Invisible Cities
Download 352.16 Kb.
|
- Navigate this page:
- Chapter 6 Discussion
- 6.1 Effects of air pollution
5.2 Goodness-of-fitHausman test This test is conducted to detect systematic difference of the coefficients between fixed effect and random effect in order to select an appropriate method to analyze this panel database. The result of Hausman test (See Appendix 1) shows a p-value 0.000, suggests that differences between fixed effect and random effect model are systematic. Thus, the fixed effect model is adopted – all the regressions showed in the above section are estimated based on this method. Heteroskedasticity of residuals Another assumption of OLS regression is that the residuals are homoskedastic. Breush-Pagan test is a typical one to detect it. However BP test is not applicable for panel dataset, as this dissertation is not statistics dominant, the heteroskedasticity of residuals has to be left for further research. Normality of residuals Assumption of normality of residuals of Model 4 (fully specified model) has been also investigated. In Figure 8, Kernel density estimate graph depicts that the middle of residuals are almost following normal distribution, but some residuals in the tails are not, which might slightly violent the assumption of normality. It is in line with P- and Q-plot, which are Figure 9 and 10, showing that there are some outliers in the tails. Figure 11 scattered the residuals, and the scatters in the upper right corner show clearly some extreme values existing in the current dataset. Figure 8: Kernel density estimate of residuals of Model 4 
Source: own elaboration Figure 9: P-Plot of residuals in Model 4 
Source: own elaboration Figure 10: Q-Plot of residuals in Model 4 
Source: own elaboration Figure 11: Scatter plot of residuals in Model 4 
Source: own elaboration There are indeed some extreme values caused this violence of normality in the tails. For example, tourist arrivals in Lhasa 2008 plunged considerably (from 223.7 to 46). It resulted from the terrorism attack happened in March, and then the visiting of tourist was almost banned in the next months. Similar case happened in Urumqi 2010. Figure 12 and 13 depict the abnormal drop of tourist arrivals in these two places. Figure 12: tourist arrivals (TA) of Lhasa 
Source: own elaboration Figure 13: tourist arrivals (TA) of Urumqi 
Source: own elaboration Outliers sometimes can largely change the estimation results, in this case, Lhasa and Urumqi are clearly the outliers suggested to be excluded in order to acquire a more reasonable regression. After dropping Lhasa, the dot in down left position has disappeared from the new Q-plot, showing in Figure 14. It implies eliminating extreme values will gradually correct abnormality of the residuals. Figure 14: Q-plot without Lhasa 
Source: own elaboration Furthermore, Model 4 is estimated again, the results do not differ much from the results we gained above (See Appendix 2). It suggests that the outliers are not problematic for the results of estimations already gained. Therefore, from the above analysis, it can be fairly concluded that the estimation results are reliable. These results presented above eventually are going to facilitate the tests of hypotheses proposed in Section 3.6. In the next chapter, the conclusions of the tests will be discussed in order to find the way to final conclusion of quantitative analysis of this dissertation. Chapter 6 DiscussionThis chapter is going to find conclusion of hypotheses made in Section 3.5 and attempt to further explain them. First it would be convenient to quickly look back to the four hypotheses, which are: H1: Lower air quality will lead to a decline on inbound tourism of Chinese cities. H1.1: Lower air quality will result in a decline of inbound tourism of Chinese cities one year later. H2: More media publicity of air pollution will lead to a decline on inbound tourism of Chinese cities. H2.1: More media publicity of air pollution in the previous year would lead to a decline of tourism in the current year. As a consequence, the following sections will answer these hypotheses separately. 6.1 Effects of air pollutionFrom the significant tests throughout the models contain variable of sulfur dioxide, it can be conclude that H1 is not rejected. In other words, it is confirmed that lower air quality had lead to a decline on inbound tourism of Chinese cities from 2005 to 2012. Regarding to H1.1, it is not rejected either, since the lagged effect model (Model 6) evidenced its significant effect on tourist arrivals. To look into the details, Model 2, 4 and 5 included the annual emission of sulfur dioxide in the present year. The magnitude of its impact keeps stable, which are -.000118, -.000125, and -.000106. Although they seem to be small numbers, the real decrease of the tourist arrivals depends on the population size, so the impact can be enlarged, especially in those cities with dense population. Using the coefficient in the fully specified model (Model 4), for instance, the population in Beijing in 2005 was 15380 thousand, so one thousand tons more sulfur dioxide emissions resulted 1923 less inbound tourists in Beijing. The coefficients of SD are both significant in 1% level in Model 2 and 4, demonstrated a strong support of its effect. In Model 5 it is only in 10% significant level. It was added an interactive variable with media publicity, which might takes the significance from SD. Regarding the time-lagged effect of the emission of sulfur dioxide, Model 6 has present a strong statistically support of the variable of SD in the previous year, which is significant in 1% level. It suggests H1.1 is not rejected, which the air pollution in the previous year will lead to a decline of inbound tourism in Chinese cities in the current year. However, a limitation has already identified that SD and SD_1 are highly collinear. Therefore, neither of the current effect nor time-lagged effect should be over interpreted, because it is unknown for which data caused this estimation result. But still, taking Beijing 2006 for example, the population was 16010, indicated that one thousand more emission of sulfur dioxide in 2005 lead to a decrease of tourist arrivals in 2006 by 2498 people. The magnitude of time lag impact is slightly larger, though the result is not confident enough, it still can be reasonable. Travelling to China usually takes a long time to prepare, due to visa application, unfamiliarity of Chinese language, and probably also financial difficulties, so it is quite possible that tourist need to decide whether going to China or not one year earlier than the departure date, and a high level of air pollution might let them quit this idea during this time period. 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:
|