4.
Data
We collected a unique dataset of Chinese automobile dealers from the Autohome website ( www.autohome.com.cn ). Our dataset comprises information on 99 auto brands and 11,761 dealers. It includes 19,310 brand-dealer pairs in 329 Chinese cities, including such major cities as Beijing, Shanghai, and Tianjin. The State Council published The Notice of the State Council on Adjusting the Standards for Categorizing City Sizes on Oct. 29, 2014. By this notice, the Chinese cities are cate- gorized into five different levels by their population. We use the same population intervals as used for city categorization to summarize the city information, such as average per capita income and population, in Table 4 . Over 60% of the cities are medium size, while 30% of the cities are large and cities with populations over 5 million. Apparently, the average per capita incomes are higher in larger cities. On average, there are around 59 brand-dealer pairs in each city. However, the distribution in the number of brand-dealer pairs is highly skewed to the large cities: there are 198 pairs on average in super cities but less than 7 pairs on average in tiny cities. Such asymmetric distribution across cities indicates that heterogeneous market conditions such as aggregate demand and income level could be important determinants for brands’ and dealers’ decisions on their exclusivity commitment. The variables measuring heterogeneous market conditions consist of two sub-categories: macroeconomic conditions, which are invariant over manufacturers, and operation cost shifters, which are variant over manufacturers. Macroeconomic conditions include the population in 10 thousand, pop , and average per capita income in 10 thousand, income , of each mar- ket (city), and are collected from the 2011 Statistical Yearbook published by the National Bureau of Statistics of China. 10 In our empirical analysis, we use their logarithmic form. The operation cost shifters refer to those factors that will change the manufacturers’ operation costs. We use the logarithmic term of distance, log ( distance ), from manufacturers to their target markets to measure this for two reasons: first, manufacturers are less familiar with the markets further away from them and so the costs of both market survey and entry could be higher; second, both transportation costs and monitor costs are higher for markets further away from the manufacturers. To obtain the distance data, we check the distance between a manufacturer’s headquarters 11 and the location of the municipal government of each market through the most recognized online map system in China –Baidu maps ( maps.baidu.com ). This map will provide three different types of paths and asso-