Oil Shocks and Kuwait s Dinar Exchange Rate: the Dutch Disease Effect

Size: px

Start display at page:

Download "Oil Shocks and Kuwait s Dinar Exchange Rate: the Dutch Disease Effect"

Amos Hill
5 years ago
Views:

1 MPRA Munich Persona RePEc Archive Oi Shocks and Kuwait s Dinar Exchange Rate: the Dutch Disease Effect A-muai, Usama and Che Sab, Normee Universiti Sains Maaysia,Schoo of Socia Sciences, Universiti Sains Maaysia,Schoo of Socia Sciences 2. October 200 Onine at MPRA Paper No , posted 9. November 200 / 4:24

2 Oi Shocks and Kuwait s Dinar Exchange Rate: the Dutch Disease Effect Abstract This study investigates the impact of oi prices on the exchange rate in Kuwait which uses the fixed exchange rate regime to the US doar. Time series data from covering a the oi shocks are used. In order to achieve the resuts of this study, the VAR mode, the Johansen-Juseius Mutivariate Cointegration test and the Granger causaity test are impemented. Due to the resuts we have arrived at, we recommend that Kuwait either maintains its exchange rate regime (pegged to a basket of currencies), or uses a crawing peg regime.. Introduction The Dutch disease theory tackes the increase in a country s revenues that comes from its natura resources (oi, natura gas) causing an appreciation of the oca currency. This makes the manufacturing goods ess competitive and the services sector entanged with the business interests. This theory emerged in 959 during the foowing of the discovery of a arge natura gas fied in the Netherands. One of the famous modes that expains the Dutch disease is the Core mode. This mode which is used to expain the Dutch disease effect was deveoped by Corden and Neary in 982. The mode consists of the non-tradabe sector (the services sector) and two traded sectors, the booming sector (oi or natura gas) and the non-booming sector (the manufacturing sector). This mode showed that during the natura resource boom in the natura resources there wi be an increase in the demand for abor in the booming sector away from the non-booming sector (the non-tradabe sector and the agging sector) and this wi cause a reduction in the production in both the agging sector and the non-tradabe sector. Aso, the increase in the country s revenue wi increase the demand for the goods of the non-tradabe sector (the services sector) and the price of these goods wi increase. To be more specific, the profitabiity of the agging sector wi fa, whie the profitabiity of the non-traded sector

3 increases due to the spending effect and fa due to the resource movement effects. However, the booming sector s profitabiity wi increase due to the resource movement effect and fa due to the spending effects. Many studies have investigated the impact of the Dutch disease during the oi shocks. The evidence shows that the Dutch disease exists. Exampes incude Corden & Neary s (982) study of Egypt, Indonesia, and the Guf countries, Bruno & Sachs (982) study of the United Kingdom, Adam s study (2003) in the African countries, Égert s (2007) in Kazakhstan, Akpan s (2007) in Nigeria, and Oomes & Kacheva s (2007) in Russia. A these researchers found that the Dutch disease exists causing a rea appreciation of the nationa currency and reducing the share of the manufacturing sector in the countries they investigated. Some studies found that a fixed exchange rate regime can aso hep to reduce the effect of the Dutch disease but with a higher infation compared with the fexibe exchange rate (Neary, 987). Aso Corden & Neary (982) in another study found that ony the abor was mobie, causing a decine in the abor in the non-booming sector and an increase in prices of nontraded and other traded goods. Neary (982) found that for the countries that use the fixed exchange rate regime, the effect of the Dutch disease is reduced but wi resut in higher infation than the countries that use the foating exchange rate. However, Lartey (2007) found that the Dutch disease exists under fixed exchange rates regime. Looney (99) found out that when the government reduces its expenditure, it heps to reduce the effect of the Dutch disease during a resource boom. However, Budina et. a. (2007) found that the instabiity in government expenditure is the cause of the reduction in the non-oi growth rather than the Dutch disease. Roemer (994) and Usui (996) both found that the effect of the Dutch disease in Indonesia is sma because the Indonesian government changed its exchange rate regime to the crawing peg regime in 987 and that heped it to maintain the rea vaue of its oca currency. In this study we wi examine if the Dutch disease exists in Kuwait. We choose Kuwait because this country uses the fixed exchange rate to the US doar for more than three decades. We woud ike to examine if the fixed exchange rate regime is an appropriate regime for Kuwait. Kuwait is an important OPEC member, and petroeum pays an important 2

4 roe in its economy. Oi contributes more that 90% of Kuwait s tota exports and 95% of its foreign earnings. 2. Methodoogy In this study, the Vector Autoregression (VAR) methodoogy wi be used. The vector autoregressive (VAR) mode is used for forecasting a system of interreated time series and anayzing the dynamic impact of random disturbances on the system of variabes. The VAR mode is usefu in this situation as it is ess restrictive compared to other modes. The goa of this study is to find out whether the oi price wi cause the Dutch disease in a country that is using the fixed exchange rate regime ike Kuwait by causing a rea exchange rate appreciation. Mode () shows the vector of endogenous variabes. Log REXCH t = α + β og OP t + β 2 og GDP+ β 3 TB t + β 4 LOEXPV t +ε t () where α is the intercept β, β 2, β 3 are the sope coefficients of the mode Log REXCH is the og of rea exchange rate (nationa currency per US doar) Log OP is the og of oi price (US doars per barre) Log GDP is the og of gross domestic product (miions of US doar) Log TB is the trade baance (miions of US doars) Log OEXPV is the vaue of petroeum exports (miions of US doar) ε is the error term. 3

5 3. Data Source The nomina exchange rate, oi price, trade baance, vaue of petroeum exports, and the gross domestic product are taken from the OPEC data statistics. 4. Estimation Procedures 4. Test for Stationarity Since we are using time series data, it is important for us to examine if the variabes are stationary or not, thus we wi use the unit root test in this study, namey the Augmented Dickey Fuer test (ADF). This test can hep us to find out if the variabes are stationary of order I (0) or I (). If we find that a the variabes are stationary at the first difference then the cointegration test wi be empoyed to find out whether the independent variabes have a ong run reationship with the dependent variabe. 4.2 Cointegration Test Johansen (988) and Johansen-Juseius (990) have deveoped an approach that can be used to find out if there is a ong run reationship between the variabes in a regression mode, therefore, we wi use the Johansen-Juseius (JJ) cointegration test in this study. The JJ procedure is based on the vector autoregressive (VAR) mode and the ag ength is determined by using the Akaike Information Criteria. The VAR mode of order p that aows for the cointegration process can be written as foows: p y t = µ + Π ky t-k + ε t (2) k= where y t is a g-vector of I() variabes, µ is a g-vector of constants, and ε t is a g-vector of white noise residuas at time t with zero mean and constant variance. For this study, the regression mode has g = 5 variabes with four independent variabes and one dependent variabe. In estimating the VAR, we wi imit the maximum ag ength to ony two ags due 4

6 to the imited number of observations in this study (n = 3). Equation (2) above can be rewritten in the first difference form as foows: p y t = µ + k= Γ k y t-k + Π y t- + ε t (3) where Γ k = (I A A k ), (k =,p ) and Π = (I A A 2 A k ) is caed the impact matrix that can give us information about the ong run reationship between the variabes. The rank (r) of Π is equa to the number of cointegrating vectors. If Π is of furank, that is r = g, then there are g cointegrating vectors. If 0 < r < g, there exist r cointegrating vectors, which means that there are r stationary inear combinations of y t. If the rank of Π is, there exists ony cointegrating vector. But if the rank of Π is zero, there is no cointegrating equation and the variabes are not cointegrated. The Johansen procedure is based on two kinds of ikeihood ratio tests, the trace test and the maximum eigenvaue test. The test statistic for the trace test is given in the foowing equation: g λ trace (r) = T n(-λ i) (4) r+ where λ i is the argest eigenvaue of the Π matrix, r is the number of cointegration vectors, g is the number of variabes and T is the number of observations. The nu hypothesis under this test means that there are ess than or equa to r cointegrating vectors and the aternative hypothesis is a genera one. For exampe, to test if there is at most ony cointegrating vector, the nu and aternative hypotheses wi be as foows: H 0 : r (there is at most cointegrating vector) against H : r 2 (there are at east 2 cointegrating vectors) If the test statistic is greater than the critica vaue, H 0 wi be rejected. 5

7 The test statistic for the second test, the maximum eigenvaue test is written as foows: λ max (r, r +) = T n( λ r+ ) (5) The nu hypothesis in this test is that there are exacty r cointegration vectors against the aternative hypothesis of (r + ) cointegrated vectors where r =, 2,..., g, g. For exampe, to test the existence of cointegrating vector, the nu and aternative hypotheses are as foows: H 0 : r = (there is exacty cointegrating vector) against H : r = 2 (there are exacty 2 cointegrating vectors) If the vaue of the test statistic is greater than the critica vaue, then H 0 wi be rejected. 4.3 Granger Causaity Test The Granger approach (969) to answer the question of whether a variabe x causes a variabe y is to see how much of the current vaue of y can be expained by past vaues of y and whether adding past vaues of x can improve in the expanation of y. The variabe x is said to be Granger-cause variabe y if the past vaues of x hep in the prediction of the present vaue of y. There is unidirectiona causaity running from x to y if the estimated coefficients on the agged vaues of x are statisticay significanty different from zero as a group in equation (6) and the set of estimated coefficients on the agged vaues of y in equation (7) beow is not significanty different from zero. k y t = k α i y t i + β i x t i + u t (6) n x t = n λ i x t i + θ i y t i + u 2t (7) 6

8 Conversey, unidirectiona causaity from y to x exists if the set of agged coefficients of y in equation (7) is statisticay significanty different from zero but the set of agged coefficients of x in equation (6) is not. Biatera causaity between x and y exists when the set of agged coefficients of x in equation (6) and the set of agged coefficients of y in equation (7) are both statisticay significanty different from zero. Lasty, there is an independence between x and y when the agged coefficients of x in (6) and the agged coefficients of y in (7) are both insignificanty different from zero. If there is at east one cointegration vector among the variabes of the mode in this study, we wi proceed with the estimation of the vector error-correction mode (VECM) to investigate the tempora short-run causaity between the variabes. On the other hand, if there is no ong run reationship (no cointegration) between the variabes in the mode, the vector autoregressive (VAR) mode wi be empoyed to examine the short-run causaity between the variabes. The VECM is a specia form of the VAR for I() variabes that are cointegrated. The VEC mode aows us to capture both the short-run and ong-run reationships. The direction of Granger causaity in the short run and the ong run can be determined by the VECM. The short-run Granger causaity can be estabished by conducting a joint test of the coefficients in the VECM, which is based on the F-test and χ 2 test. The ong-run causa reationship, on the other hand, is impied through the significance of the agged error correction term in the VECM that is based on the t test. For the purpose of this study, if the variabes are I() and cointegrated, the Granger causaity tests wi be based on the foowing VECM mode with uniform ag ength (equations (8)): REXCH t = α + β ect t- + ξ i REXCH t- + φ i og(op) t- + δ i og(oexpv) t- + λ i og(gdp) t- + µ (8) 7

9 In equation (9) above, is the first difference operator, α i is the constant term, β i, ξ i, φ i, δ i, γ i and λ i are the parameters, ect t- is the agged error correction term obtained from the cointegrating equation and µ i is the white noise error. On the other hand, if we do not find cointegration, we woud not be abe to use the VECM to examine the short-run dynamic reationship between the variabes of the mode. Instead we wi estimate a VAR mode (equation (9)) as foows: REXCH t = α + ξ i REXCH t- + φ i og(op) t- + δ i og(oexpv) t- + λ i og(gdp) t- + µ (9) 5. Empirica resuts The ADF tests showed that a the variabes in the exchange rate mode are stationary at the first difference at the % eve of significance, with the exception of the trade baance variabe which is stationary at the first difference at the 5% eve of significance. Tabe : ADF Unit Root Test Resuts Variabe Leve First Difference Intercept Intercept and Trend Intercept Intercept and Trend Log REXCH *** Log GDP *** *** Log OEXPV *** *** TB ** Log OP *** *** Note: *** denotes significance at the % eve and ** at the 5% eve 8

10 5. Johansen-Juseius Mutivariate Cointegration Test Resuts Since we found a the variabes are stationary at the first difference we can use the cointegration test to find whether a ong run reationship exists between the independent variabes and the dependent variabe. Since the cointegration test is very sensitive to the ag ength, hence, before we run the cointegration test we wi test for the optima ag ength for the rea exchange rate mode. From the ag ength resuts we find the optima ag ength is ag 4. Tabe 2: Lag Length Seection from VAR Estimates VAR Lag Order Seection Criteria Endogenous variabes: LREXCH LOP TB LOEXPV LGDP Exogenous variabes: C Date: 07/07/0 Time: 08:3 Sampe: Incuded observations: 33 Lag LogL LR FPE AIC SC HQ NA * * * * * * indicates ag order seected by the criterion LR: sequentia modified LR test statistic (each test at 5% eve) FPE: Fina prediction error AIC: Akaike information criterion SC: Schwarz information criterion HQ: Hannan-Quinn information criterion 9

11 Tabe 3: Johansen-Juseius Cointegration Test Resuts Based on the Trace Statistic Unrestricted Cointegration Rank Test (Trace) Hypothesized Trace 0.05 No. of CE(s) Eigenvaue Statistic Critica Vaue Prob.** None * At most * At most 2 * At most 3 * At most Trace test indicates 4 cointegrating eqn(s) at the 0.05 eve * denotes rejection of the hypothesis at the 0.05 eve **MacKinnon-Haug-Micheis (999) p-vaues Tabe 4: Johansen-Juseius Cointegration Test Resuts Based on the Maximum Eigenvaue Statistic Unrestricted Cointegration Rank Test (Maximum Eigenvaue) Hypothesized Max-Eigen 0.05 No. of CE(s) Eigenvaue Statistic Critica Vaue Prob.** None * At most * At most 2 * At most 3 * At most Max-eigenvaue test indicates 4 cointegrating eqn(s) at the 0.05 eve * denotes rejection of the hypothesis at the 0.05 eve **MacKinnon-Haug-Micheis (999) p-vaues Tabe 3 shows the cointegration trace statistics resuts and Tabe 4 shows the cointegration maximum eigenvaue resuts. In the cointegration rank test based on the trace and the 0

12 cointegration rank test maximum based on the Eigenvaue we find four cointegrating equations at the 5% eve of significance. This indicates a ong run reationship between EXCH and the independent variabes OP, LOEXPV, GDP, and TB. After having found a ong run reationship between the dependent and the independent variabes, the equation wi be normaized on the exchange rate, since the objective in this study is to find out whether the ong run reationship exists between the og EXCH and the other variabes. Tabe 6 beow shows the normaized cointegration vector. Tabe 5: Cointegration Equation Normaized With Respect To LEXCH LREXCH LOP TB LOEXPV LGDP C E ( ) (4.3E-07) ( ) (0.0045) ( ) From Tabe 5, the ong run og RER equation can be written as: LEXCH= ogLOEXPV ogGDP-.82E-05TB ogOP (0) From equation 0 we find that LOP and LGDP are positivey reated to Kuwait s LEXCH, whie TB and LOEXPV are negativey reated to the LEXCH. The reationship between Kuwait s exchange rate and the petroeum export vaue is negative. An increase of % in the consumer price wi decrease the exchange rate (appreciate) by 0.059%. When petroeum exports increase the foreign capita infows wi increase and that wi ead to an appreciating of Kuwait s currency. The reationship between Kuwait s exchange rate and the gross domestic product is positive. An increase of % in its gross domestic product wi increase the exchange rate (depreciate) by %. The increase in oi prices especiay during the fourth oi shocks of caused a rapid economic growth and iquidity in Kuwait. With the fixed exchange rate to the US doar, this made the monetary poicy ess effective to dea with those events, hence, the price eve increases and this eads to higher infation. According to this, the oi price is positivey

13 reated to the exchange rate with an increase of % in the oi price. This wi cause Kuwait s exchange rate to increase (depreciate) by %. We find that the trade baance is negativey reated to Kuwait s exchange rate with an increase of one miion doars in the trade baance. This eads to a fa in the exchange rate by %. Since the increase in the trade baance causes an increase in the foreign earnings this wi cause an appreciation in Kuwait s exchange rate. 5.2 Granger Causaity Resuts The F-test resuts show that due to the two variabes, namey the oi price and the petroeum export vaue Granger causes Kuwait s exchange rate in the short run. The t-test is significant for the agged error correction term (ect (-)). This indicates that a the variabes Granger causes the exchange rate in the ong run. The most important finding in the Granger causaity test is that the oi prices Granger causes Kuwait s exchange rate positivey. This means that the increase in oi prices wi cause Kuwait s exchange rate to depreciate in the short run. Tabe 6 shows the resuts for the granger causaity test. Tabe 6: Granger Causaity Resuts with LOG EXCH as the Dependent Variabe DLOGREH DLOGOP DLOGOEXV DLOGGDP DTB ect(-) f-stats ** ** ** ** Notes: ect (-) represents the error correction term agged one period. The numbers in the brackets show the optima ag based on the AIC. D represents the first difference. Ony F-statistics for the expanatory agged variabes in first differences are reported here. For the ect(-) the t-statistic is reported instead. ** denotes significance at the 5% eve and * indicates significance at the 0% eve. 6. Discussion of resuts This study aims at finding out whether the Dutch disease exists in Kuwait using time series data from covering a the oi shocks. Kuwait uses a fixed exchange rate regime. Since the Dutch disease theory can be expained by the increase in a country s revenues that come from its natura resources (oi, natura gas) causing an appreciation in the oca currency. From the cointegration anaysis, we found no evidence that the Dutch disease exists in Kuwait because the increase in oi prices caused Kuwait s exchange rate to depreciate. Aso there is no effect of the Dutch disease in the short run. This can be seen from 2

14 the Granger causaity resuts that the increase in oi prices Granger causes Kuwait s exchange rate positivey in the short run. This means that the oi price causes Kuwait s exchange rate to depreciate. The expanation above shows that the Dutch disease does not exist in the short run and the ong run in Kuwait s case. the resuts above show that the fixed exchange rate regime can hep to caution the Dutch disease effect on the economy. Neary s (982) resuts support the resuts of this study because he found that the fixed exchange rate can hep to absorb the Dutch disease effect but the consequence is high infation. This can be seen in the oi exporting countries that are pegging its exchange rate to the US doar ike the UAE, Iraq, Qatar, and Saudi Arabia during the recent years especiay in the ast oi shocks of These countries suffered from high eves of infation than the US. Kuwait government has been using the basket peg (yen, UK pound and the euro, besides the US doar) since This exchange rate regime seems to have been successfu in reducing the eve of domestic infation. So it is better for Kuwait to maintain its exchange rate regime (pegged to a basket of currencies) or use a crawing peg regime which heped Indonesia reduce the Dutch disease effect and maintain its currency vaue (Roemer, 994) and (Usui, 996). 7. Concusion This study investigated the impact of oi prices on the exchange rate in Kuwait which uses the fixed exchange rate regime to the US doar. We used time series data from covering a the oi shocks. In this study, the VAR mode, the Johansen-Juseius Mutivariate Cointegration test and the Granger causaity test are impemented. In the cointegration test, we found that the two variabes, namey, oi price and the gross domestic product have a ong run positive reationship with the exchange rate in Kuwait, whie the trade baance and the oi exports vaue have a ong run negative reationship with Kuwait s exchange rate. According to the Granger causaity test resuts we found that a the variabes granger causes Kuwait s exchange rate in the ong run, whie the oi price Granger causes Kuwait s exchange rate in both the short and the ong run. Based on our resuts, we recommend that Kuwait either maintains its exchange rate regime (pegged to a basket of currencies), or uses a crawing peg regime. 3

15 Reference Adam, K., & Bevan, D., (2003). Aid, Pubic Expenditure and Dutch Disease, Econ Papers, Akpan, E., (2007). Oi Prices Shocks and Nigeria s Macroeconomy, Bruno, M., & Sachs, J., (982). Energy and Resources Aocation: A Dynamic Mode of the Dutch Disease, The Review of Economic Studies, vo. 49, pp Budina, N., Pang, G., Wijnbergen, S., (2007). Nigeria s Growth Record: Dutch Disease or Debt Overhang?, Word Bank Poicy Research Working Paper No Égert, B., (2007). Dutch disease scare in Kazakhstan: Is it rea?, Socia Science Research Network (SSNR), Gorden, M., & Neary, P., (982). Booming Sector and De-Industriaisation in a Sma Open Economy, The Economic Journa, vo. 92, pp Lartey, E., (2007) Capita Infows, Dutch Disease Effects and Monetary Poicy in a Sma Open Economy, Review of Internationa Economics, vo. 6, pp Looney, R., (99) Rea or Iusory Growth in an Oi-Based Economy: Government Expenditure and Private Sector in Saudi Arabia, Word Deveopment, vo. 20, pp Neary, P., (987). The Determinants of the Equiibrium Exchange Rate, The American Economic Review, vo. 78, pp Oomes, N., & Kacheva, K., (2007). Diagnosing Dutch disease: Does Russia has the symptoms? Socia Science Research Network ( Roemer, H., (994). Citizen Funds and Dutch Disease in deveoping countries, Resources Poicy, vo. 3, pp Usui, N., (996). Dutch disease and poicy adjustments to the oi boom: A comparative study of Indonesia and Mexico Resources Poicy, vo. 2, pp

The Impact of Oil Shocks on Qatar s GDP

The Impact of Oil Shocks on Qatar s GDP MPRA Munich Persona RePEc Archive The Impact of Oi Shocks on Qatar s GDP Usama A-muai and Che Normee Che Sab Universiti Sains Maaysia, Schoo of Socia Sciences 24. October 200 Onine at http://mpra.ub.uni-muenchen.de/27822/