外文翻译原文--汇率波动与爱尔兰英国的贸易，1979-1992

外文--汇率波动与爱尔兰英国的贸易，1979-1992 外文翻译原文--汇率波动与爱尔兰英国的贸易， 1979-1992 Applied Economics. 2001, 33 , 249-26 5 Exchange rate volatility and Irish-UK trade. 1979-1992 ELEANOR DOYLE Department of Economics. University College Cork. Cork. Ireland i Department of Economics, University of Birmingham. Edgbaston. B15 2TT, UK E-mail: e.doyle@ucc.ie This study examines how exchange-rate volatility affected Ireland's exports to its most important trading partner, the United Kingdom, from 1979 to 1992. To ensure reliable inferences regarding income and price elasticities and the impact of exchange rate volatility on exports, the time series properties of the series used are investigated. The analysis here is conducted at both aggregate and 2-digit SITC Division levels since exchange rate volatility can reasonably be presumed to affect sectors differ- ently. Since expectations matter for exchange rate determination real volatility was generated according to a first-order GARCH process. Both real and nominal vola- tility were important determinants for over 35yo of Irish-UK trade, with positive effects predominating . This may be due to the nature of Irish firms operating in a small open economy where they have little option in dealing with increased exchange rate risk excepl to 'weather the storm' for fear of losing market share or lacing costs of either exit, re-entry, or both . 1 I N T R O D U C T I O N A number of qualifications must be made to this view in assessing the potential effects of volatility, For example, The impact of exchange-rate volatility on trade flows is covariances between currencies matter : for a country that essentially an empirical question as theoretical support exports to more than one foreign market, positive covar- exists for both positive and negative effects of volatility. iances can be expected between the home country's and It may be considered that increased risk associated with trading partners' currencies. This is particularly relevant volatility will induce risk-averse agents to direct their for smaller countries that invoice exports in a foreign cur- economic resources to less risky activities. Accordingly, rency and are . therefore, susceptible to importing economic risk-averse exporters may be expected to reduce their shocks from abroad . If exporters are free to ship more trade volumes in response to uncertainty generated by less as the home currency depreciates rises , it is possible greater exchange-rate variability, since profits from trading that the utility from profits could rise with greater exchange internationally decline in line with unanticipated exchange- rate volatility see De Grauwe, I9K8 . In practice, however, rate changes.' " Indeed, following the collapse of the exporters may be unable to respond very quickly and have Bretton Woods system, exchange-rate volatility has to contend with uncertain receipts on their pre-flxed con- increased significantly while the growth in the volume of tracts and volumes. international trade has significantly declined De Grauwe Figure 1 shows that for an exporter with the given supply 1988; Becketti and Sellon 1989 . curve, facing a given price in a foreign currency, the surplus ' In ihis paper the terms exchange-rate variability, uncertainty atid volatility are used interchangeably . ? Forward timrkets can be used for hedging but there are both limitations atid costs associated with their use. Caporale and Doroodian WM poitit out that the size of contracts can be a limitation since in tKc case of US~Canadi;in trade they must average USSI million per contract before hedging can lake plaee . A further limitation is that euslomers must keep mitiimiim deposits, usually lor multiples of 30 days. These factors indieate the difficulties for trading firms in planning the volume and timing of their international transactions to make optima ! use of forward markets . Apptied Economics ISSN 0003-6846 print/ISSN 1466--428.1 online 2001 Taylor & Francis Ltd 249 ; , . Doyle of international trade , on aggregate or bilateral bases, is that the potential for dealing with nonstutionary integrated variables was not addressed in all of the studies carried out Phillips 1986; Engle and Granger 1987 . Inferences based on ordinary least-squares estimation are invalid because the /- and f-ratio test statistics do not converge to their limiting distribution as the sample size increa.ses. Spurious inferences are generated if the levels of nonstationary vari- ables examined are cointegraled . This could have led to incorrect inferences regarding income and price elasticities and the impact of exchange rate volatility on exports in previous studies . Although most recent studies have focused on aggregate is area A . If randomness is introduced around the price trade data only, similar analysis at sectoral levels is also of and the randomness is symmetric EiA increases. Only interest since exchange rate volatility might reasonably be in Ihc case where the exportef s utility function 6/ 11 is presumed to affect industries diflercntly, Studies based on sufficiently concave will volatility be price reducing . De models of industrial organization have revealed that mar- Grauwe 1988 explained that for producers who are risk ket structure is a key variable in explaining firms' foreign averse not risk averse , increased exchange rate variability pricing behaviour Dornbusch 1987: Feenstru . 1989; increases reduces their expected marginal utility of Marston . 1990 . This implies that disparate industry income from exports , inducing more less export activity . responses to exchange rate volatility could he due . for ex- Delias and Zilberfarb 1993 use an asset portfolio ample, to the degree of competition or subslitutability in model, where the asset is a nominal unhedged trade con- the market , or the relative domestic and foreign shares in Iruct subject to exposure to changes in the exchange rate, to the market . show thai increased exchange rate volatility can increase or This study examines how exchange-rate volatility affects decrease trade depending on the nature of the risk aversion trade by empirically assessing the case of Irish exports lo parameter assumed . A convex concave function induces the UK from 1979 to 1992. The analysis begins by examin- increased reduced exports as volatility increases."* De ing aggregate exports to provide a benchmark against Grauwe 1988 . Delias and Zilberfarb 1993 and Broil which the sectoral analysis can be viewed. To focus on and Eckwert 1999 indicate that two effects are at work sectoral data , the study then examines ihe effect of in determining the outcome of the effect of volatility on exchange rate volatility on Irish exports at SITC 2-digit trade . Increased risk leaves traders less disposed to expose Division level to the United Kingdom . This involves asses- their resources to a loss a substitution effect however, sing the impact of volatility in the Irish Pound-Slerling increased risk implies that more resources should be sold exchange rate on these trade flows. The relalionship to avoid a significant loss in revenues an income effect . between volatile exchange rates and trade flows is import- The effect that dominates depends on the degree of risk ant in the analysis of how price changes are Iransmitied aversion. from one country to another , particularly for trade- No clear consensus emerges from the literature as to the dependent economies . Ireland is such an economy and its impact of exchange rate volatility on trade . For example, main trading partner remains ihe United Kingdom , despite research by Bailey et al. 1986, 1987 . Koray and Laslrupes the consistent reduction in Irish trade dependence on the 1989 and Medhora 1990 find no support for the hypoth- UK, as shown in Table 1. esis that exchange rate risk discourages trade . Nevertheless The Irish-UK example provides an interesting examin- Kenen and Rodrik 1986 . Pozo 1992 and Arize 1995 ation of the volatility question for a number of reasons, find a statistically significant negative relationship between particularly because the parity with sterling was discontin- volalilily and trade as do Asseery and Peel 1991 for 4 out ued in 1979. The significance of the UK market a.s a desti- of 5 counlries in their sample . nation for Irish exports also indicates the importance of One reason for the lack of a systematic relationship ihis issue for the Irish economy , particularly in the light between exchange-rate volatility and the value or volume of policy measures followed by progressive Irish govern- The result is shown to be robust to the presence of a forward market with transaction costs and the introduction of production into the model. '' Kor example if liie titility function displays a constant rate of risk aversioti. e.g. 1/ 1 - i C^~" where C is consumption iitid n is the coelficienl of risk aversion, an increase in volntiHty reduces trade if the coellicient of" relative risk aversion is less than utiiiy. ' See Giovannini 1988 , and Fninke 1991 for examples of models where increased exchange rate variability does not tmambiguously lead to a decline in exports. Exchange rate volatility and Irish-UK trade 25 1 Table I. Irish merchandise trade depeiulence on the UK: 1970- II . M O D E L S P E C I F I C A T I O N , D A T A , A N D A N A L Y S I S M E T H O D S % Imports % Exports Trade from UK to UK Ratio^ The model employed here is a simple example of a com- bined demand and supply model of export detertnination 1970 53 66 60 1975 49 54 52 following Goldstein atid Khan , 1978 . Monthly data are 1980 51 43 47 used to estimate long-run equilibrium export demand func- I9S5 43 33 38 tions of the form : 1990 42 34 38 1992 42 32 37 fi,- + , + + hiv, + e, 1 1993'' 36 28 32 1994 36 28 31 1995 36 25 29 where : X, denotes the natural logarithm of the value 1996 35 25 29 of Irish exports to the UK.; export values in Irish pounds 1997 35 25 28 were obtained from the Trade Statistics of helatid 1998 34 22 27 published by the Irish Central Statistics Ofiicc . The 1999" 33 26 export price deflator is taken from Interniitional Nates: "Irish exports tt plus imports from UK as a % of Irish Monetary Fund IFS . Dummy variables were used exports to plus imports from ihe UK . to seasonally adjust the series , ip, is the logarithm of ''The change in the collection of trade data following 1992 with industrial production seasonally adjusted in the United the inlroduciion of Intrastat means thai post-1992 data are not Kingdom, obtained from the International Monetary slrictly comparable to pre-1992 dala . Fund IFS . Industrial production is used to proxy real ' Data for 1999 refer lo the period from January to November . Source: Data from 1970 to 1994 taken from Bannon . 1996 based income as the analysis is conducted using monthly data on Gallagher and McAleese, 1994 and Irish C:SO dala . Data similar to Caporale and Doroodian , 1994 . UK from 1995 to 1999 arc based on author' s estimates using data GDP and industrial production are plotted in Fig . AI from the Irish CSO . in the appendix which indicates that the series do not deviate significantly from each other over the period mcnts in recent decades to reduce Irish trade reliance on considered, re, is the logarithm of the real sterling price the UK. market . Evidenee of this is clear from Ireland's of the Irish pound using wholesale prices . The nominal Huropcati Monetary System EMS entry discussions exchange rate and the Irish WPl were obtained from when poor UK economic performance and high inflation the International Monetary Fund IFS . The UK WPI were identified as key reasons not to maintain the sterling was obtained from the Monthly Dige.st of Statistics link that had prevailed since 1922. Inflation, in particular . produced by the NSO . The real exchange rate is was identified as a probletii since Irish intlation was effec- RE^ [ ,"C/*/|RL /C/'/UK 1 where E is the sterling price lively being imported from the UK Geary . 1976; Bradley . of the Irish pound , v, is a measure of exchange-rale 1977 . The Irish authorities were enthusiastic about joining volatility, discussed further below, e, is a white-noise dis- the European Monetary System in 1979 because of the turbance term . monetary stability it was expected to provide and especially The data period selected is from March 1979. when the as the UK was also apparently set to join . Even with the Irish pound broke the parity link with Sterling until UK opting out , the Irish government's assessment was that December 1992. This choice of end-date is based on the ihe benefits of joining outweighed the costs . During recent introduction of the EU"s method for the collection of trade currency crises the strategy of over-reliance on one market , data Intrastat which commenced in 1993. Post-1992 data ihe UK . was criticized as an inappropriate strategy for are nol strictly comparable with earlier data . Irish firms to follow Kavanagh et ai. 1998 . Yet despite ihese factors, exports to the UK remain high, relative to Each sector i represents an individual SITC Section other destinations . or Division all codes are presented in Table A2 while X denotes total exports . Equation 1 is assumed to be Monetary stability is also sought in the Irish move to the the long-run cointegrating equation . If we assume that single European currency, the Euro . With sterling remain- demand for exports will rise if UK domestic income rises, ing outside the Euro zone, for the moment, it is of interest /3 is expected to be positive . An increase in the real to ascertain just how the variability of the Irish pound - exchange rate i.e. a depreciation of sterling in excess of sterling exchange rate influences Irish trade flows. relative price changes can be expected to lead to a Section II provides a description of the model used in the reduction in exports so that a negative relationship is empirical analysis and addresses both statistical and econo- expected between these variables . Since the efiect of metric issues of relevance. Section III reports the results of exchange rate risk may be positive or negative, the expected the empirical analysis and Section IV presents some sign of b is ambiguous . concluding remarks . 252 E. Dovle Measuring volatility: concepts and estimates The residuals u, are the estimated residuals e, ― xji and the T~, are the estimated variances y,a. Various measures of exchange-rate volatility have been To use the mode ! empirically, assumptions must be made computed and used in attempt s to measure exchange-rate regarding the variables constituting vectors .v, and v,- on risk for examples see Pagan et al.. 1983: Holland , 1984; which the mean of the exchange rate is conditioned . The Asseery and Peel, 1991 . These estimates include the stan- exchange rate was assumed to be generated by an autore- dard deviation, deviations from trend, the dilTerence gressive process of the form:** between previous forward and current spot rates, the Gini mean difTerence coefficient GMD . the coefficient of re, 7o ? + lh 5 variation, the scale measure of variability . It is also a.ssumed that disturbances from Equation 5 are Jansen 1989 was critical of unconditional measures of not autocorrelated and that Equation 3 is modelled as a volatility since they lack parametric models for the time- first-order GARCH process : varying variance of exchange rates . The time-varying vari- ance tnay be attributed to factors such as rumours , political changes, changes in government monetary and fiscal poli- cies that may lead to the variance of forecast errors of the The predicted values from Equation 6 were used as the exchange rate to be nonconstant and autocorrelated . measure of real exchange-rate volatility.^ The inclusion of Goodhart and Giugale 1993 observed systematic pattern s the lagged variance in Equation 6 is based on Bollerslev's in intraday volatility finding that volatility was smaller in 1986 extension of Engle's 1982 work , where the con- intervals where trading volume was smaller - over the ditional variance is an ARMA process . The basic ARCH weekend and lunch-hour - and was particularly larger dur- framework is extended to the generalized ARCH process - ing the first hour of Monday trading for currencies in their called GARCH - that allows for both autoregressive and own domestic market . moving average component s in the heteroscedastic vari- ance. By using the Engle 1982 model, known as the autore- gressive conditional heteroscedasticity ARCH model, The applicability of such effects in terms of the export demand functions is clear. If traders need to forecast the exchange rate risk can be proxied by specifying the vari- ance of the exchange rate as a linear function of the exchange rate to estimate their stream of profits from trad- expected squares of the lagged value of the error term ing, their trading contract s will depend among other fac- from an auxiliary regression determining the mean of the tors on the forecast of the exchange rate and uncertainty exchange rate.'' Given that expectations matter for regarding the accuracy of the forecasts . Trader s will not exchange rate deterrnination . the ARCH approach was waste any useful information available and rather than applied in this study, if it is assumed that the conditional using unconditional information such as the historical vari- mean and variance of the exchange rate may be represented ance 0^ the exchange rate, as rational agents they will use information such as conditional means and variances . as: To measure nominal exchange rate variability VH, the re, x,ft + u, 2 following measure was used : v,a 3 vn, ― |3 period forward/,.^ - .v, 7 f, denotes the natural logarithm of the forward sterling/ where re, is the change in the real exchange in logs rate Irish pound exchange rate obtained from Financial between month s t - 1 and t\ x, is a vector of variables in the Statisiics HMSO Table 13.2: Middle closing spot rates set of information available at / that contributes to the and 3 month s forward margins in London and Table conditional mean [x,li of re, : and v, is a vector of variables 7.1b for 1992 provided by the Bank of England . ,v, denotes also in the information set at / that contributes to the con- the natural logarithm of the middle closing spot sterling/ ditional variance a; of re,. Estimates of the parameter Irish pound exchange rate obtained from the same source vectors d,/iJ are made by imizing the log-likelihood as above . Emphasis is placed on the outliers through the function for the sample, i.e. use of the squared term . Figure 2 presents the estimates of real and nominal volatility . ln L -0.5 M X ln 27r - 0.5 i ; /f/f 4 The spread of real volatility was significantly less than the nominal measure although both measures peaked in ^ Medhora 1990 discusses tfie limitations associated with these measures in detail . This measure was used by Caporale and Doroodian 1994 and Arize 1995 . ^ In the presence of a lagged dependent variable in Equation 5 the Durbin-/i statistic was used to test for first order correlation . The calculated /-statistic of 0.45 was not statistically significant. ^ The estimated coefficients of this equation and the predicted values are all positive . Exchange rate volatility and Irish-UK trade 253 ference between the variables in question in the long run 0.014 In and Menon , 1996 . 0,012 vnom According to the Granger representation theorem Engle 0.01 vwpi and Granger , 1987 , when a vector on n / I time series X, O.ooe are cointegrated with a cointegrating vector \, there exists 0 006 an error-correction representation : 0 004 ? ? 0.002 A L AX, ^ ~ictX,_i ?,0 L ,, 8 0 where A L is a matrix polynomial in the lag operator L with A 0 ― In. 7 is a n x I non-null vector of constants , \'oic: vnom denotes nominal volatility: virpi denotes real voisUility f3 L is a scalar polynomial in L and , is a vector of white measured using the real exchange rale based on relative Wholesale noise errors . In the short run , any deviation from the long- Priees. run equilibrium o'A' 0 will impact on changes in X, Fig. 2. Volatility of sterling!Irish powul e.Kchartge rate: 1979-92. and lead to movement back to equilibrium . If some element of the vector X is being driven by the equilibrium error so .lune 1983 following a 9% devaluation within the ERM in that the relevant element of ? is non-zero , such a feedback March and declined significantly from then onwards to response exists. However, if the /;th element of- is zero, the rise again in late 1992 coinciding with the Irish 'currency ?th element responds only to short term shocks to the crisis'. The period of stability corresponds to the Irish gov- stochastic environment Agenor and Taylor, 1993 . ernment's adoption of sensible monetary and fiscal con- Inclusion of error-correction terms allows for adjust- tractionary policies that contributed to a credible ments of changes in variables in the vector .Y to their exchange rate target . Nominal volatility rose in March long-run equilibrium values to be identified, providing 1986 preceding the devaluation of 3 % in April while no information on the speed of adjustment to disequilibrium significant effect on volatility corresponds with the unilat- errors. When an error-correction term has a statistically eral devaluation of 8% later that year . Sterling weakness of significant coefficient and displays the appropriate i.e. late 1989 coincides with a rise in nominal volatihty . negative sign, the hypothesis of an equilibrium relation- ship between the variables in the cointegrating equation is valid. Falling to take account of coinlegraiion between the variables would lead lo misspecification in the dynamic Cointegration and error-correction modelling structure underlying the model of interest Arize, 1995 . To provide valid statistical inference, it is important to A general-to-specific modelling exercise is used to obtain address the time-series properties of the variables used parsimonious dynamic models for changes in sectoral because all data series considered must be of the same imports AA",, . The error-correction term generated from order of integration to avoid problems of spurious relation- the Johansen cointegration procedures is included as an ships, and incorrect inferences Phillips, 1986; Ohanian . additional regressor to avoid the loss of potentially relevant 1988 . Without verifying the order of integration of the information. variables, evidence of simultaneous correlations rather than long-run relationships may be found . Harris 1995 explains that similar time trends in nonstationary variables III . E M P I R I C A L R E S U L T S could lead to such spurious relationships being found, in which case standard /- and F-statistics do not have the As a preliminary step lo cointegration analysis, the statio- standard distributions generated by stationary series. T o narity of each of the variables was tested using augmented infer causal long-run relations