# The credibility of the European exchange rate mechanism.

1. IntroductionFor most economists the collapse of the Exchange Rate Mechanism (ERM) of the European Monetary System (EMS) in 1993 came not as a surprise. Indeed, in the past many authors had argued that without the introduction of capital controls the EMS could never survive.(1) De Grauwe (1994) has pointed out that the fundamental reason for the fragility of the system lies in the fact that the promise to fix exchange rates lacks sufficient credibility. Nominal and real divergence will sooner or later undermine the credibility of a fixed exchange rate arrangement. Countries like Italy and Spain suffered, for instance, from losses in competitiveness as they had higher rates of inflation than Germany, which was not compensated for by a devaluation of their currencies. Furthermore, the credibility of exchange rate arrangements may be undermined because member countries may disagree with the monetary stance taken by the leading country. In fact, one explanation of the breakdown of the ERM is that due to their high and increasing levels of unemployment countries like the UK and France were unwilling to follow the restrictive policy stance taken by the Bundesbank after German reunification.(2)

Various authors have examined (the factors that have influenced) the credibility of the ERM. Most empirical research focuses upon the credibility of exchange rate policies of individual member states. Attempts in this direction for a number of different European currencies include Koen (1991), Lindberg et al. (1991), Bartolini (1993), Edin and Vredin (1993), De Haan et al. (1993), Chen and Giovannini (1993), Caramazza (1993), Thomas (1994), Drazen and Masson (1994), Eichengreen et al. (1994), Knot and De Haan (1995), Hochreiter and Winckler (1995), Masson (1995), Siklos and Tarajos (1996), and Knot (1997).

Two recent studies have examined the credibility of the system as a whole. Using annual data for the period 1979-93, De Grauwe (1994) concludes that long- and short-term interest differentials vis-a-vis Germany, which are used as indicators for the credibility of the system, are significantly influenced by the average unemployment rate in ERM countries, and by the standard deviation of inflation rates in the ERM. The credibility effect of diverging unemployment rates is of the same order of magnitude as the effect of the dispersion of inflation rates. In an earlier study Rose and Svensson (1994) have estimated a Vector Autoregression (VAR) panel model, using monthly data for six ERM currencies. Realignment expectations are measured by short-term interest differentials and interest differentials adjusted for expected exchange rate movements within the band. Their sample period is March 1979 to October 1992. Variables taken into account include money, output, inflation, international reserves, trade balances, and the real exchange rate. Rose and Svensson conclude that ERM realignments are only weakly related to standard macroeconomic phenomena. Only inflation affects ERM realignment expectations in a systematic way.

It is remarkable that previous studies on the credibility of the system do not mention divergent fiscal policies as a possible factor undermining the credibility of the ERM. According to some theoretical models, budget deficits may cause speculative attacks (e.g. Krugman, 1979). Some studies on exchange rate policies of individual countries report that government budget deficits help explain interest differentials (e.g. Caramazza, 1993; Knot and De Haan, 1995). In this paper it is argued that, apart from unemployment and inflation, budget deficits in member countries also affected the credibility of the ERM. In line with most previous research, credibility is measured by short- and long-term interest differentials vis-a-vis German rates. We have followed the suggestion of De Grauwe (1994) to focus on the credibility of the system. De Grauwe argues that movements of credibility of individual exchange rates are strongly correlated in practice and that econometric analysis of the credibility of individual exchange rates therefore fails to take into account the systematic nature of movements in credibility. Unlike De Grauwe, however, we employ quarterly data and, following Rose and Svensson (1994), employ a panel data approach. Whereas Rose and Svensson only use indicators for short-term credibility, we also employ long-term interest differentials. In comparison with De Grauwe we use a much wider set of variables that might affect the credibility of the ERM. To analyse short-term credibility, we employ both the unadjusted interest differential and the differential adjusted for movements of the exchange rate within the band.

The paper is organised as follows. Section 2 briefly reviews the relevant literature. After a brief discussion of the model and our data in the third section, we present empirical results in Section 4. Finally, Section 5 offers some concluding comments.

2. Theoretical background: review of the literature

Within a target zone, interest differentials vis-a-vis an anchor country can roughly be subdivided in three explanatory pieces: devaluation expectations, expected movements of the exchange rate within the band, and country-specific risk premia (see Knot and De Haan, 1995, for a derivation from first principles). From the target zone model it can be inferred that all three components must ultimately be determined by (the expected trend in) fundamentals. The interpretation of the concept of fundamentals depends on the specific model of exchange rate determination one has in mind. Available theoretical models are inconclusive as to what these fundamentals are, so they can be suggestive at best. For example, in an 'escape-clause' model like that of Giovannini (1990), or in a 'projection-equation' approach as in Chen and Giovannini (1993), any variable will do, provided only that it appears in the central bank's reaction function or in the market participants' information set, respectively. Instead of resorting to a particular model, we take an eclectic (and pragmatic) attitude towards the selection of potential determinants of interest differentials. The objective then is to find proxies that cover a broad spectrum of fundamentals and that capture the essence of expected exchange rate changes, within the band or of the central parity. The set of possible determinants should also include certain policy variables that aim to measure intervention activities on the part of the monetary authorities, like reserve positions.

In order to select additional explanatory variables we consider various models of speculative attacks and subsequent balance-of-payments crises. Krugman's (1979) model, for example, assumes that an exogenous government budget deficit lays at the root of the crisis. Expansionary fiscal policy is financed by domestic credit.(3) According to this model the monetary authorities are prepared to defend the exchange rate peg until reserves reach a pre-specified lower bound. As investors permanently rebalance their portfolios, some of the additional domestic assets will be exchanged for foreign reserves of the central bank. So it becomes inevitable that eventually reserves approach their lower limit. It is straightforward to extend the model to a (semi) small country setting by relaxing the assumption of purchasing power parity (Goldberg, 1991, 1994). In that case a shift to more expansionary fiscal policies will increase the demand for domestic goods, driving up their price level and leading to a real exchange rate appreciation and a decline in competitiveness. The empirical implication of these first generation speculative attack models, then, is that one should observe expansionary fiscal and monetary policies, and/or real exchange rate appreciation prior to a switch in regime, which may consist of a simple devaluation or a move towards a floating exchange rates system.

The second generation of speculative attack models has different empirical implications. Flood and Garber (1984b) and Obstfeld (1986) have formalised the idea that a pegged exchange rate can be successfully attacked even without any apparent problem with its sustainability due to insufficient reserves. In their models multiple equilibria exist in the foreign exchange market because of the contingent nature of the authorities' policy rule. In the absence of an attack, monetary and fiscal policies are in balance, and nothing prevents continuation of the currently prevailing exchange rate peg. An attack may occur, however, if investors rationally anticipate that, after a speculative attack, monetary and fiscal policy will become more expansionary. In that case the exchange rate is expected to be reduced. The main difference with the first generation of speculative attack models is that this occurs even though the fundamentals may not be wrong.

Internal and external macroeconomic balance are two (medium-term) objectives that must be reconciled with 'fundamental equilibrium exchange rates' (Williamson, 1991, 1994), and are likely to dominate the authorities' (contingent) policy rule. External balance is often measured by the current account position of the country concerned. Internal equilibrium is normally defined as the situation where the underlying level of potential output is realised; information on internal balance may be gathered from domestic inflation and the level of unemployment (Clark et al., 1994). Ozkun and Sunderland (1994) postulate a trade-off in the authorities' policy rule between the level of the domestic interest rate and the rate of unemployment. They show that high and rising unemployment might lead the government to abandon the peg. Anticipations of such a policy change could lead to an immediate attack. The unemployment rate signals the stance of government policy and the costs of maintaining a particular stance. Therefore, expectations of a realignment also reflect pressures to increase employment and growth after a period of restrictive policies. Similarly, De Grauwe (1994) stresses the role of different preferences for the weights to be attached to inflation and unemployment in the interpretation of internal balance in the various EMS countries. This asymmetry in national preferences produces conflicts of policies about the appropriate stance of monetary policy for the system as a whole, thereby increasing doubts about the commitment towards fixed exchange rates.

3. The model and data

Based on the review of the literature presented in Section 2 we have included in our set of fundamental determinants of the interest differential ([Delta]) the domestic rate of consumer price inflation ([Pi]); export price competitiveness (EPC) measured as the real exchange rate relative to Germany in terms of export prices; the stock of foreign exchange reserves in relation to GDP (F/Y);(4) the government budget deficit in relation to GDP (D/Y); the current account in relation to GDP (CA/Y); and the rate of unemployment (U)

[Delta] = f([Pi], EPC, F/Y, D/Y, CA/Y, U,...) (1)

Not only does the literature not give clear guidance concerning the choice of the variables to be included, the effect of the variables on credibility is also not always clear. For instance, a high rate of unemployment could imply that in the near future it is likely that a government will no longer stick to its exchange rate policy, thereby reducing the credibility of the exchange rate system (De Grauwe, 1994). However, it could also be interpreted as a signal from government that it is tough (hard nosed) - it is after all willing to accept a higher unemployment rate - which implies that credibility will increase (Drazen and Masson, 1994).

We aim to detect a relationship between various fundamentals and the credibility of exchange rate parities, following the suggestion of De Grauwe (1994) to focus on the credibility of the system. De Grauwe has pointed out that movements of credibility of individual exchange rates are strongly correlated in practice (see also Rose and Svensson, 1994). Econometric analysis of the credibility of individual exchange rates therefore fails to take into account the systematic nature of fluctuations in credibility. This may be especially important when the source of such fluctuations is the emergence of policy conflicts. Unlike De Grauwe, who has used simple averages of variables in the ERM countries to test for the credibility of the system, we employ a panel data approach, which enables us to differentiate between system and country-specific effects. The countries included are: Belgium, Denmark, France, Italy, Ireland, and the Netherlands. We focus on these countries as they have been participating in the ERM for a sufficiently long time to enable testing.(5) Following Rose and Svensson (1994) we employ a panel data approach, although in quite a different way as will be explained below.

Credibility is measured in three ways. First, the one-month interest differential vis-a-vis Germany ([[Delta].sub.1]). According to interest parity, this differential is equal to the expected future exchange rate change. The problem with this measure is, however, that the expected change consists of the sum of the expected exchange rate movement within the band and the expected realignment. Therefore, following Svensson (1993) we have adjusted short-term interest differentials for movements of the exchange rate within the band. This yields our second measure of ERM credibility ([[Delta].sub.a1]). The procedure and the estimation results for the expected exchange rate movement within the band are explained in Appendix 1.(6) Finally, we have taken long-term interest differentials vis-a-vis Germany as credibility measure ([[Delta].sub.1]).

Figure 1 gives an indication of ERM credibility. To emphasise credibility of the system, the figure shows weighted interest differentials vis-a-vis Germany, using GDPs as weight. It follows from Fig. 1 that interest differentials have declined over time, hence credibility has increased. It also follows from Fig. 1 that fluctuations in the adjusted interest differential are much higher than in the other credibility indicators. At one moment, the adjusted interest differential is even negative, which casts some doubt on the reliability of this measure. Finally, Fig. 1 suggests that after the agreement of Basle-Nyborg credibility has increased; in particular, long-term interest differentials have declined further since then.

We use quarterly data, mainly from the IFS-database of the IMF.(7) Data on government budget deficits, foreign exchange reserves, and current account positions will be measured as ratios to (quarterly) GNP. The estimation period is 1979.Q2 to 1992.Q2 for testing short-term credibility. Our sample period stops in the second quarter of 1992 due to the end of ERM participation of Italy.(8) Still, the period includes the years after the German reunification and the lead up to the crisis in 1992 (the Danish and French referenda).

Before estimating the model we analysed the order of integration of the time-series concerned using the Augmented Dickey-Fuller test. It appeared that the variables are I(1).(9) Since the variables concerned are cointegrated, it is possible to estimate an error-correction model (ECM)

[Delta][Delta] = [[Alpha].sub.0] + [[Alpha].sub.1] ([Delta] - Z[Beta]) + [[Alpha].sub.2]X (2)

The first variable denotes the difference between [Delta] and our set of included fundamental variables Z. The second set of variables X includes up to four lags of the first differences of all possible explanatory variables that we have distinguished, including the lagged interest rate differentials. We have included the stock of foreign reserves in X. As we have argued in footnote 4, this variable affects the credibility of the exchange rate regime mainly in the short-run.

An ECM has various advantages. For one thing, it allows to differentiate between short-term and long-term effects. This may be important if these effects are different. For instance, if the central bank raises its interest rate in order to resist exchange rate pressure, this will reduce credibility in the short run according to our measures of credibility. However, in the long run it may enhance credibility.(10)

The long-run model simply consists to the panel data model with most explanatory variables included. We estimated both random and fixed models. The Hausman (1978) test clearly showed that there were no country-specific effects, thereby supporting the view that the credibility of the system as a whole is important.(11) The ECM's have been determined as follows. We started with four lags of all (first differences) of the variables included. The non-significant terms were dropped subsequently, until the last lag included was significant.

4. Credibility of the ERM

Table 1 first presents our estimation results of the long-term model, using the three credibility measures described in the previous section. All explanatory variables listed in Section 3 have been included, except for foreign reserves which is probably only important in the short-run and is therefore only included in the ECM.

Various conclusions can be drawn from Table 1. First, the differences for the various measures of credibility used are quite large. So, the way the credibility of the ERM is measured matters apparently. The claim of Rose and Svensson (1994) that the measures 'typically deliver the same message' is not supported by our findings. So it is important to check the sensitivity of the outcomes for the way credibility is measured, which is generally not done in the literature. Especially the [TABULAR DATA FOR TABLE 1 OMITTED] results for the adjusted interest rate differentials are quite different from those for the other credibility indicators. Furthermore, we do find anything but evidence that fundamentals are more relevant for drift-adjusted devaluation expectations than for raw interest differentials.

Second, the coefficient of export price competitiveness is never significantly different from zero. However, the other coefficients are all significant in the models for the unadjusted interest differentials. In line with the findings of De Grauwe (1994) we find that higher unemployment reduces the credibility of the exchange rate system. We also find that higher budget deficits reduce the credibility of the system (except for the adjusted interest differential), as does the inflation rate (again with the exception of the adjusted interest differential) and the position of the current account.

To test more formally whether the Basle-Nyborg agreement affected credibility, we have added a dummy variable which is one after, and zero before the Basle-Nyborg agreement. The test-statistic of the hypothesis that this agreement did not affect credibility is shown in Table 1. It follows that this hypothesis is rejected in all cases, which suggests that this agreement may have furthered the process of increasing convergence and credibility.

Table 2 reports the outcomes of the ECMs. The error-correction terms have the expected sign and are all significantly different form zero. It also follows that the coefficients of these error-correction terms are rather low, suggesting that it takes quite some time before the long-run equilibrium positions are reached. It also follows from Table 2 that the stock of foreign reserves has the expected effect on credibility. As a higher stock of reserves implies that the central bank will be able to defend its exchange rate longer, it will increase credibility, i.e. reduce interest differentials.

5. Concluding comments

The basic philosophy of the Maastricht Treaty on monetary integration in Europe is that the process of monetary unification is to proceed in stages. At the start of the [TABULAR DATA FOR TABLE 2 OMITTED]

third and final stage of EMU, currencies of qualifying member states will be irrevocably fixed. The transition to this final stage is made conditional on a number of convergence criteria, including criteria for the inflation rate and the government budget deficit. Our results based on observations for the period 1979-92 provide support for the choice of both criteria. We find that inflation and budget deficits in member countries affected the credibility of the ERM. Our results also indicate a deficiency in the Maastricht convergence criteria, since unemployment and current account positions also consistently have affected the credibility of the ERM. Both real and nominal divergence may thus pose a threat on the road to monetary unification in Europe.

Acknowledgements

The views expressed in this paper are the authors' only and do not necessarily represent the views of De Nederlandsche Bank. We would like to thank two referees for their stimulating comments.

1 See e.g. Padoa-Schioppa (1988), Giavazzi and Giovannini (1989), and Portes (1992). However, not all economists were pessimistic about the stability of the ERM. Gros and Thygesen (1992, p. 166) argue, for instance, that 'despite the lack of historical experience with fixed exchange-rate systems without capital controls there is reason to be optimistic about the stability of the EMS. The main reason for optimism is that governments in Europe are aware of the fundamental importance of acquiring a credible exchange-rate commitment. ... A factor that could further reduce the danger of speculative attacks in the intermediate stages is the binding rules for budgetary policy that were proposed for stages II and III of the Delors report. These may require or induce member governments to take corrective actions well before their financial situation becomes unsustainable in the eyes of the markets.'

2 Masson (1995) develops a model in which maintaining the parity exchange rate in the face of inflation that is higher than in the anchor country affects credibility in two opposite ways. A resulting high unemployment signals a tough government, but also makes it less likely that even a tough government will maintain the existing parity. See also Drazen and Masson (1994).

3 Compare also the studies of Flood and Garber (1984a) and Claessens (1991), which introduce uncertainty about the rate of domestic credit creation, thereby increasing the likelihood of a regime shift, so that reserve losses may even exceed increases in domestic credit.

4 It seems likely that this variable affects credibility mainly in the short-run. Indeed, in our long-term model this variable never turned out to be significant. In the estimation results as reported in the paper the variable is therefore only included in our short-term model. See below for further details.

5 Although in principle it would have been possible to use an unbalanced dataset, we have decided not to include new members in our panel, given the relatively short period of membership. See also Rose and Svensson (1994) who include the same group of countries in their analysis as we do, for the sane reason. For instance, Spain entered the ERM only in June 1989, which would imply only 12 available observations. It is also not possible to lengthen the estimation period as the fluctuation band was increased in August 1993. Increasing the number of observations by using monthly data is also not an option since some of the crucial variables that we include (like the government budget deficit) are not available on a monthly basis. The use of quarterly data also implies that no cyclically adjusted budget deficit figures could be employed, as they are not available on a quarterly basis.

6 This Appendix is based upon Knot (1996).

7 Sources and definitions of the variables are to be found in Appendix 2.

8 Rose and Svensson (1994) employ the same estimation period.

9 As the debt ratio is I(2) it is not possible to include the public debt-to-GDP ratio as an additional explanatory variable. Only when this variable is differenced twice does it become stationary and then it is no longer discernable from the budget deficit. As a consequence the selected fundamentals are mainly of a flow type.

10 This point was suggested by one of the referees.

11 The qualitative results of the fixed-effects model were very similar to those of the random-effects model. Results are available on request.

References

Ball, C.A. and Roma A. (1994). 'A Jump Diffusion Model for the European Monetary System', Journal of International Money and Finance, 12, 475-92.

Bartolini, L. (1993). 'Devaluation and Competitiveness in a Small Open Economy: Ireland 1987-1993', Working Paper, No. 83, IMF, Washington, DC.

Caramazza, F. (1993). 'French-German Interest Rate Differentials and Time-Varying Realignment Risk', IMF Staff Papers, 40, 567-83.

Chen, Z. and Giovannini, A. (1993), 'The Determinants of Realignments under the EMS: some Empirical Regularities', Working Paper, No. 4291, NBER, Cambridge, MA.

Claessens, S. (1991), 'Balance of Payments Crisis in an Optimal Portfolio Model', European Economic Review, 35, 81-101.

Clark, P., Bartolini, L, Bayoumi, T, and Syrnanski, S, (1994). 'Exchange Rates and Economic Fundamentals: a Framework for Analysis', IMF Occasional Paper, No. 115.

De Grauwe, P, (1994). 'Towards European Monetary Union without the EMS', Economic Policy, 149-85.

Delgado, F. and Dumas, B. (1991). 'Target Zones: Broad and Narrow', in P. Krugman and M. Miller (eds), Exchange Rate Targets and Currency Bands, Cambridge University Press, Cambridge.

Drazen, A. and Masson, P.R. (1994). 'Credibility of Policies versus Credibility of Policymakers', The Quarterly Journal of Economics, 109, 735-54.

Edin, P.A. and Vredin, A. (1993). 'Devaluation Risk in Target Zones: Evidence from the Nordic Countries', The Economic Journal, 103, 161-75.

Eichengreen, B., Rose, A.K., and Wyplosz, C. (1994). 'Speculative Attacks on Pegged Exchange Rates: an Empirical Exploration with Special Reference to the European Monetary System', Working Paper No. 4898, NBER, Cambridge, MA.

Flood, R.P and Garber, P. (1984a). 'Collapsing Exchange-Rate Regimes: some Linear Examples', Journal of International Economics, 17, 1-13.

Flood, R.P and Garber, P. (1984b). 'Gold Monetization and Gold Discipline', Journal of Political Economy, 92, 90-107.

Giavazzi, F. and Giovannini, A. (1989). Limiting Exchange Rate Flexibility: the European Monetary System and Monetary Union, MIT Press, Cambridge, MA.

Giovannini, A. (1990). 'European Monetary Reform: Progress and Prospects', Brookings Papers on Economic Activity, 2, 217-74.

Goldberg, L.S. (1991). 'Collapsing Exchange Rate Regimes: Shocks and Biases', Journal of International Money and Finance, 10, 252-63.

Goldberg L.S. (1994). 'Predicting Exchange Rate Crises: Mexico Revisited', Journal of International Economics, 36, 413-30.

Gros, D. and Thygesen, N. (1992). European Monetary Integration, Longman, London.

Haan, J. de, Knot, K.H.W., and Sturm, J.E. (1993). 'On the Reduction of Disinflation Costs: Fixed Exchange Rates or Central Bank Independence?', Banca Nazionale del Lavoro Quarterly Review, 87, 429-43.

Hausman, J.A. (1978). 'Specification Tests in Econometrics', Econometrica, 46, 1251-71.

Hochreiter, E. and Winckler, G. (1995). 'The Advantage of Tying Austria's Hands: the Success of the Hard Currency Strategy', European Journal of Political Economy, 11, 83-111.

Koen, V.R. (1991). 'Testing the Credibility of the Belgian Hard-Currency Policy', Working Paper No. 79, IMF, Washington, DC.

Knot, K.H.W. (1996). Fiscal Policy and Interest Rates in the European Union, Edward Elgar, Aldershot.

Knot, K.H.W. (1997). 'Fundamental Determinants of Interest Rate Differentials in the ERM', Applied Economics, forthcoming.

Knot, K.H.W. and Haan, J. de (1995). 'Interest Differentials and Exchange Rate Policies in Austria, the Netherlands, and Belgium', Journal of Banking and Finance, 19, 363-86.

Krugman, P.R (1979). 'A Model of Balance of Payments Crises', Journal of Money, Credit, and Banking, 11, 311-25.

Lindberg, H. and Soderlind, P. (1993). 'Intervention Policy and Mean Reversion in Exchange Rate Target Zones', IIES Seminar Paper, Institute of International Economic Studies, Stockholm.

Lindberg, H., Svensson, L.E.O., and Soderlind, P. (1991). 'Devaluation Expectations: the Swedish Krona 1982-1991', Working Paper No. 3918, NBER, Washington, DC.

Masson, P.R. (1995). 'Gaining and Losing ERM Credibility: the Case of the United Kingdom,' The Economic Journal, 105, 571-82.

MacKinnon, J.G. (1991). 'Critical Values for Cointegration Tests', in R.F. Engle and C.W.J. Granger (eds), Long-Term Economic Relationships. Readings in Cointegration, Oxford University Press, Oxford.

Obstfeld, M. (1986). 'Rational and Self-Fulfilling Balance of Payments Crises', American Economic Review, 76, 72-81.

Ozkun, F.G. and Sunderland, A. (1994). 'A Model of the ERM Crisis', Discussion Paper No. 879, CEPR, London.

Padoa-Schioppa, T. (1988). 'The European Monetary System: a Long-term View', in F. Giavazzi, S. Micossi, and M. Miller (eds), The European Monetary System, Cambridge University Press, Cambridge.

Portes, R. (1992). 'EMS and EMU after the Fall', The World Economy, 16, 1-15.

Rose, A.K. and Svensson, L.E.O. (1994). 'European Exchange Rate Credibility before the Fall', European Economic Review, 38, 1185-224.

Sildos, P.L. and Tarajos, R. (1996). 'Fundamentals and Devaluation Expectations in Target Zones: some New Evidence from the ERM', Open Economies Review, 7, 35-59.

Svensson, L.E.O. (1991). 'The Term Structure of Interest Differentials in a Target Zone: Theory and Swedish Data', Journal of Monetary Economics, 28, 87116.

Svensson, L.E.O. (1993). 'Assessing Target Zone Credibility: Mean Reversion and Devaluation Expectations in the ERM, 1979-1992', European Economic Review, 37, 763-802.

Svensson, L.E.O. (1994). 'Why Exchange Rate Bands? Monetary Independence in Spite of Fixed Exchange Rates', Journal of Monetary Economics, 33, 157-99.

Thomas, A.H. (1994). 'Expected Devaluation and Economic Fundamentals', IMF Staff Papers, 41, 262-85.

White, H. (1980). 'A Heteroskedasticity-consistent Covariance Matrix Estimator and a Direct Test for Heteroskedasticity', Econometrica, 48, 817-38.

Appendix 1

Ball and Roma (1994) found that for the various ERM-currencies the process for the exchange rate within the band can be modelled most appropriately as a meanreverting Ornstein-Uhlenbeck process. Moreover, Svensson (1991) has shown that the relationship between the expected rate of depreciation over a finite horizon equal to a month or longer and the position of the exchange rate within the band can be approximated by a linear function for fairly reasonable parameter values. In addition, in a model of optimal intervention policy in a managed-float regime like in Delgado and Dumas (1991) and Lindberg and Soderlind (1993), the expected rate of depreciation within the band is generally not only a function of the current exchange rate; at least domestic and foreign interest rates should be taken into consideration (Svensson, 1994, p. 190). Hence, [E.sub.t][[Delta][x.sub.t + [Tau]][where]NR]/[Tau] can be estimated using the fitted values of a regression of the actual rate of depreciation within the band on observable variables [x.sub.t], [Mathematical Expression Omitted], and [Mathematical Expression Omitted]

[Mathematical Expression Omitted] (A.1)

where the variable [d.sub.j] is a dummy for each period between two realignments, in order to allow the intercepts to vary. The error term [[Eta].sub.t + [Tau]] is uncorrelated with the regressors, heteroskedastic, and serially correlated for overlapping time-intervals ([Tau] [greater than] dt). Ordinary least squares will give consistent estimates of the coefficients, but their standard errors may be inappropriate.

Equation (A.1) is estimated separately for the cross-Deutschmark exchange rates of Belgium (BF/DM), Denmark (DK/DM), France (FF/DM), Ireland (IP/DM), Italy (IL/DM), and the Netherlands (NG/DM), for a time to maturity of one month ([Tau] = dt = 1/12 year). The period covered starts at the beginning of the EMS in March 1979 and ends in December 1992, except for Italy (August 1992). Months before a realignment occurred are excluded from the sample in order not to include potential jumps. So the estimation results presented in Table A.1 must be interpreted as conditional upon no realignment. As heteroskedasticity in the error terms [[Eta].sub.t + [Tau]] of all regressions is likely, White's (1980) covariance matrix has been used to calculate standard errors.

It follows from the first column of Table A.1 that the Belgian franc has witnessed seven realignments since the beginning of the EMS: the intercepts vary between 0.122 and 0.302 and they are all significant. Similar results are found for the Danish krona, the Irish punt, and the Dutch guilder. This could point at a systematic (positive) drift in the exchange rates vis-a-vis the Deutschmark, albeit this magnitude of the drift varies substantially over the various regimes. The same holds for the relative magnitude of the intercepts if a comparison is made between the various currencies. For France and Italy, on the other hand, such a systematic drift in the cross-DM exchange rate cannot be discerned. Significant mean reversion can be asserted in all currencies.

For France, Ireland, and Italy it seems that the coefficient [[Beta].sub.1] hardly indicates a mean-reverting process in [x.sub.t]. It is well known, however, that if the true coefficient [[Beta].sub.1] is zero, there is a downward bias in the estimate of the coefficient and the usual t-distribution does not apply. The critical level for a standard Dickey-Fuller test on a 5% significance level is - 2.88 for this sample size (MacKinnon, 1991, p. 275), so that our test cannot reject the hypothesis of a unit root in [x.sub.t] for France, Ireland, and Italy. Exchange rate policies pursued in these countries have lacked credibility within the one-month horizon at several points in time during the 1980s. Consequently, the position of the exchange rate within the band may have fuelled devaluation expectations, thereby limiting the degree of mean reversion. Furthermore, accompanying estimates of [[Beta].sub.2] and [[Beta].sub.3] suggest that monetary policy has been least effective in France and Italy, compared to the other EMS countries. France and Italy constitute the larger countries in the EMS, in which the willingness to follow the German interest rate 'dictation' has presumably been more limited than in smaller member countries that depend more on Germany as a partner in foreign trade. Consequently, the vigour with which the exchange rate policy is being pursued might well have been somewhat more restricted by conflicting interests in those countries.

With the estimates of Table A.1, we can adjust short-term interest differentials for expected movements of the exchange rate within the band.

[TABULAR DATA FOR TABLE A1 OMITTED]

Appendix 2

Definitions and sources of the variables used

[i.sup.l] 1-month Euromarket interest rates, being period averages of working days, from the Bank of International Settlements (BIS).

[i.sup.l] Long-term interest rate, measured as government bond yields, International Financial Statistics (IFS) line 61.

S Spot exchange rate against the Deutschemark, period averages, computed from IFS line rf.

C Central parities against the Deutschemark. For Belgium, Denmark, France, Ireland, Italy, and the Netherlands this series was taken from De Nederlandsche Bank, Kwartaalbericht, various editions.

[Pi] CPI-inflation quarter to quarter, computed from IFS line 64.

EPC Export price competitiveness relative to Germany, measured as the real exchange rate in terms of export prices (index, 1980 = 100). Export prices were taken from IFS line 74.

Y Gross National Product, IFS, line 99a for Austria and Denmark. This series has been used as a scaling variable and is not available on a quarterly base for Denmark before 1987 and for Belgium, France, Ireland, and Italy over the whole sample. We therefore interpolated the annual series, using quarterly index numbers on total industrial production for each year from OECD, Main Economic Indicators (MEI) or IFS line 66. For the Netherlands we used data from the Centraal Bureau voor de Statistiek (CBS), Kwartaalrekeningen.

F Foreign exchange reserves, IFS line 1l.d (Total Reserves minus Gold).

D Government budget deficit, IFS line 80. For Belgium these data were kindly provided by Mrs. Frieda de Wit from De Nationale Bank van Belgie. For Denmark this series was constructed from data on Governments Bonds Outstanding, kindly provided by Soren Carlsen from Danmark Statistik.

CA Current account balance, IFS line 77ad.

U Standardised unemployment rate (percentage of labour force), OECD Main Economic Indicators.

Printer friendly Cite/link Email Feedback | |

Author: | Knot, Klaas; Sturm, Jan-Egbert; Haan, Jakob de |
---|---|

Publication: | Oxford Economic Papers |

Date: | Apr 1, 1998 |

Words: | 5695 |

Previous Article: | International environmental agreements under uncertainty. |

Next Article: | Verdoorn's law and increasing returns to scale in the UK regions, 1968-91: some new estimates based on the cointegration approach. |

Topics: |