As a first step in the ARDL technique, the determination of any long run relationship among the variables is estimated. The Bounds F test result in Table 5.2 shows the results of this first stage with the estimated F-test value indicative of the presence of the long run relationships among the variables. As the calculated F-statistic of 3.31 exceeds the upper bound value of 2.943, the null hypothesis of no long run relationship is rejected irrespective of whether the variables used are integrated of order one I(1) or zero I(0). This indicates that a long run relationship exist amongst the variables in equation (5.1).
Table 5.2 Bounds F-Test Results for Tourism and Growth Nexus in Fiji
Critical value band (90%) Model K- degrees of
freedom I(0) I(1)
Estimated F test value
Pass/Fail Equation (5.1) 5 1.825 2.943 3.31 Pass
In the second stage, the ARDL, long run and the short run ECM coefficients are estimated by using the Akaike Information Criteria (AIC) to select the appropriate lags.9 Given the time series period from 1968 to 2006, the lag length has been set to maximum order of 2, as also selected by the AIC. The estimated results for tourism-growth nexus in Fiji is reported in Table 5.3. The adjusted R2 value of 0.50 indicates that tourism growth model explains about 50 percent of the variations in Fiji’s economic growth. The model’s diagnostic tests for serial correlation, functional form, normality of the residuals, and heteroskedasticity do not indicate any concern.
The estimated long run and short run relationships are noted. The coefficients for the standard production function factors of labour force (GLF) and merchandise exports (GX) have positively contributed to economic growth. However, the estimated negative and statistically significant total investment coefficient suggest that investment has declined over time. This has been due to the political instability since 1987 whereby the results indicate that low levels of investment have an adverse effect on growth.
The estimated long run coefficients for labour force suggest that a 1 percent increase in the labour force leads to a 4.12 percent increase in GDP. This result is similar to Gounder (2002) but differs from that of Narayan and Smyth (2005). Narayan and Smyth obtained mixed and insignificant results for the long run and short run labour force coefficients. The estimated long run growth in merchandise exports (GX) coefficient is positive and significant, it suggest that a one percent increase in merchandise goods exports increases GDP by 29 percent. This result differs from the per capita results of Doessel and Gounder (1996) where the merchandise exports coefficient was negative but insignificant for the period 1980 to 1993.
The estimated export of tourism services (GTR) coefficient is positive and statistically significant at the 1 percent level in the short run. The long run coefficient indicates that a 1 percent increase in tourism exports leads to a 12 percent increase in the growth of Fiji’s economy. This finding is similar to the results derived by Doessel and Gounder (1996) where a one percent change in tourism receipts leads to a 14 percent increase in economic growth and Narayan’s (2004b) Computable General Equilibrium (CGE) results where a 10
9
percent increase in tourist expenditure increases GDP by 0.5 percent in dollar value terms. This result also supports the recent finding of Kumar and Prasad (2007) who analyzed the case of the aggregate services exports-growth nexus for Fiji. He found that total service exports, of which tourism exports is a major component, has had a positive impact total output level in both the short run and long run.
Table 5.3 Results for Tourism-Growth Nexus in Fiji
Dependent Variable: Growth in Gross Domestic Product
ARDL Estimates Long Run Estimates ECM Short Run Estimates Variable Coefficient Coefficient Variable Coefficient
4.12 4.12 4.12 GLF (3.05)*** (3.05)*** ∆GLF (3.05)*** -5.36 -5.36 -5.36 LIY (-2.00)* (-2.00)* ∆LIY (-2.00)* 0.15 0.29 0.15 GX (2.10)** (2.45)** ∆GX (2.10)** 0.03 -0.12 GXt-1 (0.41) ∆GXt-1 (-1.97)* 0.12 0.12 0.12 GTR (2.75)*** (2.75)*** ∆GTR (2.75)*** -2.12 -2.12 -2.12 DV (-1.44) (-1.44) ∆DV (-1.44) -2.51 -2.51 ∆DVCY -2.51 DVCY (-2.03)** (-2.03)** (-2.03)* 8.34 8.34 Constant (1.06) ∆Constant (1.06) -1.30 ECMt-1 (-7.62)*** Adjusted R2 0.50 0.80 SER 3.35 SCχ2(1) 0.78 FFχ2(1) 0.05 Nχ2(2) 0.67 Hχ2(1) 0.00
Note: ***, ** and * are the levels of significance at the one, five and ten percent levels of the t-ratios given in brackets. The description of the test statistics are as follows: Adjusted R2 is the coefficient of determination, adjusted for degrees of freedom. SER
is the standard error of the regression. SC stands for Serial Correlation. FF is Functional Form. N is Normality of residuals and H stands for Heteroskedasticity. The critical values of the chi-square distribution for the tests are as follows: χ2(1) = 6.63, χ2(2) = 9.21.
In terms of the impact of political instabilities (DV) and incidences of tropical cyclones (DVCY) coefficients, both indicate negative effects from coups and cyclones that adversely impact the economy. The estimated coup dummy coefficient shows a weak negative significance level (16 percent level) that suggest that coups in Fiji have led to substantial decline in growth in the long run. On the other hand the estimated long run
cyclone coefficient suggests that the adverse effects of tropical cyclones lead to a 2.5 percent decline in growth. This finding is similar to that of Chand (2000) and Gounder and Katafono (2004). The ECM results for the movement of the variables in relation to the previous period’s gap from long-run equilibrium show that the error correction term (ECMt-1), coefficient is negative and statistically significant at the one percent level. This
suggests that during short term disturbances, the speed of returning to equilibrium is fast.