CYCLICAL MOVEMENTS OF TOURISM INCOME AND GDP AND THEIR TRANSMISSION MECHANISM: EVIDENCE FROM GREECE Bruno Eeckels, Alpine Center, Athens, Greece beeckels@alpine.edu.gr George Filis, University of Winchester, UK George.Filis@winchester.ac.uk Costas Leon, Democritus University of Thrace, Greece kleon@ierd.duth.gr 1
Abstract In this study we examine the transmission mechanism of the cyclical components of Greek GDP and international tourism income for Greece for the period 1976-2004. Using spectral analysis we find that cyclical fluctuations of GDP have a length of about 9 years and international tourism income has a cycle of about 7 years. The volatility of tourism income is more than three times the volatility of the GDP cycle. VAR analysis shows that the influence of tourism on GDP is rather small, but the influence of GDP on tourism is large, a fact that may be explained on the basis that higher GDP improves the tourist infrastructure and 2 attracts more tourism income.
Structure Purpose of the Study Background Data and Statistical Tests Estimates and Findings Conclusion 3
Purpose of the Study To examine the transmission mechanisms of the cyclical components of GDP and tourism income in Greece for the period 1976 2004. 4
Background of the Study 1/5 Tourism is an important sector of the Greek Economy: 17% of GDP in 1997 (Pavlopoulos, 1999); and generates approximately 700.000 jobs (Greek Ministry of Economy and Finance, 2006). Tourism has numerous economic impacts: foreign exchange, employment sector, business sector, income sector, fiscal sector, etc (Dritsakis, 2004). 5
Background of the Study 2/5 The majority of the existing studies focus mainly on the long run influence of tourism on economic development of a country. Two opposite views: Tourism-led growth and Growth-led tourism. Export-led growth (tourism-led growth): tourism brings foreign exchange which can be used to finance imports, contributing to economic growth (McKinnon, 1964). If imports are comprised of capital goods or basic inputs, tourism will be beneficial to non-tourist regions. Tourism activities also enhance efficiency of local firms as they have to compete with international firms (Bahgwati and Srinivasan, 1979; Krueger, 1980) as well as the exploitation of economy of scale for local firms (Helpman and Krugman, 1980). 6
Background of the Study 3/5 Durbarry (2004) noted that the opposite view, i.e. a growth-led tourism, also exists: economic growth generates further exports. Results from empirical research do not provide a clear picture in favour of either the tourism-led growth or growth-led tourism. Ballaguer and Catavella-Jorda (2002), using cointegration and Granger Causality test, found evidence that supports the tourism-led growth hypothesis for Spain for the period 1975-1997. 7
Background of the Study 4/5 Using a similar approach, Dritsakis (2004) found evidence that supports both the tourism-led growth and growth-led tourism for Greece (1975-1997). Durbarry (2004) found that the export-led growth hypothesis is suggested by the data for Mauritius, and that tourism has a strong impact on the Mauritian economy. On the other hand, Chi-Ok Oh (2005) suggests that GDP Granger causes tourism for Korea. 8
Background of the Study 5/5 The most frequent econometric techniques used are: Cointegration and error corection model (EG - VECMs), and Computable General Equilibrium (CGE) that allows to work with low quality and small length data. In this paper we focus on the short run relationship between tourism and economy, analyzing the statistical properties of the cyclical components of the Greek tourism receipts and Greek GDP in terms of synchronization, duration and persistence, as well as the transmission mechanism of these two macroeconomic variables. 9
Data Annual data series of the Greek: Gross Domestic Product (Y), Tourism Income (TI). Series are expressed in constant 2000 prices, in euros. Data have been obtained from OECD (www.oecd.org) and World Tourism Organization (www.world-tourism.org). Both variables are in logarithms and cover the period 1976 2004, which is translated into 29 observations. 10
First Views on Series 1/2 Long run trend is estimated by the HP filter with smoothing parameter = 100 GDP, Tourism Income. 7.3 4.8 7.2 7.1 4.4 7.0 4.0 6.9 6.8 3.6 6.7 6.6 3.2 6.5 1980 1985 1990 1995 2000 2.8 1980 1985 1990 1995 2000 11
First Views on Series 2/2 Cyclical component = Actual data - Estimated trend. Denoted as C_GDP, C_TI..04.4.03.3.02.2.01.1.00.0 -.01 -.1 -.02 -.2 -.03 -.3 -.04 1980 1985 1990 1995 2000 -.4 1980 1985 1990 1995 2000 12
Descriptive Statistics Table 1: Maximum, Minimum, St. Deviation Variable Max. Min. St.Dev. C_GDP 0.03-0.03 0.05 C_TI 0.35-0.36 0.17 C_TI has a higher amplitude compared to the amplitude of C_GDP. This is expected as tourism income exhibits greater volatility compared to the volatility of GDP (Table 1). Indeed, for GDP is 5% whereas for tourism income is 17%, i.e. more than 3 times higher. This can constitute evidence of the volatile nature of the tourist sector. 13
Co-movement Analysis 1/2 1.co rr( y, x ) = 0.21 t t 2.co rr( yt, xt 4) = 0.43 3.co rr( yt, xt 5) = 0.49 14
Co-movement Analysis 2/2 1. Almost inexistent contemporaneous comovement. 2. Significant co-movements at lags 4 and 5. Tourism and GDP are counter-cyclical variables. Tourism is leading indicator of GDP. 15
4 Tests for Stationarity 1. ADF t-statistics For C_GDP: -3.86. For C_TΙ: -4.80. 2. Roots of the characteristic polynomial: 0.63, 0.52. Moduli: 0.63, 0.52. ADF suggests stationarity at 5% and 10% s.l. Roots of the characteristic polynomial suggests stationarity. 3. Cointegration test with trace statistic suggests stationarity at 1% and 5% s.l. 4. Cointegration test with max. eigenvalue statistic suggests stationarity at 1% and 5% s.l. All 4 tests suggest stationarity! 16
Spectral Densities C_GDP C_TI 0.0018 0.0018 0.25 0.25 0.0016 0.0016 0.0014 0.0014 0.20 0.20 0.0012 0.0012 Spectral Density 0.0010 0.0008 0.0010 0.0008 Spectral Density 0.15 0.10 0.15 0.10 0.0006 0.0006 0.0004 0.0004 0.05 0.05 0.0002 0.0002 0.0000 0.0000 0 2 4 6 8 10 12 14 16 18 20 22 24 26 28 30 Period 0.00 0.00 0 2 4 6 8 10 12 14 16 18 20 22 24 26 28 30 Period Spectral density estimates suggest that the dominant length of the GDP cycle is 9.3 years and for the IT is 7 years. 17
Univariate Analysis: Propagation Mechanism in the Model yt 4 y = ay + ε t i t i t i = 1 is the cyclical component. Persistency may be evaluated on the basis of the magnitude of the estimated parameters in a fourth order autoregressive scheme. The higher (close to but below unity) the estimated parameters, the higher the persistence. For GDP it appears that the first, second and the fourth autoregressive parameter are statistically significant, although only the first lag is relatively high (0.74). For Tourism Income it appears that the first and the second lags are significant (0.81, -0.50). 18
Univariate Analysis: Propagation Mechanisms The dynamic adjustment of the two series is different. With the estimated values of the auto-regressive parameters, we present a simulation of the adjustment to equilibrium after a positive and permanent increase of 1% in each of the cyclical components individually. Simulation shows that: GDP exhibits a oscillating pattern and tourism income a monotonic pattern. Required time for full adjustment is over than 20 years for GDP and 4 years for Tourism Income. 2.0 1.45 1.6 1.44 1.2 1.43 0.8 1.42 0.4 1.41 0.0 1980 1985 1990 1995 2000 1.40 1980 1985 1990 1995 2000 19
VAR Analysis Transmission mechanism is analyzed by means of a VAR model. As the VAR model is stationary, it is estimated in the level terms of the series. The purpose of the VAR is mainly to examine the dynamic adjustments of each of the involved variables to exogenous stochastic structural shocks. 20
Innovation Innovation Accounting: Accounting: Transmission Mechanisms through through Impulse Impulse Response Response Functions Functions Transmission Transmission Mechanisms Mechanisms (TMs) (TMs) depicted depicted by by the the impulse impulse response response function. function. TMs TMs 1 1 2 2 refer refer to to response response of of C_GDP C_GDP to to C_GDP, C_GDP, C_TI C_TI stochastic stochastic shocks. shocks. TMs TMs 3 3 4 4 refer refer to to response response of of C_TI C_TI to to C_GDP, C_GDP, C_TI C_TI stochastic stochastic shocks shocks.016 1.016 2.012.012.008.008.004.004.000 2 4 6 8 10 12 14.000 2 4 6 8 10 12 14.16 3.16 4.12.12.08.08.04.04.00 2 4 6 8 10 12 14.00 2 4 6 8 10 12 14 21
Impulse Response Functions: Dynamic Adjustment Transmission Mechanism Convergence Pattern Dynamic Adjustment 1 Monotonic 11 years 2 Monotonic 11 years 3 Monotonic 9 years 4 Monotonic 11 years Transmission Mechanisms (TMs) 1 2 refer to response of C_GDP to C_GDP and C_TI stochastic shocks. TMs 3 4 refer to response of C_TI to C_GDP, C_TI stochastic shocks. Dynamic Convergence refers to time required to achieve equilibrium in years. 22
Impulse Response Functions: Description In TM 1 the convergence of GDP is monotonic and approximately 11 years are required for equilibrium to be restored. In TM 2 the response of C_GDP to C_TI stochastic structural shock is lower in magnitude than in TM 3. The higher response in TM 3 (response of C_TI to the structural shock of c_gdp) might be explained on the basis that higher GDP means better infrastructure, translated into the provision of better and more efficient tourist services. This is, in fact, an important reason to attract higher tourism income. TM 4 shows the reaction of the cyclical component of tourism income to its own structural shock. The response of the cyclical component of Tourism Income to its own shock is much higher in magnitude than in the TM (C_GDP to its own shock). However, overall convergence to equilibrium is not very different in length of time, since in 3 out of 4 transmission mechanisms, adjustment needs 11 years for equilibrium to be restored. 23
Conclusion 1/2 In this paper we attempted to identify the transmission mechanism between the cyclical components of GDP and tourism income in Greece with data covering the period 1976 2004. Spectral analysis indicates that GDP has a cycle of 9.3 years and tourism income has a cycle of 7 years. The volatility of the tourism income cycle is more than three times the volatility of the GDP cycle. Based on the cross-correlation coefficient we see that there is no contemporaneous movement between the two series, but tourism income is a countercyclical variable and acts as a leading indicator for GDP. 24
Conclusion 2/2 To identify the transmission mechanism, we used two different approaches: a univariate analysis and a VAR analysis, yielding different results. VAR analysis indicates that the influence of tourism on GDP is rather small, but the influence of GDP to tourism is large. This might be attributed to the fact that higher GDP improves tourist infrastructure and attracts tourism income. This cannot necessarily be regarded as a contradiction to other studies showing the large influence of tourism on economic development in Greece. Furthermore, Dritsakis (2004) also found that in the long run there is a Granger causal relationship between economic growth and international tourism earnings for Greece. Propagation mechanism by means of the univariate analysis shows that the adjustment path is oscillating and takes several years to converge (20 years for GDP and 4 for Tourism). On the other hand VAR analysis shows that the time required for the equilibrium to be restored is long, estimated in about 11 years in 3 out of 4 transmission mechanisms. 25
Limitations Further Study We used annual data. Probably higher frequency data will provide better estimates. Since our estimates depend on the filter used (HP) for the de-trending of our variables, cross correlations as well as univariate frameworks do not allow for simultaneous interactions with other possible explanatory variables, and given the difference with the results obtained from the VAR analysis, we consider the above analysis as indicative and tentative. 26
We thank you for your attention! Bruno Eeckels George Filis Costas Leon 27