Invisible Cities
Download 352.16 Kb.
|
- Navigate this page:
- Chapter 5 Results
- 5.1 Models
- 5.1.2 Comparisons between different regions
4.3 Model SpecificationThere is a general demand model proposed by Lim (1997) typically for econometric analysis of international tourism, stated that tourism demand is a function of:
Considering what have been demonstrated in the last two sections, the estimation model for our analysis is quite different from this general model. Income of tourists is neglected from the discussion. It is difficult to be included, because the origins of tourists are too many and income level might differ too much, there is no proxy can represent this variable. Transportation cost is excluded because of the worry of dominant effect and problem of endogeneity. Currency exchange rate, which is BIS in this case, not get involved because the time demeaned variables already controlled its effect. Besides, relative prices and qualitative factors in destinations are taken into account for the following analysis. As a result, the baseline mode of this dissertation is formulated as follows, which only includes price level and some qualitative factors of destination: Baseline Model (Model 1): 
The two hypotheses are testing the effect of sulfur dioxide and media publicity, thus two other models should be formulated respectively, and there is a fully specified model with these two variables at the same time. The fully specified model is going to be used for verifying assumptions of OLS regression and other tests of hypothesis. Sulfur dioxide effect model (Model 2): 
Media publicity effect model (Model 3): 
Fully specified model (Model 4): 
Since there is a wonder of exaggerate effect of media publicity, an interactive variable of sulfur dioxide and media publicity is created to test it. Model of moderated media publicity (Model 5): 
Since there showed an approximate one year recovery time of negative incidents in Section 3.5, leaves a necessity of testing time lagged effect of air pollution, accordingly, media publicity need also to take values of the previous year. Model 6 tests that sulfur dioxide in the previous year might influence the tourist arrivals in the current year, and Model 7 tests that sulfur dioxide in the previous and current years both influence the tourist arrivals in the current year. Time lagged effect model with previous year only (Model 6): 
Time lagged effect model with previous and current years (Model 7): 
The results of tests of significance and assumptions are going to be presented in the next chapter. Chapter 5 ResultsThis chapter will present the results of the models formulated in the last chapter, in order to facilitate find the conclusion of hypothesis. Meanwhile, some relevant tests of assumptions of OLS regression will also be conducted. Therefore, Section 5.1 will present the result of models, and Section 5.2 will present results of goodness-of-fit of the fully specified model. 5.1 Models5.1.1 Results in generalFirst, in order to guarantee the fixed effect model is more appropriate than the random effect model, a Hausman test is carried out using the baseline model. It shows a P-value equal to 0.000, which suggests there is systematic difference between fixed and random effect model. Therefore it is reasonable to use fixed effect model for the following analysis. Second, the independent variables – sulfur dioxide and media publicity – are added in step by step into the model. Table 11 shows the results from Model 1 to Model 4. It does not include year dummy variables because those are all insignificant. Table 11 Results from Model 1 to Model 4
Source: own elaboration Note: * Significant at 10% level, ** Significant at 5% level, *** Significant at 1% level It can be found that two variables in the baseline model keeps significant in 1% level, which are FDI and property price. Mega Event is significant except for in Model 2, indicating that adding sulfur dioxide takes away its impact. After controlling media publicity it turns to significant again, suggests that the focus of media of air pollution was more intensive when mega events taken place. Property price has a positive impact on tourist arrivals, which is slightly beyond the expectation, since it is a proxy of price level, indicating that higher travelling expenses would keep tourists away. This might be explained by low price for China in the aggregated level, as discussed in the prices of a Big Mac hamburger, China is less costly comparing with the major source countries of international tourism. Therefore, Chinese cities are still affordable, though the price level keeps increasing. Moreover, those expensive cities (three hotspots – Beijing, Shanghai and Guangzhou in particular) are more famous and in rich of tourist products for international tourists, enabling them still willing to visit there. Hotel number is insignificant in all the models, thus can conclude that it has no impact on tourist arrivals. One reason might be its relatively small changes over the years. A city usually has hundreds of hotels, and there were about 10 hotels newly opened or closed in each year, after dividing by population size (on thousand basis), the magnitude of change becomes very small. All the coefficients need to be elaborated carefully. As Model 4 is the fully specified one, it is taken for the example. Keep in mind that some variables are corrected by population size, which will have impacts on the interpretation. FDI: For each year and each city, if FDI increases by one million US Dollar, tourist arrivals will increase 670 on average, ceteris paribus. ME: If there is a city holding a mega event in the certain year, there are 266 tourist arrivals than the year that no mega event is held, ceteris paribus. PP: For each year and each city, if property price increase by one thousand Chinese Yuan per square meter, tourist arrivals will increase by 8 on average, ceteris paribus. SD: For each year and each city, if emission of sulfur dioxide increases by one thousand tons, tourist arrivals will decrease by 0.125*population size, on average, ceteris paribus. MP: For each year and each city, if there is one thousand more webpages mentioned air pollution, tourist arrivals will decrease 1.089* population size, on average, ceteris paribus. Model 5 to Model 7 has checked the interaction between sulfur dioxide and media publicity, time-lagged effects of air pollution and time–lagged effects of media publicity. Table 12 presents the results. The result from model 4 is presented again to make comparison easier. Table 12: Results from Model 5 to Model 7
§Source: own elaboration Note: * Significant at 10% level, ** Significant at 5% level, *** Significant at 1% level Model 5 confirmed that there is no statistically significant impact of interaction between sulfur dioxide and media publicity, indicating media is not a moderate variable. The other variables stayed almost the same as Model 4, which is another support of no impact of interaction. Model 6 showed that both sulfur dioxide and media publicity in the previous year have significant influence on tourist arrivals in the current year, but media publicity in the previous year receives less statistical support comparing with the others. After adding both the sulfur dioxide and media publicity in the present year back to the model, the coefficients of SD, SD_1, MP and MP_1 have become all insignificant. This is resulted from multicollinearity problem. Table 13 shows that the correlation between SD and SD_1 is 0.9896, between MP and MP_1 is 0.9795, indicating Model 7 is a high collinear model. In this case, either the data from the current year or the previous year should be dropped. Furthermore, this correlation table warns the reliability of the result of the other models, because in this highly correlated model case, it cannot be distinguished that which one (previous/current year) leads to the impact, and it decreases the confidence of the estimations. However, there is barely method to solve this problem, which implies further research is needed. Table 13 Correlation of variables
Source: own elaboration 5.1.2 Comparisons between different regionsIn Section 2.2.3, the uneven geographical distribution of international tourism in China has been identified. In short, three hotspot cities attracted more tourists than the others, and the east regions attracted more tourists than the west. Therefore, it is also meaningful to check if air pollution has produced different effects in different geographical locations. The fully specified model has been applied to the investigation for this s section. Note that the correlation problem of variables is not mitigated yet, which might still defect the confidence of these results. Three hotspots vs. other regions It is clear that three hotspots (Beijing, Shanghai and Guangzhou) received more tourists than others. However it is difficult to formulate a regression with only three cities because of lack of observations (only 24). Instead, a regression excluded these three cities is built with 210 observations. The result of regression is showed in following Table 13, the original result of Model 4 is also showed for comparison. Table 13: Results of in/excluding three hotspot cities
Source: own elaboration Note: * Significant at 10% level, ** Significant at 5% level, *** Significant at 1% level
Unequal distribution of wealth was an important reason resulting difference of international tourism between eastern and western China. The border of east and west is drawn according to Figure 2, the cities in white area are included in the regression of eastern China, and the cities in grey area are included in the regression of western China. As a result, in eastern regions, there are 156 observations used for regression, while western are 76. Comparable results of regression models are presented in Table 14. Table 14: Results of eastern and western China
Source: own elaboration Note: * Significant at 10% level, ** Significant at 5% level, *** Significant at 1% level There are several differences have to be pointed out from this table of result. Regarding significance perspective, eastern China has almost the same result gained from the estimation with the complete dataset. ME lost its significance, and hotel gained a little significance. The reasons behind it might be that in this smaller size of the dataset, the differences between cities in the east are also smaller. The estimation with only data from western cities has very different result from the other two, and most importantly, SD and MP are not significant any more. But this result is short of reliability, and the critical reason behind is lack of observations. This reason also can be applied to Hotel, which has a quite surprising significance in the estimation of west. Additionally, ME is omitted again since there is no mega events held in western China. From the perspective of magnitude, first, FDI has less magnitude in the east than west. It might because of rich tourist products from the east, attracting tourist visiting for various reasons other than FDI. The magnitude of Hotel changed a lot, which might result from discredited reliability. Directory: pub pub -> Project Scoring Template Pre-Construction Phase New Core/Branch Campus Projects pub -> Acm word Template for sig site pub -> The german unification, 1815-1870 pub -> Baltic Olympiads in Informatics: Challenges for Training Together pub -> Preparation of Papers for ieee transactions on medical imaging pub -> Forthcoming meetings and conferences pub -> Asilomar Conference on Circuits, Systems & Computers, Pacific Grove, ca, November 1978, pp. 55 58 pub -> Forthcoming meetings and conferences pub -> Harmonised compatibility and sharing conditions for video pmse in the 7 9 ghz frequency band, taking into account radar use Download 352.16 Kb. Share with your friends:
|