in Figure 5 (monthly data), was perceived as a major problem in an open economy such as ~ s r a e lBeing .~ an exchange ratebased stabilization, the 1985 programme and the policies that followed its implementation did not prevent a severe real appreciation. Rising concerns by policy-makers about the loss of external competitiveness prompted the devaluations of December 1988-January 1989 and the adoption of an exchange rate band. Stability of the real exchange rate was attained under the first band regime (with fixed central parity), albeit at an overvalued level, without an acceleration of the rate of inflation (see Figures 3 and 5). At the end of 1990 the real exchange rate appreciation was of about 15% as compared to the start of 1986. However, a partial reversal took place under the crawling band in 1992, with a 9% real exchange rate depreciation 4 The real exchange rate was calculated as the product of the nominal basket exchange rate times a (basket-weighted) foreign price level divided by the domestic price level. Elhanan Helpman, Leonardo Leiderman and Gil Bufman ISRAEL -- inn, . 8::i CHILE , . II ding band Isand Crav lannounced MEXICO 260r Figure 5. Real exchange rate Note: Israel: The real exchange rate shown is computed using Israeli wholesale prices, a currency basket-weighted composite of foreign wholesale prices and the currency basket exchange rate. Mexico: The real exchange rate shown is computed using Mexican wholesale prices, U S wholesale prices and the exchange rate of the Peso against the U S dollar. Chile: Real effective exchange rate. Source: IMF, Zntaational Financial Statistics. (measured in terms of the basket, from December 1991 to December 1992). In spite of early fears that this real exchange rate depreciation, and more generally the shift to a crawling band, would be associated with a rise in the rate of inflation, the opposite occurred. The rate of inflation reached a singledigit figure in 1992 (9.4%) for the first time in more than two decades. However, this did not eliminate the concern about potential inflationary consequences of exchange rate fluctuations, especially in light of the deviation of actual from target inflation observed in 1993. All in all we conclude that there were no special pressures from the rate of inflation and real exchange rate developments on the sustainability of the exchange rate bands. On the contrary, the recent reduction in the rate of inflation and real exchange rate depreciation have probably contributed to strengthen this regime. 275 New exchange rate bands 3.5 Estimated realignment expectations The sustainability of an existing band and realignment expectations are not directly observable. Yet, as noted above, the interest rate differential reflects expectations of (i) expected realignments and (ii) expected movement of the exchange rate within the band. In this section we use a procedure developed by Svensson (1992) (known as the drift-adjustment method) which permits to unscramble these two components. Appendix A explains how we extend the procedure to crawling bands. The pattern of the expected realignments thus computed, shown in Figure 6, is plausible. During the first band there was a gradual decrease in devaluation expectations, only partially reversed a few weeks before the end of the band. In the second band there was a gradual upward trend until the next realignment. The third band featured an initial decrease in ISRAEL 89.42 91.20 92.50 Week number CHILE 0.25~ MEXICO I 0.45 Y $ 0.20 3 0.15 A 0 0 & 0.05 ,,,mr_mn, 0.00 0ct-86 May-88 Jan-90 Aug-91 Mar-93 Daily data 92.07 ","....,.."."."~ 92.33 93.06 Week number -Realignment -Differential Figure 6. Expected realignment and the interest rate differential Note The expected realignment series shown are fitted values of equation (3) in Appendix A. 276 Elhanan Helpman, Leonardo Leiderman and Gil Bufman realignment expectations followed by a sharp rise. An upward realignment was expected throughout the fourth band. The shift to the first crawling band (i.e., the sixth band) was preceded by a sharp increase in realignment expectations. Yet, no such expectations were associated with the shift to the second and third crawling bands. Overall, Figure 6 confirms Svensson's (1992) point that movements in the interest rate differential could be misleading indicators of realignment expectations. As with European countries, fluctuations in expected realignments are much more volatile, and fit more closely realignment dates, than movements in the interest rate differential. We have also explored whether some macroeconomic indicators affect expected realignments. In particular, we were interested in determining whether the trade-off between credibility and competitiveness could be traced back in the data. The reasoning in Section 2 has led us to focus on the rate of change of the real exchange rate, on the rate of change in economic activity, and on the monthly change in foreign currency reserves. There is plausible evidence (at the 10% significance level) that past increases in economic activity, foreign exchange reserves, and the real exchange rate are associated with a significant decrease in expected realignmenk5 4. From fixed to crawling bands: a regime change? Among the three countries discussed in this study, Israel is unique in featuring an important change in the functioning of the exchange rate band, from a fixed to a crawling central exchange rate. There was, and still is, considerable controversy about the possible effects of this shift. On the one hand, by allowing the parity to change gradually over time - in a series of small preannounced daily steps - rather than in the sudden discrete changes that were characteristic of the earlier bands, the new system could have a stabilizing influence on financial markets as well as on the real exchange rate. On the other hand, these frequent realignments of the central exchange rate could weaken the role of the exchange rate as a nominal anchor and elicit inflationary expectations. Put differently, a shift toward a more accommodative exchange rate policy could, in principle, have a destabilizing impact on the rate of inflation and other nominal variables, especially in an economy with a history of pronounced inflation inertia and widespread wage - and financial - indexation. As a matter of fact, part of the aversion to exchange rate flexibility was rooted 5 However, it would be hard to consider the estimates as providing a strong case for using these macroeconomic variables in expected-realignment equations. New exchange rate bands 277 in Israel's experience with a crawling peg in the mid-1970s which was associated with a rise in the level and volatility of the rate of i n f l a t i ~ n . ~ While it is too early now to settle (ex-post) this controversy, our purpose here is to rely on various statistical indicators to provide an empirical comparison of the bands with fixed and crawling central parity and to draw some preliminary conclusions. The section proceeds as follows. First, we discuss whether there has been a change in the volatility of nominal and real exchange rates and of interest rate differentials. Then, evidence is provided on shifts in realignment expectations, and on other indicators of band credibility such as the parallel market exchange rate premium, following the policy shift. Last, we examine the impacts of this shift on inflation expectations and on the passthrough from exchange rates to prices. Figure 2 and Table 1 show that there has been a reduction in the volatility of the interest rate differential under the crawling bands. (The unconditional standard deviation of the differential was on average less than half the standard deviation before the shift to a crawl, a change that is statistically significant.) This visual impression is formally confirmed by various statistical tests presented in Appendix B. In regards to the nominal exchange rate, Figure 1 suggests that while under the earlier bands the exchange rate quite often remained distant from the central parity, it has recently not moved far from the crawling path of the central exchange rate, thus suggesting that there has been a reduction in the volatility of deviations of actual from central exchange rates (see also Table 1). This is confirmed by formal statistical tests presented in Appendix B. Another important aspect of the comparison across bands is the behaviour of realignment (or devaluation) expectations: the stronger are these expectations, the greater is the probability of collapse in the exchange rate regime. The weekly expected realignment series previously estimated and plotted in Figure 6 suggest that the move to crawling bands has been accompanied by a significant reduction in both the level of expected realignments and their variability (measured by the conditional standard deviation). Again, the formal tests presented in Appendix B confirm these impressions. Another way of making the point is by looking at the percentage of days for which Figure 6 indicates that markets may have been expecting a devaluation. Table 2 shows that the this proportion has decreased in the crawling band period relative to the earlier bands. 6 The acceleration in the rate of inflation was related to the accommodation policies, and indexation effects, that followed the oil shock of 1973-74 as well as the relatively high government budget deficits that prevailed at that time. For details, see Bruno (1993) and Leiderman (1993). 278 Elhanan Helpman, LRonardo Leiderman and Gzl Bufman Table 2. Israel: band non-credibility index Band number Non-Credibility index (%) 1 17 18 38 22 54 7 2 3 4 5 6 7 8 19 0 No& This index shows the percentage of days for which there were significantly positive devaluation expectations of the exchange raie from the maximal feasible rate under the current regime (0% = full credibility, 100% = full non-credibility) . Still further evidence pointing in the same direction is provided by two market-based indicators. First the premium on the parallel (black) market exchange rate is plotted in Figure 7. Second, the Bank of Israel sells, via auctions, options on the exchange rate of the local currency against the US dollar. Figure 8 shows the premium for three-month 'at the money' (i.e., European) options. The premium is shown as a percentage of the exchange rate prevailing on the day of the auction of the options. Increased premiums on these three-month options are ordinarily interpreted as signalling an increase in the expected volatility of the exchange rate. The reversal in the trend of these premiums -from positive to negative - following the shift to crawling bands shown in Figure 8 suggests a decrease in volatility. Did the shift rekindle inflationary expectations as often feared? A negative answer can be given once we derive inflation expectations from a comparison of yields on nominal and CPI-indexed government bonds of similar maturity (see Yariv, 1989). Figure 9 depicts the term structure of these inflation expectations one month before and after the announcement of the shift to a crawling band in December 1991. As indicated above, this announcement of the policy shift, and of the new band's parameters, was made jointly with the announcement of an official inflation target for 1992. It can be seen in Figure 9 that there has been a downward shift of the whole term structure of inflation expectations from about 19% prior to the announcement to about 14% afterwards. Finally, we ask whether the shift to a crawling band was associated with a marked change in the transmission of exchange rate (or foreign price) impulses into domestic prices of tradable goods. The results of our N m exchange rak bands Figure 7. Black market premium on the U S dollar (black market rate over official rate) Notes: The black market premium is the difference between the black market exchange rate and the official exchange rate as a percent of the official exchange rate. The currency option premium is shown as a percent of the exchange rate known on the day of the option auction (the effective premium). The premium shown is that of Smonth 'at the money' options written and auctioned by the Bank of Israel. 88.07 89.07 90.07 91.07 92.07 93.07 F i e 8. Currency option premium (premium as a percent of exchange rate) No& As for Figure 7. Elhanan Helpman, Leonardo Leiderman and Gzl Bufman 280 -5 207g.-- 19. 1 month before announcement ! 18C m I 17- S .z 16- m 'F Announced target W 13 12 - I 3 6 9 12 Months ahead -91.11 +92.01 - 14% target Dec 1991: Announcement of 9 % crawl and inflation target of 14%-15% Figure 9. Response of inflationary expectations to crawl rate and inflation target announcement Note: Inflationary expectations are market-based and are calculated using the difference between the yield to maturity of non-indexed government T-bills and CPI indexed government bonds with similar maturity periods. For further details see Yariv (1989). Source: The Bank of Israel. estimates are presented in Figure 10 which depicts the simulated impact of a one-time exchange rate 1% depreciation on the domestic price of traded goods. The effect is clearly much weaker in the recent crawling band period, in contradiction with the view that the shift to increased exchange rate flexibility has been followed by a rise in the inflationary impact of exchange rate depreciation.' To recapitulate, we find that, contrary to fears from the past, enhanced flexibility of the band's central exchange rate did not come at the cost of accelerating inflation. The shift from fixed- to crawling-central-parity bands was accompanied by a reduction in the volatility of exchange rates and interest rate differentials, a decrease in the relative size and 7 We conducted bivariate vector autoregressions for the periods 1989-91 and 1991-93. The systems included monthly data for the rate of change of traded-goods prices and for the sum of the rate of change of the exchange rate and foreign prices (measured by a trade-weighted foreign WPI). The estimated parameters and standard errors on dummy variables for the crawling bands period indicate that there has been a statistically significant reduction in the degree of passthrough in the second subperiod compared to the earlier period. Similar evidence was obtained from estimating direct passthrough equations. N m exchange rate bands -0.2 i ~mh~sei - 89-91 3 i - 6 4 j Months from impulse 91-93 6 10 12 11 +Exchange rate F i e 10. Impulse response of traded prices to a 1% shock to the exchange rate Note: Simulation based on bivariate VAR of traded goods price changes and basket exchange rate foreign WPI changes. The simulation shows the response of traded goods prices to a 1% shock of the exchange rate. + Table 3. Israel: macroeconomic indicators - Domestic inflation (CPI) (%) Foreign inflation (basket-weighted WPI) (%) Domestic budget deficit (as a % of GDP) Money growth (%) Unemployment rate (%) Real wage per employee post-business sector (% change) Change in labour productivity (%) Change in real wage per unit of output (%) Current account deficit(-), (as a % of GDP) Foreign currency reserves, end of year (billions of $) Source: The Bank of Israel, Annual Report - 20.7 3.1 4.9 36.9 8.9 17.6 3.8 3.9 30.0 9.6 18.0 -0.3 5.6 19.1 10.6 9.4 1.2 4.0 30.2 11.2 11.2 -0.7 3.0 27.5 10.0 -1.6 0.5 -2.1 1.6 -1.3 4.0 -5.1 2.1 -5.2 2.2 -7.3 -0.9 1.8 0.4 1.4 0.3 -0.1 n/a n/a n/a 5.3 6.3 6.3 5.1 7.4 - 1989, 1990, 1991, 1992. importance of realignment expectations, and a reduction in the passthrough from exchange rates to prices of traded goods. On the other hand, it would be difficult to argue that the shift to a crawl per-se had beneficial effects. Table 3 suggests that the same factors that probably led to a marked reduction in the rate of inflation in 1992-93 - such as the persistent fiscal discipline, the relatively high (and persistent) rate of 282 Elhanan Helpman, Leonardo La'derman and Gil Bufman unemployment, and the reduction in foreign inflation account for some of our findings. - could well 5. Additional evidence: Chile and Mexico 5.1 Background 5.1.1. Chile. As in Israel, the evolution of Chile's exchange rate system is closely related to efforts to reduce and stabilize the rate of inflation. During Chile's macroeconomic stabilization of 1973 the government adopted a set of orthodox policy measures consisting mainly of tight fiscal and monetary policies. The next stage was exchange rate-based stabilization. The authorities first established a tablita, i.e., a preannounced schedule of nominal depreciation. In 1979 they shifted to a fixed exchange rate against the US dollar which was maintained until 1982. The fixed exchange rate, along with the accompanying policies, resulted in a considerably lower rate of inflation, but it was also associated with a sharp overvaluation of the domestic currency. The futed exchange rate was abandoned in 1982, and considerably depreciated, when new policies and reforms were adopted. The crawling band was adopted in 1985. Chile's exchange rate is determined in an intra-bank market. It is allowed to fluctuate within a band around a reference rate set by the Banco Central (see Banco Central de Chile, 1991, and Ffrench-Davis and Vial, 1990, for details and additional references). Following two steep devaluations in 1985, the authorities adopted a policy of daily adjustments in the Peso/US Dollar reference exchange rate. At the start of each month, the authorities announce the size of the daily exchange rate adjustments for that month, based on the estimated difference between domestic inflation in the previous month and a forecast of foreign inflation. The width of the band was 2% around the central parity in the initial phase, increased to 3% in January 1988, and further widened to 5% in June 1989. On January 23, 1992 there was a discrete revaluation of 5% in the central parity and the band's width was increased to 10% around the reference rate. From July 6, 1992 the band policy is defined in terms of a basket of foreign currencies (as in Israel) and not in relation to the US dollar as in the earlier periods. Accordingly, the analysis that follows divides the sample into five sub-periods. The adoption of an exchange rate band in Mexico in November 1991 is another case of a policy shift toward increased exchange rate flexibility some time after a major disinflation has been achieved. The Mexican economy was at the edge of a major crisis in the mid-eighties, with growing inflation (143% in 1987), a high and rising public sector external debt, and limited access to international capital 5.1.2. Mexico. New exchange rate bands 283 markets. The Economic Solidarity Pact, signed in December 1987, aimed at breaking the inflationary spiral. The Pacto consisted of a social/ economic accord between workers, government, and entrepreneurs and it was supported by deep fiscal and monetary adjustments, as well as major structural reforms. These changes resulted in sharp disinflation: 18.8% in 1991, 11.9% in 1992 (i.e., the lowest rate in 17 years), and 9% in 1993. The budget swung from a deficit of 16.1% of GDP in 1987 to a surplus of 0.4% in 1992. The ratio of external debt to GDP fell from 92.5% in 1987 to 24.4% in 1992. Similarly, the ratio of internal debt to GDP was reduced by about 20% of GDP during that period. Exchange rate policy went through three main phases after stabilization. First, after an initial devaluation of 38.9%, the Pacto established a f ~ e dPeso/US dollar rate. Second, the authorities shifted toward increased flexibility by adopting a preannounced crawling peg after January 1989: the Peso/dollar rate was allowed to depreciate by 1 Peso per day in 1989,80 cents per day in 1990, and 40 cents per day in 1991. Third, an exchange rate band was adopted on November 11, 1991. Mexico's band differs from those of Israel and Chile in three main respects. First, while the band's ceiling is subject to a daily pre-announced depreciation by a fixed amount, the band's floor remains fixed. Consequently, the width of the band has been increasing with time. Second, the rate of crawl of the band's upper limit is specified in nominal terms, a futed number of cents per day, rather than in percents. Third, there is no official announcement of a central parity. Only the band's upper and lower limits are officially announced. The authorities also announce every day before trading begins their narrow band target range for that day. The band's upper limit was depreciated at the rate of 20 cents per day in 1992. Its total width increased from 1.2% in November 1991 to 4.3% in December 1992. So far there has been one 'realignment' of the band's parameters: in October 1992, upon renewal of the Pacto, the rate of crawl of the upper limit was increased to 40 cents per day in order to allow for an annual depreciation of 4.6% in 1993. The band's width reached 8.7% at the end of 1993 - a width that is not very different from the 10% in Israel. Under the October 1993 renewal of the Pacto, the floor of the exchange rate band remains f ~ e dand the ceiling continues to be depreciated by 40 cents a day. By the end of 1994 the exchange rate band width will have become 14%. 5.2 Exchange rates 5.2.1. Chile. Key indicators for Chile's exchange rate bands are provided in Table 4, and Figure 1 depicts daily observations of the Peso/US Dollar Table 4. Chile: band characteristics, exchange rate, interest rates, prices and foreign currency reserves PERIOD 1 Band Width = 2% 1986.01-88.04 Period PERIOD 2 Band Width Increased to 3% 1988.05-89.05 PERIOD 3 Band Width Increased to 5% 1989.0691.12 PERIOD 4 PERIOD 5 Band Width Exchange rate Increased to 10% Basket adopted 1992.01-92.06 1992.07-93.06 Band characteristics and the exchange rate; with daily data: Average mid-band rate (1) Band width (in %) Annualized average crawl of mid-rate of the band Average exchange rate (1) Average deviation of exchange rate from mid-band (in %) Standard deviation from mid-band (in %) +/- 204.0 2 22.0 205.4 0.67 1.24 Interest rates using weekly data, in percent per month terms: Average domestic interest rate (short-term deposits) Standard deviation of domestic interest rate Average foreign interest rate (LIBID) (2) Average interest rate differential (2) Annualized average interest rate differential (2) Standard deviation interest rate differential (2) 0.47 0.49 1.16 14.8 249.2 3 323.3 5 381.6 10 409.1 10 5.3 247.4 27.9 316.2 13.9 351.8 15.4 384.5 -0.70 1.20 -2.22 4.02 -7.82 2.07 -6.03 2.28 f; J 2 & g P E iz 3 Infition; with monthly data, in percent per month &ms: Average domestic inflation rate (CPI) Average foreign inflation rate (CPI) (2) Average inflation rate differential (2) Annualized average inflation rate differential (2) Standard deviation of inflation differential (2) 1.35 0.24 1.10 14.1 0.54 1.07 0.58 0.49 6.0 0.74 1.72 0.34 1.37 17.8 0.89 0.52 0.25 0.27 0.94 0.19 0.75 3.3 9.4 0.72 0.62 Foreign currenq resmes (non-gold); with monthly data, in millions of dollars per month terms: Average change in reserves Standard deviation of change in reserves 6.8 128.8 54.7 197.7 119.0 220.6 211.1 226.2 143.2 137.2 100.0 106.4 108.1 99.4 100.7 1.6 1.7 2.2 3.2 1.8 Real effective exchange rate Average real exchange rate (index, 86.01-88.04 avg. = 100) Standard deviation of real exchange rate change (in %) Notes : ( 1) For periods 1-4 the exchange rate vis-a-vis the dollar is shown. For period 5 the exchange rate vis-a-vis the Chilean currency basket is used. The base of this index is such that the exchange rate of 1$ on 1July 1992 = the exchange rate of 1 currency basket unit on 1July 1992. The rate of change of this index equals that of the Peso/$ exchange rate up 30 June 1992 and that of the Peso/basket rate from 1July 1992 onwards. ( 2) For periods 1 4 these values are vis-a-vis the U S dollar. For period 5 these values are basket-weighted values using the Chilean currency basket weights. Sources : Banco Central de Chile, Boletin Mensual ; IMF, International Financial Statistics . s R 3 % 3 ?c f% 286 Elhanan Helpman, Leonardo Leidman and Gzl Bufman rate from the inception of the crawling band until March 1993. Two salient features arise from Figure 1: (i) there is a marked upward trend in the exchange rate, which is characterised by a sizeable depreciation over the entire period; and (ii) exchange rate fluctuations make full use of the band's width. In particular, there were several episodes in which the exchange rate persistently stayed at either the upper or lower limit of the band. This second point can be illustrated by a particular episode. During the second half of 1989, the exchange rate stayed at the upper limit of the band when the authorities intervened in the foreign exchange market in order to weaken the Peso. The aim was to reduce the demand for imports and preserve external competitiveness that had declined in previous months. The resulting depreciation of 9% in real terms between May 1989 and the end of 1989 was reversed as high domestic real interest rates along with other factors attracted about $1 billion of short-term capital inflows in the first half of 1990. The appreciation prompted the central bank to purchase about $4.5 billion of Pesos in 1990 and 1991. The outcome was that, for several years now, the exchange rate has remained very close to the lower limit of the band and yet it depreciates along with the crawling depreciation of the central parity. These persistent pressures toward appreciation have contributed to the latest official realignment on January 23, 1992 when the reference rate was revalued by 5% while the band's width was enlarged to +/-lo%. Overall, a detailed analysis shows that the Chilean experience partly backs the simple target zone model since most of the time the exchange rate was very close to either the upper or lower limit of the band but also contradicts it since volatility has been higher near the limits (see Helpman et al., 1993). 5.2.2. Mexico. Exchange rate behaviour under the Mexican band is plotted in Figure 1 and basic data are presented in Table 5 (the figure includes also the latest observations for the crawling peg in 1991 that preceded the band). Intervention inside the band is the rule rather than the exception and this is manifested in the daily announcement of an intra-marginal intervention band. It is not surprising, therefore, that the exchange rate has stayed most of the time near the middle of the band. Along with more volatility near the limits, the Mexican experience also contradicts the simple target zone predictions. 5.3 Interest rate differentials 5.3.1. Chile. Differentials based on creditory nominal interest rates (on deposits for 30 to 89 days) in Chile are plotted in Figure 2. Except for the Nau exchange rate bands Table 5. Mexico: band characteristics, exchange rate, interest rates, prices and foreign currency reserves Period 11 Nov. 9120 Oct. 92 Band characteristics and the exchange rate; with daily data ( i n percents): Total band width at: Beginning of period End of period Annualized crawl of upper limit of band Annualized crawl of calculated mid-band Average exchange rate (Peso/dollar) Average deviation of exchange rate from calculated mid-band Standard deviation from calculated mid-band Average deviation from upper bound Standard deviation from upper bound Interest rates using weekly data, i n percent per month t m : Average domestic interest rate (28-day CETES T-bill) Standard deviation of domestic interest rate Average foreign interest rate (LIBID, U S dollar) Average interest differential (Mexico-U S) Annualized average interest differential Standard deviation of interest rate differential Inflation; un'th monthly data, i n percent per month t m : Average domestic inflation rate Average U S inflation rate (CPI) Average inflation differential Annualized average inflation differential Standard deviation of inflation differential Fara'gn currency reserves (non-gold); with monthly data, i n millions of dollars per month t m : Average change in reserves Standard deviation of change in reserves Real effective exchange rate Average real exchange rate (index, 91.11-92.10 avg.= 100) Standard deviation of real exchange rate change (in %) Source: Banco de Mexico; IMF, International Financial Statistics. 21 Oct. 9230 June 93 288 Elhanan Helpman, Leonardo Leiderman and Gzl Bufman first half of 1990, these rates did not exhibit a marked trend, but rather fluctuated around an almost constant mean with a relatively stable degree of volatility (means and standard deviations are presented in Table 4). The sharp rise in rates in the first half of 1990 was mainly due to two factors: (i) a contractionary monetary policy aimed at reducing the very rapid expansion of aggregate demand; and (ii) the sterilization of sizeable capital inflows. 5.3.2. Mexico. The interest rate differential in Figure 2 exhibits long cycles since the adoption of the band: it decreased until the first quarter of 1992, went onto a strong upward trend which reached a peak with the renewal of the Pacto and since then has moved down. The amplitude of these fluctuations is quite large: from a low of about 6.5% in early 1992 to a high of abow 16% in October 1992. This pattern confirms the important role of devaluation expectations and of anticipations of a Pacto renewal. Overall, both Chile and Mexico contradict the basic model's predictions since the interest rate differential is higher at the margins and there is no systematic relation between the level of the interest rate differential and the deviation of the exchange rate from mid-band. 5.4 The real exchange rate and inflations 5.4.1. Chile. The real effective exchange rate of the Chilean Peso is plotted in Figure 5. From 1985 to 1988 the crawling band regime was associated with a considerable real effective depreciation. From 1988 to 1992, in contrast, there was a weak trend of real appreciation as the consequence of large and persistent capital inflows. This pattern of capital inflows and real exchange rate appreciation is not unique to Chile as documented in Calvo et al. (1993). In order to reduce the scope for excessive short-term capital inflows in the form of domestic borrowing abroad, the central bank imposed on June 1991 a capital import tax of 20% that applies to loans with maturity of less than one year. The tax was raised to 30% in May 1992. The real effective exchange rate returned in the first quarter of 1993 to the level that prevailed in early 1987, which was rather depreciated relative to the early 1980s. Overall, the Chilean band allowed for considerable fluctuations in the real exchange rate. The economy's performance since the inception of the crawling band (and the adoption of other policies) has been impressive. During the 8 At variance with Israel, we do not discuss here the behaviour of foreign exchange reserves because it was primarily determined by the authorities' intervention in the face of capital inflows from abroad and not so much by considerations having to do with the band per se. New exchange rate bands ISRAEL CHILE -s a, 2 0 80 70 60 50 g a, 40 30 20 10 0 -1 0 Jan-86 Jan-88 Jan-90 Jan-92 Jan-87 Jan-89 Jan-91 Jan-93 MEXICO 80 70. 6050. 0 - a~ Jul-90 Jul-91 Jul-92 Jul-93 Rate of CPI change over: Previous month Corresponding month in previous year -+- Figure 11. Rate of inflation (annualized terms) seven years 1986-92 GDP grew at more than 6% per year, inflation averaged 16% per year (see Figure l l ) , unemployment fell from 12% of the labour force in late 1985 to less than 6% in 1992, non-copper exports rose by about 14% per year in volume, and there was a sharp reduction in the ratio of external debt to exports. Underlying these developments was a supportive budget with near fiscal balance (the budget deficit was only about 1% of GDP between 1985 and 1992). 5.4.2. Mexico. The marked reduction in the rate of inflation in Mexico since 1990 shown in Figure 11 happened in spite of increased flexibility of the exchange rate policy and of the possibility that this flexibility would lead to a weaker role of the exchange rate as an inflation-expectations anchor. At the same time, the considerable real appreciation of the Peso 290 Elhanan Helpman, Leonardo hderman and Gil Bufman of about 20% in less than two years after the implementation of the band, shown in Figure 5 and in Table 5 has given rise to mounting pressures on policy-makers to increase the rate of crawl of the exchange rate. 5.5 Devaluation expectations As for Israel, we have estimated the expectation of devaluations for Chile and Mexico. The results are reported in Appendix A, Tables A1 and A2, and plotted in Figure 6.' Clearly devaluation expectations rise as the exchange rate moves toward the upper limit of the band. Again we find that interest differentials do not match expected devaluations. In Chile, there were marked expectations of revaluation for quite some time before revaluation actually took place, e.g. in January 1992, despite the fact that no major change occurred then to the interest rate differential. We also find (Table A2) that increases in economic activity, in foreign exchange reserves, and in the real exchange rate (i.e., real depreciation) are associated with reductions in expected devaluations, though these effects are imprecisely estimated. 6. A comparison and policy implications From comparison of the experience of Chile, Israel and Mexico we draw five main conclusions. First, crawling bands per-se do not eradicate expected devaluations. There always exist circumstances in which unexpected shocks or other factors lead to policy actions that renege on the band's commitment. In all three countries under review there have been some waves of speculative foreign exchange purchases (or sales) as well as discrete realignments of the whole band (as opposed to the built-in crawl). However, the case of Israel shows that the shift from a fixed to a crawling band may reduce realignments uncertainty and may be associated with a reduction in the volatility of the interest rate differential, at least when inflation is high. Moreover, when the realignments have a transparent objective and do not provoke sizeable capital gains, the regime's credibility is not seriously damaged. Second, exchange rate bands are compatible with wide varieties of real exchange rate experiences. In Mexico, the crawling band was associated u 9 We suggest caution in interpreting Chile's estimated expected devaluation because of the low explanatory power of the underlying regression. N m exchange rate bands 291 with a marked real exchange rate appreciation in 1992-93, Chile underwent a sharp real depreciation during the first years of its crawling band, and in Israel the real exchange rate first remained stable and then depreciated. These differences correspond to alternative methods of setting the rate of crawl. In Israel the authorities announce the daily crawl for the next year based on the difference between their inflation target and a forecast of foreign inflation. In Chile the authorities announce each month the planned daily rate of depreciation of the central parity based on the difference between inflation in the previous month and a forecast of world inflation. Thus, if inflation accelerates in month t , real appreciation will be larger in Israel than in Chile where the exchange rate is more likely to be accommodative in month t 1. The counterpart is that inflation shocks are more likely to persist in Chile than in Israel. Mexico is similar to Israel in that the exchange rate target is preannounced on an annual basis, except that this target applies only to the band's upper limit. In practice, the Mexican authorities have devalued at a slower pace than the inflation differential, hence the real appreciation. Third, despite the considerable width of the bands under study, only in Chile was the exchange rate allowed to widely fluctuate. Israel and Mexico conducted extensive intra-marginal intervention which kept the exchange rate within an inner band narrower than the official band.'' Such interventions are often explained by the risk of destabilizing expectations of inflation and devaluation as the exchange rate approaches its upper limit. To assess this issue, we use our earlier estimates of devaluation expectations to simulate the effects of movements of the exchange rate within the band." The simulated paths are shown in Figure 12. In Israel and Mexico we observe a rapid and strong response of expected realignments to exchange rate depreciations within the band. On the other hand, in Chile such an effect is absent. Thus, it is precisely those countries where exchange rate movements appear to have strong impacts on realignment expectations that intra-marginal interventions have been actively used. However, there could be a problem of causality here. It may well be that expectations are relatively more destabilizing in Israel and Mexico than in Chile because of agents' knowledge of the authorities' preference for exchange rates near the central parity. Fourth, in the three countries the shift toward more flexible exchange rate regimes, under a relatively tight fiscal discipline, was not associated + 10 The standard deviation for Chile is about twice the size of that for Israel, which in turn is about three times the level of Mexico's (see Tables 1, 3, and 5). 11 Specifically, we use the estimated coefficients on the xvariables in Table A l , assuming all else constant, and plot the calculated expected devaluations for hypothetically feasible values for x depending on each country's band characteristics. 292 Elhanan Hebman, Leonardo La'derman and Gil Bufman .. A A MEXICO A Position within band (%) xt (CHILE, ISRAEL), yt (MEXICO) Figure 12. Expected devaluation as a function of the exchange rate's position within the band Note xt is defined as the deviation of the exchange rate from central parity. is defined as the deviation of the exchange rate from the upper limit of the band. Based on equation (3) and parameter estimates shown in Table A1 in Appendix A. with a rise in inflation. On the contrary, the rate of inflation exhibits a downward trend in recent periods, and has reached singledigit figures in Israel and Mexico. Moreover, in Israel the shift from fixed to crawling bands was associated with a reduction in market-based inflation expectations and in the passthrough from exchange rates to the prices of traded goods. When there is considerable uncertainty about the exact trend of disinflation, it may be useful to have a band flexible enough to accommodate various low-inflation outcomes, such as in Mexico. Finally, in all cases bands are sustained by various combinations of policy choices regarding movements in foreign reserves, domestic interest rates, and exchange rates within the band. Of the three countries studied, Chile exhibits the greatest degree of flexibility in exchange rates and interest rates. In Israel and Mexico, disturbances are primarily met by foreign exchange market interventions that attempt to stabilize exchange rates and interest rates. It is only when the shocks persist that large exchange rate fluctuations within the band or interest rate changes are allowed. The implication is that monetary policy has been awarded a very New exchange rate bands 293 limited degree of independence in Israel and Mexico despite the existence of relatively wide bands. We believe that these countries could allow more exchange rate flexibility within their bands in order to provide themselves with added monetary policy flexibility. Discussion George Alogoskoufis Athens School of Economics and Ministry of National Economy This is an interesting paper, which attempts to investigate the nature of exchange rate regimes in high-inflation, small open economies. The focus is on the use of exchange rate bands, after the first stages of disinflation, which started by limiting the rate of depreciation of the exchange rate. As the authors put it, the main objective of the paper is to document and analyse the experience with exchange rate bands of three small open economies, Chile, Israel and Mexico, in an attempt to draw empirical regularities and relevant policy implications. The typical bands that are examined feature a crawling central parity, a unilateral commitment by the authorities to intervene in support of the band, and wide fluctuation margins. The conclusions from the empirical analysis can be summarized as follows: - The existence of crawling bands did not resolve the problem of lack of credibility of exchange rate policy. There were instances of speculation and discrete realignments. However, crawling bands were more credible than conventional bands. - Crawling bands were compatible with a wide variety of real exchange rate and inflation experiences. As expected, these experiences depended on the rate of accommodation of past inflation differentials. - In the two economies (Israel and Mexico) where there was extensive intra-marginal intervention, there is a strong response of expected realignments to exchange rate depreciation within the band. - The shift towards more flexible exchange rate regimes was not associated with a rise in inflation. I shall divide my comments into two parts. In order to highlight some of the comments I shall make them sharper than is justified by the paper itself, which is written in a balanced and careful manner. First, I argue that the crawling bands themselves can hardly be distinguished from typical managed floating regimes. There were only half-hearted and 294 Elhanan Hebman, Leonardo Leiderman and Gzl Bufman short-lived attempts to defend both the central parity and the band. In many respects, therefore, such regimes are more akin to managed floating than to exchange rate bands. Second, I examine the analysis and the conclusions. I argue that it would have been useful to examine more in depth the process of wage and price formation under the different exchange rate regimes, although I recognise that data problems do not make it feasible for these particular countries. 1. Is the new breed of exchange rate bands really distinguishable? The authors argue that they are examining a new breed of exchange rate bands. For short they call them crawling bands, and relate them to the literature on exchange rate bands, based on the famous Krugman (1991) paper. However, the fact that there were frequent realignments of the central parities and relatively wide bands should, right from the start, put a question mark on whether this literature really has relevance for the regimes examined in this paper. In many respects, it is irrelevant that the authorities announce a sliding central parity and a band. The behaviour of the exchange rate and of interest rates will depend not on the existence of the band, but on the expected rate of depreciation, whether announced or unannounced. The market will try to anticipate the factors that determine the actual rate of depreciation, and when disturbances occur it will ignore the announced central parities and bands, knowing that the authorities are bound to allow changes in parities without much of a fight. The standard literature on optimal exchange rate management under uncertainty might have sufficed as a framework for examining the experience of these countries. The empirical results in Tables A1 and A2 broadly confirm this. The statistical significance of the dummy variables indicating the nature of the regime is lower when additional macroeconomic variables are included, as in Table A2. Thus the reason why the regime appears to have been significant for expected depreciations may have more to do with the different evolution of the main macroeconomic variables under the different regimes, than with the regimes themselves. The same can be said about the significance of quadratic or cubic terms in deviations from central parities. These are far less significant when additional macroeconomic variables are included. Of course, I do not want to go so far as suggesting that the announcement of central parities and bands has no significance for the very short-term behaviour of exchange rates. What may be more relevant, though, is the set of indicators used to trigger changes in actual exchange rates, and not the characteristics of the crawling bands themselves. New exchange rate bands 2. Is wage and price-setting affected? This is in some ways a rhetorical question, because of data problems in the countries under investigation. It would have been interesting to investigate the link between exchange rate regimes and wage and pricesetting. Some attempt is presented for traded goods prices in Israel. It is reassuring for crawling bands that they were not associated with a rise in inflation or traded g$ods markups. Overall this is an interesting paper, as it documents the experience of small open economies with exchange rate regimes that allow more flexibility than exchange rate bands. It is reassuring that the flexibility of the regimes resulted in a more realistic path for the real exchange rate, combined with the containment of inflation. As such, these regimes might have contained the inflationary pressures that would have followed a switch to free floating. Vittorio Grilli Minister0 del Tesoro, Roma This article extends beyond the usual sample of European countries the large body of literature on exchange rate band systems. The three countries analysed here - Israel, Mexico and Chile - are geographically far apart, but they share the common feature of having relied on the exchange rate as part of a wider stabilization plan. Beyond obvious dissimilarities with Europe (e.g economic conditions) it is important to remember that the exchange rate band system in these countries is a unilateral, not a multilateral, commitment. The main assertion is that, in the context of a macroeconomic stabilization programme, a moving exchange rate band may be preferable to a peg. The authors argue that the moving band system provides an effective nominal anchor and, at the same time, is flexible enough to allow sufficient nominal exchange rate movement to prevent severe real misalignment. Put differently, the moving band system provides an optimal trade-off between inflation control and real exchange rate stability. In evaluating the experience of these three countries, several issues must be addressed. First, does the system provide any of the stabilization properties predicted by theory? Second, is a crawling band system credible? Third, is it possible to control inflation under such an exchange rate arrangement? And finally, can the real exchange rate be stabilized? The three countries under study do not differ much from European countries. There is no support either for the simple target zone theory. 296 Elhanan Helpman, Leonardo Leiderman and Gil Bufman Three explanations are usually advanced to explain the empirical failure of this theory. First, that the bands are not perfectly credible. Second, that monetary authorities adopt foreign exchange intervention strategies other than the assumed infinitesimal interventions at the band limits. Third, that international capital mobility is less than perfect. It has been shown that the combination of less than perfect credibility, intra-marginal interventions and capital controls can produce almost any type of behaviour of exchange and interest rates, thus severely undermining the predictive power of the simple target zone model. In the countries under study all three factors have been present to some extent, which makes the results unsurprising. It is clear from the results at hand that these three countries did not achieve perfect credibility with their band systems. The concept of credibility is harder to define with moving bands than with traditional fixed bands. The authors suggest using as an indicator of credibility estimates of the expectation of a realignment of the band. In all three countries expectations of realignment are never zero, confirming that the systems are not fully credible. However, in the case of Israel the authors find that the shift from a fmed to a crawling band was associated with a reduction in realignment uncertainty and a lesser expectation of realignment. Whether this can be interpreted as evidence of greater credibility of the monetary authority or just as the consequence of having less ambitious policy goals is debatable, but perhaps not important. What is more important is the finding that, despite the shift to a crawling band regime, inflation in all three countries has tended to fall. Thus more exchange rate flexibility did not translate into inflationary pressure. It is in this sense that a crawling band system may enhance the effectiveness of monetary policy. However, it should be recalled that the decrease in inflation occurred in a period, the early 1990s, of worldwide recession and deflation, and thus the evidence should also be interpreted cautiously. We may have to wait until a full business cycle has been completed to draw robust conclusions. Finally, is it true that a crawling band helps the real exchange rate to be stabilized? The evidence here is not conclusive. The three countries display different patterns: a real appreciation in Mexico, a real (initial) depreciation in Chile, and real exchange rate stability in Israel. In conclusion, the evidence presented in this paper is interesting, but cannot yet be used as a clear guide to policy choices in other countries. General discussion Stanley Fischer suggested that the effectiveness of crawling bands should first be assessed in terms of inflation and the extent to which New exchange rate bands 297 overvaluation of the exchange rate could be avoided. He regretted that the authors did not address this aspect more fully, while recognizing the difficulty of evaluating inflation performance at a time of world-wide disinflation. Charles Wyplosz wondered whether the institutional differences between Israel and Chile could not be better exploited; in particular, he wondered whether a crawl based on past information, as in Chile, had different properties from one based on predictions, as in Israel. Along similar lines, David Begg suggested that the different experiences of Chile and Israel could be associated with differences in the starting point. He presumed that the initial stock of credibility enjoyed by these countries should affect the final outcome. Fischer added that the two countries also had different objectives in the first place, with Israel focusing on inflation and Chile focusing on the exchange rate. According to the authors, this difference in objectives should, however, not be overemphasised as the difference had been quickly narrowing over time. Some panel members also wondered about the appropriate theoretical framework in which the properties of crawling bands can be formulated. Victor Norman suggested that in terms of principles the main benefit arising from bands is to induce a process of mean-reversion in the exchange rate. The effectiveness of bands should therefore be assessed in terms of whether they induce such reversion. One salient fact observed in the paper is indeed that mean-reversion was not induced. Accordingly, one should, in his opinion, take a rather more sober view than the authors regarding the effectiveness of bands. Richard Portes suggested that the key differences between crawling pegs and crawling bands should be highlighted, in particular because the experience with crawling pegs in the 1970s is far from conclusive. Vittorio Grilli insisted that imperfect capital mobility should be taken into account more systematically. He wondered how many of the results derived in the target zone literature would be affected in the presence of imperfect capital mobility. If they are, the authors might be evaluating the experience of crawling bands against an inadequate benchmark. The authors acknowledged this, while insisting, however, that measuring the degree of capital mobility is a difficult exercise. Appendix A. Methodology used to estimate realignment expectations At each point in time, the log of the spot exchange rate et can be expressed as the sum of two components: the log of the central parity ct and the log of the deviation of the exchange rate from the central = AX^+^ A Q + ~Taking . parity xt; i.e., et = xt Q. Accordingly, expectations of both sides of this equation conditional on time t + + 298 Elhanan Helpman, Leonardo La'derman and Gzl Bufman information yields an expression for the expected rate of exchange rate depreciation: That is, the expected rate of depreciation equals the expected rate of depreciation within the band E ~ A X ~plus + ~ the expected rate of realignment EtAct+l. Now assume that uncovered interest parity holds and that the foreign exchange risk premium is negligible. Then the lefthand side of (1) equals the interest rate differential (it - 2,'). In the case that the authorities pre-announce a rate of crawl for the central parity for a given time period, this built-in expected adjustment is included in the second term on the right hand side of (1) and it will be denoted by EtAcy+l (where 'w'stands for within the present regime). And in the presence of a fured-central-parity band this term is equal to zero. Thus, the method is general enough to adjust comparisons across various bands for built-in differences in their rates of crawl. We can now subtract from the interest rate differential both the expected rate of depreciation within the band and the expected crawl of the central parity within the existing regime in order to obtain a measure of the expected rate of realignment: Et(zt+1- zt) = (it - i,') - EtAet+l - EtAcy+, The first term on the right-hand side of (2) can be measured directly. For the last term, it is reasonable to rely on the rate of crawl announced by the authorities. This leaves the term for the expected depreciation within the band, which is unobservable. Accordingly, in order to obtain econometric-based time series of this measure and thereby of expected realignments, we use estimates of the following pro'ection of the lefthand side variable on a set of information variables: a where d j is a dummy for 'regime j defined as a period j between realignments or any other changes in the parameters of the band. These equations are estimated by least squares with the covariance matrix corrected for heteroskedasticity. Table A1 presents the various parameter estimates of equation (3) using weekly data for Israel and Mexico, and daily data for chile.13 In the 12 Svensson (1992) provides the rationale for including non-linear xterms as explanatory variables in equations of this type. 13 As in Svensson (1992), data corresponding to realignment dates were not included in the regressions. N m exchange rate bands Table Al. Expected devaluation (estimates of equation (3)) n E t ( z t + l - zt)= CDO + ~ +~ ~+~ + hxt 82.: P~X: D,(zt - .-I) j=1 ISRAEL Parameter estimate Standard error CHILE Parameter estimate Standard error MEXICO Parameter estimate Standard error Note Weekly data used for Israel and Mexico, daily data used for Chile. xt is defined as the deviation of the exchange rate from central parity. In the case of Mexico, st is replaced by y, which is defined as the deviation of the exchange rate from the upper bound of the band. case of Israel, based on preliminary results, the number of dummy variables was reduced from eight to five: the first dummy is for bands 1 and 2; the next three dummy variables are for bands 3-5; and the last dummy variable is for bands 6 to 8. For Chile, the sample period includes four band sub-periods which differ in the width of the band. For Mexico the sample period includes two regimes: from the initiation of the band in November 1991 until the renewal of the Pacto in October 1992, and from this date to the present.'4 The estimated parameters indicate that there is a positive and significant role of nonlinearities in the exchange rate's deviation from 14 Because of the asymmetric nature of the Mexican band, equation (3) was slightly modified and the variable x (i.e. the percentage deviation of the exchange rate from the central parity) was replaced by the deviation of the exchange rate from the band's upper limit. This enables us, by construction, to capture expectations of devaluation due to changes both in the central parity rate and in the width of the band. 300 Elhanan Helpman, Leonardo L a ' d m n and Gil Bufmn mid-band in accounting for devaluation expectations. As the exchange rate rises from the mid-band toward the upper limit this has a more than proportional impact on devaluation expectations. In addition, lagged devaluation expectations have positive and significant explanatory power for current devaluation expectations, a finding which could correspond to a 'Peso problem'. In order to explore the role of macroeconomic indicators as explanatory variables for expected realignment, we expanded the projection in equation (3) to include the rate of change of the real exchange rate, the rate of change in economic activity, and the Table A2. Macroeconomic variables and expected devaluation (estimates of equation (4)) ISRAEL Parameter estimate CHILE Standard error Parameter estimate Standard error Note: The expanded equation is as follows: Where Yt is the logarithm of an index of economic activity, R is non-gold foreign currency reserves (millions of US dollars) and RER is the log of the real effective exchange rate. In the case of Israel, Yt is the log of the Bank of Israel state of the economy index. In the case of Chile the activity variable used is the log of the industrial production index and the macroeconomic variables were lagged one period beyond that shown in equation (6). For both countries monthly data were used in the estimation of the equation. 30 1 New exchange rate bands monthly change in foreign currency reserves. The expanded equation is given by: where Y is the log of an index of economic activity, R is non-gold foreign currency reserves (in millions of dollars), and RER is the log of the real effective exchange rate. Estimates of this equation based on monthly data are reported in Table A2. Appendix B. Quantitative indicators of reduced volatility in the crawling band period We present here the tests used in Section 4. B1. Interest rate differential A first series is based on the time series procedure employed by Schwert (1989), which is a generalisation of a rolling standard deviation estimator (see e.g. Fama, 1976, for a similar estimator) and is similar to the autoregressive-conditional-heteroskedasticity (ARCH) model of Engle (1982). Using this method, an eight-week long rolling conditional standard deviation was constructed. The rolling conditional standard deviation was then regressed on a constant and a dummy variable for the crawling band period (using weekly data). The regression results are: (figures in parenthesis are t- statistics), sion = 0.0024. = 0.12; n = 245; S.E. of regres- The parameter of DCRAWL has a significant negative sign thus indicating a significant reduction in interest rate differential volatility in the crawling band period. B2. Exchange rate deviation from mid-band We use Schwert's (1989) method on the basis of daily data for exchange rates. A one-month rolling conditional standard deviation was regressed 302 Elhanan Helpman, Leonardo La'derman and Gzl Bufman on a constant and a dummy variable for the crawling band period. The regression results are: xt = 0.002 - 0.0008 DCRAWZ, (47.485) (16.553) If = 0.16; n = 1,218; S.E. of regression = 0.0009. The parameter of DCRAWL has a significant negative sign thus indicating a significant reduction in exchange rate volatility in the crawling band period. B3. Realignment expectations The expected realignment series that was estimated in a previous stage (see Appendix A) was regressed on a constant and a dummy variable for the crawling band period (using weekly data). The regression results are: If = 0.02; n= 245; S.E. of regression = 0.0023. The parameter of DCRAWL has a significant negative sign thus indicating a significant reduction in realignment expectations in the crawling band period. Table 2 provides information as to the proportion of observations with expected realignment values significantly higher than zero. It appears that this proportion has decreased in the crawling band period in comparison to that of the previous band regime. A rolling conditional standard deviation of expected realignment was regressed on a constant and a dummy variable for the crawling band period. The regression results are: SDEG = 0.001 - 0.0004 DCRAWZ, (20.129) (4.097) @ = 0.12; n= 124; S.E. of regression = 0.0005. The parameter of D C R A W has a significant negative sign thus indicating a significant reduction in realignment expectation volatility in the crawling band period. New exchange rate bands 303 B4. Black market exchange rate premium The black market premium was regressed on a constant and a dummy variable for the crawling band period (using monthly average data). The regression results are: @ = 0.31; n = 62; S.E. of regression = 0.026. The parameter of DCRAWL has a significant negative sign thus indicating a significant reduction in exchange market premium in the crawling band period. B5. Kolmogorov-Smirnov tests The Kolmogorov-Smirnov test is used to detect shifts in the distributions of the variables of interest - here: the exchange rate (measured in terms of deviation of actual from central parity rates), the interest differential and the real exchange rate - before and after the presumed regime change. The test applies to distributions whose exact form is not known and can not be parametrized. Samples from two independently distributed populations of a variable x are characterised by the (unknown) cumulative density functions H(x) and J(x). The functions H and J are estimated from the empirical distribution functions Hn (x) and J,(x), where m and n are the numbers of observations in the samples. The Kolmogorov-Smirnov statistic for testing the null hypothesis that H = J is given by D = max, I Hn (x)-J, (x) 1, and the critical value for the 0.01 significance level is approximated by 1.63[(n m) /nm]0.5. The results are shown in Table B1. They always reject the null that the distributions have remained unchanged. + B6. Price equations Another approach is the estimation of a price equation with 12 monthly dummy variables and several lags of the rate of exchange rate change and of rate of foreign WPI change. Dummy variables for the crawling band period are included as well in the regression in order to quantify the changes in the equation during the crawling band period. The main results are as follows. First, it appears that pass-through from the exchange rate to traded prices is rapid and is completed within two 304 Elhanan Helpman, LRonardo Leiderman and Gzl Bufman Table B1. Israel: tests for changes in band characteristics. Crawling band period against horizontal band period A Tests for differences in mean values, unweighted variances and distributions of the deviations of the exchange rate from central parity Difference in bands 6 8 in comparison to bands 1-5 Means Unweighted variance Distribution (Kolmogorov-Smirnovstatistic shown) Statistically significant? 0.039% -3.52E-04 0.20 No Yes Yes B Tests for differences in mean values and unweighted variance changes of the interest rate differential Difference in bands 6 8 in comparison to bands 1-5 Means Unweighted variance Statistically significant? -0.002% -2.35E-07 No Yes C Tests for differences in mean values, unweighted variances and distributions of the real exchange rate Difference in bands 6 8 in comparison to bands 1-5 Mean of real exchange rate Unweighted variance of real exchange rate change Distribution of real exchange rate change (Kolmogorov-Smirnovstatistic shown) Statistically significant? 2.24% -8.53E-05 0.00 Yes Yes Yes Note: Differences are shown as bands 6 8 less bands 1-5. months from the change in the exchange rate. The degree of passthrough for the 1989-9 1 period is within the interval 0.15 to 0.39 (i.e. 15% to 39% of the changes in the exchange rate are translated to traded goods prices change), for the period 1991-93 the interval is -0.19 to 0.22. Based on an Ftest, the hypothesis that the decrease in the degree of passthrough is not significant is rejected with a high degree of confidence (97.3%). Also, the hypothesis that the contribution of wages to the explanation of traded goods price changed is rejected. New exchange rate bands 305 Appendix C. Description of data used in the econometric applications C1. Israel Domestic interest rate -weekly observations of the 1-week monetary auction rate for the period between the first week of 1989 and the last week of 1993 (260 weekly observations). Source: Bank of Israel data base. Foreign interest rate - daily observations of Eurocurrency 1-month rates, weighted using Israel's effective currency basket weights and compounded into weekly averages (the official weights of the basket are: 0.6 US dollars, 0.4177 DMs, 0.067 UK pounds and 7.7 Yen). Source: DRI data base. Exchange rate data - daily data of the exchange rate of the NIS against the currency basket and the exchange rate band limits and mid-band rate were used. Source: Bank of Israel data base. Black market exchange rate data is based on daily observations. Source: Bank of Israel data base. Currency option premiums or those of Smonth 'at the money' options written and auctioned by the Bank of Israel. Source: Bank of Israel data base. C2. Chile Domestic interest rate - daily observations of the 30-89 day non-indexed deposit rate for the period January 2"*, 1986 to June 28th, 1993. Foreign interest rate - daily observation of Eurocurrency 1-month rates. For the currency basket period Uuly, 1992 onwards) a basket weighted interest rate was constructed using the effective currency basket weights. For the period prior to the currency basket, the dollar Eurocurrency rate was used. Source: DRI data base. Exchange rate data - daily data of the exchange rate of the Peso against the U S dollar for the period up to June 3oth, 1992 and weighted basket rate for the period of July 1, 1992 onwards (the official weights of the basket are: 0.3497 US dollars, 0.3182 DMs and 17.45 Yen). Daily exchange rate band limits and mid band rate were used. Source Boletin Mensual- Banco Central de Chile. C3 Mexico Domestic interest rate - weekly observations of the 28day CETES nonindexed T-Bill auction rate for the period between the 46th week of 1989 and the 26th week of 1993 (86 weekly observations). Source: Banco de Mexico. Elhanan Hebman, Leonardo Leiderman and Gil Bufman Foreign interest rate - daily observations of dollar Eurocurrency 1month rates compounded into weekly averages. Source: DM data base. Exchange rate data - daily data of the exchange rate of the Peso against the dollar and the exchange rate band limits were used to compute weekly averages. Source: Banco de Mexico. References Argy, V. and P. De Grauwe (1990). Choosing an Exchange Rate Regime - The Challenge for Smaller Industrial Countries, International Monetary Fund. Banco Central de Chile (1991). Evohcim d t la Economia a 1991 y Perspectivas Para 1992. Banco de Mexico (1993). T h Mexican Economy 1993, Mexico. Bertola, G. and R. J. Caballero (1992). 'Target Zones and Realignments,' American Economic Review. Bruno, M. (1993). Crisis, Stabilization and Economic R e f m : Therap) by C o n s e w , Oxford: Oxford University Press. Calvo, G. A., L. Leiderman and C. M. Reinhart (1993). 'Capital Inflows and Real Exchange Rate Appreciation in Latin America,' IMF StaffPapen. Cukierman, A,, M. Kiguel and L. Leiderman (1994). 'Choosing the Width of Exchange Rate Bands: Credibility vs. Flexibility,' CEPR Discussion Paper No. 907. Dornbusch, R. (1982). 'PPP Exchange Rate Rules and Macroeconomic Stability,' Jouml of Politico1 Economy. Engle, R. F. (1982). 'Autoregressive Conditional Heteroskedasticity with Estimates of the Variance of United Kingdom Inflation,' Ecmometrica. Fama, E. F. (1976). 'Inflation Uncertainty and Expected Returns on Treasury Bills,'Jouml ofPolitical Economy. Ffrench-Davis, R. and J. Vial (1990). 'Trade Reforms in Chile: Policy Lessons for the Nineties,' presented at the seminar, Latin America: Facing the Challenges of Adjustment and Growth, EDI, The World Bank, Caracas. Helpman, E., L. Leiderman and G. Bufman (1993). 'A New Breed of Exchange Rate Bands: Chile, Israel, and Mexico,' Discussion Paper No. G93, The Pinhas Sapir Center for Development, Tel Aviv University. Krugman, P. R. (1991). 'Target Zones and Exchange Rate Dynamics,' Quarterly Jouml of Economics. Leiderman, L. (1993). Infition and Disinfition: The Israeli Experiment, Chicago: University of Chicago Press. Lewis, IC IC (1991). 'An Empirical Exploration of Exchange Rate Target Zones: A Comment,' Carnegie-Rochester Conference Series on Public Policy. Schwert, G . W. (1989). 'Why Does Stock Market Volatility Change Over Time?,' Journal o f F i n a n ~ . Svensson, L. E. 0. (1992). 'An Interpretation of Recent Research on Exchange Rate Target Zones,' Journal of Economic Perspectives. Yariv, D. (1989), 'Estimating Inflation Expectations in Israel,' Bank of Is~aelEconomic Review. Zahler, R. M. (1992). 'Politics Monetaria en un Contexto de Apertura de la Cuenta de Capitales,' Banco Central de Chile, Boletin Mmual. http://www.jstor.org LINKED CITATIONS - Page 1 of 2 - You have printed the following article: A New Breed of Exchange Rate Bands: Chile, Israel and Mexico Elhanan Helpman; Leonardo Leiderman; Gil Bufman; George Alogoskoufis; Vittorio Grilli Economic Policy, Vol. 9, No. 19. (Oct., 1994), pp. 259-306. Stable URL: http://links.jstor.org/sici?sici=0266-4658%28199410%299%3A19%3C259%3AANBOER%3E2.0.CO%3B2-S This article references the following linked citations. If you are trying to access articles from an off-campus location, you may be required to first logon via your library web site to access JSTOR. Please visit your library's website or contact a librarian to learn about options for remote access to JSTOR. [Footnotes] 3 An Interpretation of Recent Research on Exchange Rate Target Zones Lars E. O. Svensson The Journal of Economic Perspectives, Vol. 6, No. 4. (Autumn, 1992), pp. 119-144. Stable URL: http://links.jstor.org/sici?sici=0895-3309%28199223%296%3A4%3C119%3AAIORRO%3E2.0.CO%3B2-I 12 An Interpretation of Recent Research on Exchange Rate Target Zones Lars E. O. Svensson The Journal of Economic Perspectives, Vol. 6, No. 4. (Autumn, 1992), pp. 119-144. Stable URL: http://links.jstor.org/sici?sici=0895-3309%28199223%296%3A4%3C119%3AAIORRO%3E2.0.CO%3B2-I 13 An Interpretation of Recent Research on Exchange Rate Target Zones Lars E. O. Svensson The Journal of Economic Perspectives, Vol. 6, No. 4. (Autumn, 1992), pp. 119-144. Stable URL: http://links.jstor.org/sici?sici=0895-3309%28199223%296%3A4%3C119%3AAIORRO%3E2.0.CO%3B2-I References NOTE: The reference numbering from the original has been maintained in this citation list. http://www.jstor.org LINKED CITATIONS - Page 2 of 2 - Target Zones and Realignments Giuseppe Bertola; Ricardo J. Caballero The American Economic Review, Vol. 82, No. 3. (Jun., 1992), pp. 520-536. Stable URL: http://links.jstor.org/sici?sici=0002-8282%28199206%2982%3A3%3C520%3ATZAR%3E2.0.CO%3B2-F PPP Exchange-Rate Rules and Macroeconomic Stability Rudiger Dornbusch The Journal of Political Economy, Vol. 90, No. 1. (Feb., 1982), pp. 158-165. Stable URL: http://links.jstor.org/sici?sici=0022-3808%28198202%2990%3A1%3C158%3APERAMS%3E2.0.CO%3B2-%23 Inflation Uncertainty and Expected Returns on Treasury Bills Eugene F. Fama The Journal of Political Economy, Vol. 84, No. 3. (Jun., 1976), pp. 427-448. Stable URL: http://links.jstor.org/sici?sici=0022-3808%28197606%2984%3A3%3C427%3AIUAERO%3E2.0.CO%3B2-P Target Zones and Exchange Rate Dynamics Paul R. Krugman The Quarterly Journal of Economics, Vol. 106, No. 3. (Aug., 1991), pp. 669-682. Stable URL: http://links.jstor.org/sici?sici=0033-5533%28199108%29106%3A3%3C669%3ATZAERD%3E2.0.CO%3B2-E Why Does Stock Market Volatility Change Over Time? G. William Schwert The Journal of Finance, Vol. 44, No. 5. (Dec., 1989), pp. 1115-1153. Stable URL: http://links.jstor.org/sici?sici=0022-1082%28198912%2944%3A5%3C1115%3AWDSMVC%3E2.0.CO%3B2-C An Interpretation of Recent Research on Exchange Rate Target Zones Lars E. O. Svensson The Journal of Economic Perspectives, Vol. 6, No. 4. (Autumn, 1992), pp. 119-144. Stable URL: http://links.jstor.org/sici?sici=0895-3309%28199223%296%3A4%3C119%3AAIORRO%3E2.0.CO%3B2-I NOTE: The reference numbering from the original has been maintained in this citation list.