Appendix: Data Sources

This paper uses the firm-level annual survey data for PRC's large and medium-sized industrial enterprises during 1995-2002, which are collected and maintained at the National Bureau of Statistics (NBS) in Beijing. The firm data set allows us to compare enterprise performance across region, ownership, industry, and time. Fuller details on the dataset are in the Appendix of Xiao (2005).

The NBS survey covers more than 20,000 industrial enterprises classed as medium and large. There are some unusable observations due to incomplete data reporting or small enterprises, which were classified as large and medium-sized historically based on their design production capacity. The classification standard for the size of industrial enterprises was first issued in April 1988 by a number of government agencies including the State Planning Commission and the National Bureau of Statistics. It includes detailed specifications based on the measurement of the output quantity or capacity in technical terms, instead of in value. The original standard is a legacy of the centrally planned economy and is being phased out. It now only applies to state-owned enterprises. For private enterprises, the National Bureau of Statistics is using sales value as the only variable to determine the size classification of enterprises.

In this study, we work with the full NBS sample of medium and large enterprises. However observations meeting one of the following screening conditions are regarded as unusable and deleted from the sample:

1. Net value of fixed assets < RMB100,000;
2. Intermediate inputs < RMB100,000;
3. Number of employees < 30;
4. Gross value of industrial inputs at current price < RMB100,000;
5. Sales < RMB100,000;
6. Total assets < RMB100,000;
7. Total assets – liquid assets < 0;
8. Total assets – gross fixed assets < 0;
9. Total assets - net value of fixed assets < 0;
10. Accumulated depreciation – current depreciation < 0;
11. Significant missing data.

The unusable observations are evenly distributed across ownership, industry, and region. Hence, excluding them from the usable sample should not create much bias in our analysis. However the sample does not have the same population over time. As in principle it should cover the entire large and medium-sized industrial enterprises sector, as defined above, so enterprises that become smaller and no longer qualify for the group exit from the sample every year. This means that this is an unbalanced panel and there may be a risk of ‘survivor bias’ as we do not have data on enterprises that exit the sample.

The sample represents an important part of the economy. The value added of the sample enterprises is as high as 43.3% of total industrial value added in 2002 and 19.2% of GDP. Employment of the sample enterprises is 16.7% of total industrial employment. The total liabilities of the sample enterprises are 43.6% of total bank loans.

Enterprise data at current prices are deflated by price indices for output, fixed capital, and intermediate inputs. In the data analysis, we use an aggregate price index for fixed capital, but apply industry level price indices to deflate output and intermediate inputs.

Download this Discussion Paper [ PDF 168.5KB| 25 pages ].

Comment(s)

There are [0] comment(s) for this entry. Post a comment.