Invisible Cities

4.2.3 The independent variables

The following variables listed will be included into the estimation equations going to be made, and the details of these variables will be illustrated one by one.

• 
  Sulfur dioxide
  
• 
  Media publicity
  
• 
  Foreign direct investment
  
• 
  Hotels
  
• 
  Mega events
  
• 
  Property Price
  

Sulfur Dioxide

This is the most important independent variable because it measures air quality, confronting the essence of the main research question.
In Chapter 2, it indicated that the concentration of PM 2.5 can measure air quality as well, and many countries (U.S. and western European countries) have adopted it as a measurement, called AQI (air quality index), measuring how microgram of PM 2.5 concentrated in the air per cubic meter on average. However, AQI had not accepted by the official department of China Environment Protection as an indicator of air quality before 2014. Meanwhile, PM 2.5 is a relatively new concept to Chinese public, most of websites provide real time statistics for each city, and historical data is rare. A Chinese website² provides historical data from October 2013, but no data available from 2005 to 2012, indicating that there is not sufficient data to sustain a reliable result of estimation. Moreover, the official measurement of air quality in China was API, abbreviated for air pollution index; the difference between it and AQI is that API is measuring the concentration of PM 10. Some historical data of API can be pursuit in the website of Ministry of Environmental Protection of PRC³, but lack of completeness for the entire needed time span (2005 to 2012) will deteriorate the reliability of estimation, thus it is not adopted as well, unfortunately.
Finally, the annual emission of sulfur dioxide on thousand tons basis for each prefecture city was found in CEIC database, thus it is adopted in this empirical model as the indicator of air quality. Sulfate, mostly sulfur dioxide, which has been discussed in Section 2.5.1, is an important composition of air pollutant PM 2.5. It can be a proxy of measurement because they are in the same direction, more sulfur dioxide probably causes more PM 2.5 generation. Comparing with PM 2.5, this proxy might have less visible to the public, since it does not directly reported by media. It might lead the influence of media publicity is overestimated.
Still, the emission of sulfur dioxide is an appropriate proxy, since sulfur dioxide influences the air quality. It thus will also influence tourists’ behavior and decisions, since it is a threat to human’s health, e.g. increase the probability of CVD. Similarly to PM 2.5, WHO also set up guidelines for sulfur dioxide, which duration of exposing to sulfur dioxide more than 500 micro gram per cubic meter on average should not exceed 10 minutes (World Health Organization, 2006).
There is no need for city-size correction because it is not definite that bigger city will generate more sulfur dioxide based on the evidence found about PM 2.5, the cities in north are more polluted than the south, but both north and south have big cities.

In addition, since time lagged effect of air pollution is also going to be investigated, the one-year lagged value of sulfur dioxide will be also included as a independent variable.

Confronting with the second hypothesis, media publicity of air pollution has to be included. There is no definite way to measure this variable, or quantify it. In this research, it is assumed that more media publicity will generate more impact on tourists, so the number of information available would be an appropriate method for measurement. Therefore, certain terms were searched in the Internet searching engine Bing.com, and the total numbers of results found were recorded. The terms used to search were following the structure as “City name +’air pollution’ + year”, for example:

Third, the Internet is just one of the mass media channels, other information sources, such as the local newspapers, are excluded from the results. However, newspapers is much more reachable for aged people than the Internet.

It is an indicator referring to market ties. In the previous context we measured it by GDP. FDI and GDP are all economic indicators, which can be used as proxy of market ties. Figure 6 depicts GDP (in billion Chinese Yuan) and FDI (in million US dollar) for all Chinese prefecture level cities from 2005 to 2012. It is clear that they are positively correlated, which suggests that only one of these two indicators can be included in the estimation model.

Accommodation is always important of travelling arrangement. It has been discussed that international visitors prefer qualified hotels, which indicates that 4-star or 5-stars hotel should be taken into account by measuring international tourism demand. However when looking for the data of 4 and 5-star hotels, there are still some values missing. In CEIC China Premium Database, the number of all registered hotel in the city is found, thus it is adopted as a proxy of qualified accommodation. It is also corrected by the city size (population).

As stated in Section 2.2.3, China started to participate holding mega events. To control the impact of mega event on tourism demand, this is set up as a dummy variable. If the city held a mega event in the year, then its value equal to 1, otherwise equal to 0. There are two mega eents took place in Chinese cities in time span from 2005 to 2012, which are Beijing Olympics 2008 and Shanghai Expo 2010. Value for Beijing in the year of 2008 and Shanghai in the year of 2010 equal to 1, all the rest cities in the rest years equal to 0.

Price level of the destination also has impact on tourism demand, because the basis of demand and supply relationship suggests a higher price leads to a lower quantity. An expensive destination, to some extent, probably will keep tourists away, and vice versa.

The best variable for the price level of tourism is Consumer Price Index (CPI), because tourism activities covers a wide range of consumer products, and CPI is the one that comprehensively evaluated the price level in a certain place. It is a surprise that a comprehensive CPI is difficult to find. CEIC provides retail prices for agricultural, food, industrial, consumer goods, and service charges in very details (e.g. daily price of eggs per kilo), but they are average values in the country level. The only prefectural level data is only available for property price, so it is taken as a proxy of consumer price level. It is counted based on Chinese Yuan per square meter.
Property price can be relevant to tourists’ expenditure, because hotel price is largely influenced since they are both related to real estate sector. Accommodation expenditure also has great proportion in total travel expenses, so property price can partly represent how expensive a destination is. The limitation of this proxy is obvious, it ignores the prices of other tourist products. However it is still difficult to include every product the tourists might consume, so another two representatives are picked up. The retail price of egg is an indicator of normal food expense, and the price of taxi per kilometer is an indicator of service/transportation expense. Figure 7 illustrates the changes of price level for property, egg and taxi in Chinese cities from 2005 to 2012(price in Jan 2008 equals to 1). It shows that property price has a greater increase than the other two products, and taxi price almost remains stable over years. Thus taking property price as a proxy of price level may lead to overestimation of the impact of price level on tourism demand.

Figure 7: Price changes of property, egg and taxi over years

Exclusion

Likewise, the following variables are excluded from this research; the reasons will also be explained for each of them.

• 
  Trade-weighted effective exchange rate index (BIS)
  
• 
  Number of heritages
  
• 
  Transportation cost
  
• 
  Airport ranking
  
• 
  Income of tourists from origins
  

Trade-weighted effective exchange rate index (BIS)

BIS is an overall exchange rate index that complies weighted average of exchange rate of home currency against foreign currencies with the weight of each foreign country equal to its share in trade. Exchange rate between Chinese Yuan and other currencies is another determinants of how expensive for travelling in Chinese cities, which can be seen as a composition of price level, indicating an impact on tourism demand.
In fact, the exchange rate has changes quite an extent in these years, due to the financial crisis in western countries and strong performance of Chinese economy. Table 8 illustrated the changes of BIS over years, which can be found a trend of appreciation of Chinese Yuan.
Table 9: BIS of Chinese Yuan from 2005 to 2012 (2010=100)

Year	2005	2006	2007	2008	2009	2010	2011	2012
BIS	87.2467	89.24	90.38197	96.37333	102.0108	100	100.0833	105.6792

Source: CEIC China Premium Database (2014)
The reason of dropping it is that BIS is an universal number for every city. Although it changes over years, but in this sense it can only explain the changes in the number of visitors in different years, which occurs in all cities. Meanwhile, as time-demeaned variables are added, which already have caught characteristic differs over the years but remains the same for every city; BIS as such a variable is not necessarily to be added into the model.

Number of heritages

Cultural and natural heritages are important attractions for international tourists. It is not picked into the model because the number of heritages keeps constant over time, because they cannot be created. In panel data estimation, this type of effect is automatically omitted because the impact does not change.
Transportation cost

Lower transport cost leads to an expansion of tourism – it might be true for short distance international travelling, like a city break travel in European countries. For example, the introduction of low-cost airline in Alghero significantly boosted the number of tourists, because of reduction of transportation cost resulted from greater accessibility (Pulina & Cortes-Jimenez, 2010). However, it is inferred not applicable for travelling to Chinese cities, because expense on air transport is always generating the most transportation cost, and the ticket prices from Europe or US to China do not differ too much. Another reason to exclude it is the difficulties of data collection, because ticket prices vary in a large range, and depend on many factors, e.g. booking time prior to travel date, the airline company selected, the travel agency contacted.

Since accessibility is one of the major determinants of destination attractiveness, the existence of international airport and how many passengers it receives seem that need to be taken into account.

Subsequently, even if you could construct a variable that exactly measures the number of destinations reachable from an airport, it would probably suffer from endogeneity. For example, if a city attracts many tourists in the previous year, its airport is likely to have a bigger number of destinations in the next year to cope with the high number of incoming tourists. Endogeneity is a problem of causation, it keeps unknown that whether an accessible airport causes more visitors to come to the city, or whether more visitors coming to the city causes the airport to become more accessible by having an increased number of connections/destinations. Therefore, it is very difficult to add a variable for transport into the estimation.

In the previous discussions, the property prices of Chinese cities are included as a sign of travel expenses in China, which will matter for destination selection. If the expenses of travelling to Chinese cities are far more than the price they can afford, then tourists will visit other places instead. The affordability is certainly linked to income, in this case, is income of the tourists from the origins. However, since the origins cover all over the world, the types of inbound tourists are diversified, and the proportion of different types varies across different years, and again very detailed data is needed because 31 Chinese cities received different numbers of tourists. Therefore, it is very difficult to find an appropriate proxy to measure their incomes, or to find a suitable weighted factor to leverage all those incomes in one variable. In order to avoid disturbance of model reliability, this variable have to be left out.