• Tidak ada hasil yang ditemukan

Cointegration Analysis

Modelling Tourism Demand in Tunisia Using Cointegration and Error Correction Models

5.3 Cointegration Analysis

Most of the empirical works on the least-squares regression (OLS) estimates of tourism demand has used variables in levels rather than first differenced or other- wise filtered data. This method is appropriate for stationary time series data, whereas tourism demand and a lot of explanatory variables such as income and prices are not stationary. The risk of the regression of non-stationary variables is to obtain spurious results consisting on a high value of R2and a small value of DW with misleading results of the standard t-tests and F-tests. To avoid the problems of spurious regres- sions, our study tends to model tourism demand by using cointegration technique and Error Correction Models (ECM).

Cointegration is a technique introduced by Granger (1981), Granger and Weiss (1983) and developed by Engle and Granger (1987) to test common trends in series over the long-run and to find eventual stationary relation by a linear combination of two (or more) non-stationary time series.

To estimate possible long-run relationship between variables, they must have common trends and must move together in the long-run (Kulendran 1996). The existence of common trends among the variables is possible through the application of unit root tests such Augmented Dickey-Fuller test (ADF) or Philips-Perron (PP) test to determine whether variables are stationary or not.

Using the ADF tests, we find that all variables have unit roots in level terms, but are stationary in the first difference; therefore, it is possible to apply the Johansen’s likelihood cointegration procedure.

Table 5.1 Definition of the variables used in the model Variable Definition

LNTj,T,t the log of the number of nights spent by residents of origin country j=(FR:

France; GER: German; the UK or IT: Italy) in Tunisia (T) at the instant t LYj,t the log of GDP per capita of country j (in constant dollars of the 2000 year) CPIT the consumer prices index of Tunisia (2000=100)

CPIj the consumer prices index of one of origin country j (2000=100) EXT/j index of Tunisian currency per units of currency of each country j and it

measures the units of the country of origin that are needed for the purchase of one Dinar (2000=100)

CPIC the consumer prices index of the competing country C; C={Spain, Morocco, Egypt or Cyprus} (2000=100)

EXT/C index of Tunisian Dinar per units of currency of the country C (2000=100) αc the weight of the competing destination c

D86∗∗ dummy for controlling the effects of the economic recession in Tunisia in 1986

D91∗∗ dummy to capture the effects of the Gulf War in 1991

D02∗∗ dummy for controlling the effects of the terrorist attack of Djerba in the year 2002

ut random disturbance term

a1 constant term

ai(I=2,. . .7) unknown parameters which are expected to have the following signs:

a2, a40; a3, a5, a6, a70

Spain, Morocco, Egypt and Cyprus (and Turkey) are considered as the major competing destinations to Tunisia.

∗∗Dummy variable takes a value of 1 in the year of the event took place and 0 otherwise.

5.3.1 Estimation Results by Johansen Cointegration Procedure

To check whether the variables of a system are cointegrated and then identify long- run relationships between them, Johansen (1988) and Johansen and Juselius (1990) provided a maximum likelihood estimation approach to cointegration. This proce- dure can detect more than one long-run relationship between tourism demand and the explanatory variables.

Using the Eviews 4.0 program, the results of maximum likelihood estimation show that the most appropriate cointegrating models (i.e. that have an expected sign and which are economically significant) of tourism demand for Tunisia are as follow:

LNTFR,T,t=38.59+5.43LYFR,t−0.93LRT/FR,t+0.20LSPT/FR,t (5.3) LNTGER,T,t=22.73+3.84LYGER,t−0.02LRT/GER,t+0.35LSPT/GER,t (5.4) LNTUK,T,t=14.15+2.86LYUK,t−1.17LRPT/UK,t+2.89LSPT/UK,t (5.5) LNTIT,T,t=29.04+4.47LYIT,t−0.03LRPT/IT,t+1.25LSPT/IT,t (5.6)

Taking into account the above results, we can state that all variables exhibit cor- rect sign. In fact, it seems that income elasticity is particularly high and that relative prices do not play a major role in tourism demand, the British model being apart.

Moreover, substitute prices are elastic only in the cases of the UK and the Italian models.

5.3.2 Tests for Weak Exogeneity

In order to determine the variables that undergo an adjustment process, we use the exogeneity test. Tables 5.2, 5.3, 5.4 and 5.5 show that dependant variable LNTj,T,tis not weakly exogenous in our four models.

From the tables above, we can observe that all adjustment coefficients of tourist nights in the Tunisian hotels “LNTj,T,t” are negative and statistically significant, so the variables LNTFR,T,t, LNTGER,T,t, LNTUK,T,t and LNTIT,T,tare not weakly exoge- nous. In others words, they witness a correction of their deviations with respect to the equilibrium level during the estimation period (1965–2005). Also, we can con- clude that explanatory variables contribute in the correction of the disequilibrium of

Table 5.2 Model of French demand

LNTFR,T,t LYFR,t LRPT/FR,t LSPT/FR,t

Adjustment coefficients –0.355568 0.038298 –0.108900 –0.038684 t Student (–2.40193) (3.77315) (–2.09020) (–0.55150)

Table 5.3 Model of German demand

LNTGER,T,t LYGER,t LRPT/GER,t LSPT/GER,t

Adjustment coefficients –0.604775 0.003929 0.064350 0.010670 t Student (–5.71582) (0.35480) (1.54250) (0.23797)

Table 5.4 Model of English demand

LNTUK,T,t LYUK,t LRPT/UK,t LSPT/UK,t

Adjustment coefficients –0.436130 0.008913 0.040561 0.012305 t Student (–5.26843) (1.58971) (1.22880) (0.60409)

Table 5.5 Model of Italian demand

LNTIT,T,t LYIT,t LRPT/IT,t LSPT/IT,t

Adjustment coefficients –0.694616 0.047355 –0.180411 0.006592 t Student (–3.26026) (1.98242) (–1.42772) (0.07769)

the tourist nights (35% per year in the case of France, 60% in the case of Germany, 43% in the case of the UK, and 69% in the Italian model).

The small values of adjustment coefficients in the cases of France and the UK could be interpreted in such a way that explanatory variables like income and prices do not probably play an important role to equilibrate the variations of the volume of nights in Tunisian hotels from their equilibrium level. It is possible that other factors such marketing efforts undertaken by the authorities or transport coasts could influence more significantly tourism demand.

5.3.3 Empirical Results and Policy Recommendations

5.3.3.1 Income Elasticity

The empirical results indicate that income of the visitors has a positive and signifi- cant impact on tourism demand in Tunisia. According to the cointegration equations (5.3), (5.4), (5.5) and (5.6), income elasticities range between 2.46 for the UK and 5.83 for France. The estimated income elasticity shows that an increase of 1% in real GDP per capita of the UK results in a 2.46% increase in stays at Tunisian hotels.

The values of income elasticity are strangely high, which means that Tunisian tourism is largely influenced by the level of revenue in the European countries. In some ways, this result is reasonable; because economic theory considers foreign holidays as “superior goods” so its income elasticity is generally higher than a unit.

Also, Crouch (1994) argued that, in many cases the estimated income elasticities of tourism demand studies were well above 2.0. Furthermore, many similar studies of European tourism demand in the Mediterranean countries showed that long-run income elasticity of tourism demand is high. For example, Dritsakis (2004) found that income elasticity for German demand to Greece tourism is 2.16 and 6.03 for the English demand; also, Muñoz and Montero-Martín (2007) found that income elasticity for German demand to Spain is equal to 5.40.

5.3.3.2 Relative Price Elasticities

According to the results of main tourism demand studies which concluded that prices exhibit generally a non-significant elasticity, our results showed that tourism in Tunisia is price inelastic because the value of elasticities is less than a unit (apart from the UK model). This result is possibly due to the limitation of the approxi- mation of tourism prices through CPI. In this case, Divisekera (2003) pointed that

“trends in general price levels as implied by Consumer Price Indices measures may not necessarily coincide with that of tourism”.

In the cases of Germany and Italy, relative price elasticities are equal to –0.02 and –0.03 respectively, which means that German and Italian tourists choose Tunisia as their destination with almost no regard to the prices of the hotel’s night in Tunisia.

German tourists are not sensitive to the prices, probably, because they do not have large possibilities of domestic holidaying (essentially in terms of sun and sea).

The little value of price elasticities is compliant with the idea that international tourism is more and more popular. This fact is due to the development of the orga- nized trip and packaged tours, which makes the price of holiday less and less high since tourism operators could minimize their costs through a massive selling.

Tourists from France (–0.93) and UK (–1.17) seem to be more responsive to price changes than tourists from the other countries, which confirms that French and British tourists make arbitration between tourism in their respective countries and tourism in Tunisia before holidaying. This result seems to be realistic in the French case, because it is recognized that France is the first destination in the world and it is common that French holidaymakers choose between a domestic tourism and an international one.

Our results are in line with those of the majority of tourism demand studies.

For example, Crouch (1996) found that the average of price elasticity of 77 stud- ies reviewed is –0.63 which is close to our price elasticity-average which is equal to – 0.53.

5.3.3.3 Cross Price Elasticities

The great value of cross price elasticity of British tourists (2.89) is consistent with their relative price elasticity (–1.17) and emphasizes that Spain, Morocco, Cyprus and Egypt are perceived by the British tourists as an important substitute to Tunisia. In other words, they seem to be very sensitive to a change in the rapport of tourism prices between Tunisia and these destinations. Also, there is an important substitution effect in the case of Italian market (1.25).

The little value of cross price in the case of France (0.20) illustrates the loyalty of French tourists to Tunisia. This fact can be interpreted as a natural consequence of a colonialist past and the important links between the two countries, moreover, it is known that Tunisia is one of the first “non-European” destinations of French tourists.

The value of cross price elasticity of tourism demand of Germany (0.35) may indicate that residents of this country are not very sensitive to the prices practised by Tunisia’s competing destinations since they decide to visit it. This could signify that an enhancing-price policy might encourage more French and German tourists to travel to Tunisia, ceteris paribus.

5.4 Error Correction Models