• No results found

Chapter 2. The pattern of industrial agglomeration and changes Evidence from China

5. Data Description

In this chapter, we utilise the Industrial Enterprises Dataset collected by the National Bureau of Statistics of China (NBS dataset) which has been widely applied by other researchers. It includes all of the state-owned enterprises (SOEs) and non-state-owned enterprises (groups, limited companies, joint stock enterprises, Hong Kong, Taiwan and Macau-investment enterprises (HTM) and FDI enterprises(overseas)) with a turnover larger than 5 million RMB. The dataset covers over 40 2-digit, 90 3-digit and 600 4-digit industries. The total GDP of the enterprises in the dataset accounts for over 90% of the whole GDP of the manufacturing industries in China. It is the most comprehensive annual data at the firm level in China. Two variables are required to calculate the MS index – the number of industries and the number of regions. We clean the data and aim to maximise the number of observations. Previous studies on the patterns of industrial agglomeration are generally based on the rankings of EG index values at the 4-digit and 2-digit industry level. We too have followed this approach in our study so we can make easy international comparisons with previous works. GB/T 4754_1994 and improved GB/T 4754_2002 (also called SIC) are the official industry classifications set up by the National Bureau of Statistics of China in 1994 and 2002, providing the coding system of ISIC (International Standard Industrial Classification). These coding systems were officially applied in 1995 and 2003. China, as a very dynamic economy, adjusts its industry classification over time. The adjustment is consistent with the development of economic growth in China. Some emerging and new established industries in China have also been included on the SIC (Standard Industry Classification). The most recent version of SIC GB/T 4754-2011 was applied in 2011. Since the restructuring of the industry coding system by NBS in China in 2002, it is difficult to compare value changes of the MS index from 1998 to 2007. In the previous study, most of the studies focused at the 3-digit industry level, which is the most detailed level to overcome the recoding issue. Lin et al. (2011) discussed agglomeration at the 4-digit industry only focusing on the textile industry. Although Brandt et al.(2014) give the concordance tables for China, it does not include natural resources and energy and water supply industries. Therefore, we

30 / 346

prefer to look at changes of the MS index of each 4-digit industry within the time period of the same coding system. That gives us two time periods: from 1998 to 2002 and from 2003 to 2007. Doing so offers detailed information on the changes of industrial agglomeration in this emerging economy, ruling out changes in coding. In general, most of the 4-digit, 3-digit and 2-digit industries are identical over all time periods. However, some merge and recoding creates differences for some industries at various industry levels.

We first target our time period of analysis. The number of 4-digit industries before 1998 varied significantly year by year in this dataset. It is therefore difficult to include earlier years at the 4-digit industry level. Since we aim to calculate the value of the MS index at the most detailed industry level – the 4-digit industry level – our starting year is 1998. On the other hand, although we also have data for 2008 and 2009, the missing 4-digit industries make datasets incomplete for both years. Hence the ending year of our analysis is 2007. After that, as the dataset includes inactive enterprises, we need to exclude those enterprises not still in operation. There are seven types of enterprises business status defined by NBS5. The definition of business status "Open" relates to enterprises that have been in production for over three months in the year of the survey.

"Close down" designates those enterprises that have been in production for less than three months or were completely out of business in the survey year but still retain their business license6. "Preparation" refers to an enterprise that is still in the starting process or in trial operations; we also treat enterprises with business status "Preparation" that have positive turnover as enterprises in business and include them in the dataset. This judgement is important when we examine the dynamic of agglomeration using survival analysis. We only keep those enterprises that are in business activity. Finally, we keep

5 According to the type of enterprise status defined by NBS, enterprises can be classified as “Open”,

“Close down”, “Preparation”, “Cancel”, “Close down in the statistic year”, “Bankruptcy in the statistic year”, and “Others”.

6 The Business license is applied by the Chinese government to register different type of business units in China. The license is free for any person or legal person. The type of business needs to satisfy the amount of capital investment and related law. In the situation of business "Close down", the business unit must either have suffered financial crisis or legal violation. The business unit is allowed to keep its license while solving its financial or legal issues. Therefore, the "Close down" firms can be reopened in the next year although the business status shows "Close down" in the previous year.

31 / 346

4-digit industries with over six enterprises in each year and the duration is five years.

In the judgement of plants status, we treat plants with business status "Open" and

"Preparation" together with turnover greater than zero as inactive and keep them in the dataset while the rest are excluded. We assume "t" as the survey year, then the "Entrance"

category means that the enterprise did not appear in t-1 but appears in t; the "Exitor"

category means that enterprises appear in year t but not in t+1; the "One-year" category represents enterprises appearing in t but not in t-1 nor t+1; the "Survivor" category represents enterprises that appear in both t and t+1. Note that if a plant had a turnover greater than zero and its business status was "Preparation" in t-1, we do not treat it as

"New entry" if its status is "Open" in t. We also follow Devereux et al.’s (2004) definite

"job creation" and "job destruction" factors. The total number of new employment positions is defined as the number of workers employed by new entrants together with the additional number of workers for plants that increased employment compared with the previous year. Job creation is defined as the total new employment over total employment. Job destruction is also estimated as the total reduction in employment over total employment where total reduction in employment is defined as the total number of employees of the "Exitors" together with the total reduction of employment for those plants who reduce their workforce.

Some cleaning of the geographical data was also required. There are two regional code systems in China: the administration division code and postcode. The administrative division code is officially called the ISO-3166-2: CN. It uses a 12-digit code to include all urban and rural regions in China. The NBS update the code of several regions every year.

The first two digits represent the province, municipalities, autonomous region (Xinjiang, Tibet etc.) and Special Administrative Regions (Hong Kong and Macau). The third and fourth digits of the code represent the associated city, prefecture, autonomous prefecture, Mongolian league, municipal city and county regions. The fifth and sixth code represents city districts, county-level cities and the banner areas of the Inner Mongolia region level.

The seventh to ninth digit is the township in the rural area or street in the city. The last

32 / 346

three digits give the village in a rural area or community area in a city. We aggregate the urban and rural areas of the four municipalities and treat them as one city (equal to 4-digit administration division). Note that the administrative division is based on political enquiries; therefore, the area across divisions may vary due to historical, cultural background and other reasons. We also applied postcode regions for the general industrial agglomeration in China to compare the results with those based on administrative division classifications. These comparisons are also applied at postcode level in international comparisons.

In a previous study, Lu and Tao (2009) merged the regions based on the administration division codes from 1998 to 2005. Although there were some code changes for certain county areas during the period, they still merged the county areas and used those areas each year from 1998 to 2005. However, it may still lead to recent developed county areas being missed where they were recorded into administration divisions. China has set up additional county areas since 1998 and the newly developed regions may be important to trace. We would like to analyse the changes for the whole of China on the pattern of industrial agglomeration, not just those regions recorded for many years. We, therefore, keep all the enterprises located in 31 provinces in mainland China to maximise our dataset, which also helps us to trace the industrial agglomeration changes in China. Since the MS index gives smaller weight to regions with fewer production activities, the newly established regions will have a limited impact on the value of the MS index. Rapid growth in China has resulted in a dynamic map of industrial geographic distribution.

There have been newly established manufacturing industry zones and large-scale entry in the traditional industry cluster regions every year. Hence our analysis would be biased if we only keep traditional industry clustered regions, dropping the new regions that appear in the dataset. Bai et al. (2004) investigate the local protectionism within administration divisions in China. Therefore, we calculate the MS index at different administration division codes to capture if local protectionism also affects the degree of changes in the MS index over time.

33 / 346

Finally, following Devereux et al. (2004) with the assumption that the plant location decision is independent, enterprises owned by the same owners, located in the same place and producing the same products cannot be treated as different enterprises. Hence, we combine two or more enterprises into one "firm" by aggregating their number of employees and other main indicators in the descriptive part.

In summary, our data cleaning process enabled us to achieve our three main aims: to study the trends of industrial agglomeration at a very detailed industry level; to make international comparisons at the same industry level and various region levels; and to study the regional development of the manufacturing industries over time.