3.3 Data
3.3.3 Sample description
3.3.3.1 Descriptive statistics
31This study defines top officers or top 10 shareholders as being politically connected if they were currently
The sample in this study is finalised as consisting of 381 PCFs on the Chinese stock market. This study has matched connected firms with non-connected firms one by one. The sample includes 8387 firm-year observations, covering the period from 1992 to 2011. The data used in this paper include information on accounting values, stock price and corporate governance. This study obtained data on accounting, corporate governance and stock price for both connected and non-connected firms from the Chinese Stock Market and the Accounting Research (CSMAR) database provided by the Shenzhen GTA Information Technology Corporation.
Table 3.1 presents the descriptive statistics for the connected firms in this study; and Panel A provides the IPOs of these firms. The sample period is from 1992 to 2011, which largely reflects the overall IPO pattern in China. Two decades ago there were only 40 newly listed stocks, and then the number rapidly increased to 209 in 1997. Even though PCFs are unevenly distributed across the sample period, the number of connected firms increased from 1992 to 1996, but remained around 20 from after 1996 until 2005, when it fell to 2. Note that in 2005, there were only 15 newly listed stocks. The abnormally low level of IPO activity in 2005 reflects the impact of several regulatory actions that restricted the flow of IPOs to the market32. The Chinese stock market had undergone substantial changes in 2005. One of them is the NTS (non-tradable share) reform, which gradually makes all non-tradable shares become tradable. Before 2005, the Chinese government still partially controlled share issue privatisation (SIP) firms, which impacts on the market’s proper development and expansion (People’s Daily, June 28, 2005). As a result, China has started a split-share structuring reform (referred to as the NTS reform), thereby reducing IPO numbers. Therefore, the sample period is divided into two phases: 1992 to 2004; and 2005 onwards. This is one of the reasons for the robustness tests on political connections by these two sub-periods in Section 3.6.2.
Panel B reports a particular pattern of cross-industry variation among PCFs. The Industrial category has the highest percentage of PCFs (57.74%), followed by Conglomerates (23.62%), Public Utilities (7.61%), Commerce (6.56%), and Properties (4.46%). Moreover, Panel C breaks down the subsamples of Industrial firms. Of the total 57.74% that are
32From time to time, market observers have identified some problems with China’s IPO system, such as
underpricing caused by fixed P-E ratios or issuance quotas. To overcome these problems the CRSC has issued many corresponding reforms. One such freeze was implemented in 2005, lasting from May 2005 to June 2006.
Industrial firms, food, medicine and chemical accounts for nearly 15.49%; this figure is followed by electrical manufacturing and transportation & power, which account for 13.39% and 12.34%, respectively.
Table 3:1:The sample Panel A: By year
listing Year No. of connected firm’s IPO Total No. of A- share's IPO % of total A- share IPOs 1992 5 40 12.5% 1993 13 129 10.1% 1994 18 107 16.8% 1995 6 28 21.4% 1996 34 206 16.5% 1997 39 209 18.7% 1998 24 104 23.1% 1999 16 96 16.7% 2000 27 131 20.6% 2001 15 75 20.0% 2002 16 71 22.5% 2003 21 67 31.3% 2004 21 100 21.0% 2005 2 15 13.3% 2006 14 66 21.2% 2007 25 126 19.8% 2008 12 77 15.6% 2009 21 99 21.2% 2010 27 349 7.7% 2011 20 282 7.1% 2012 5 155 3.2% total 381 2532 15.0%
Panel B: By industry classification A % of total connected firms
Public utility 29 7.61
Properties 17 4.46
Conglomerates 90 23.62
Commerce 25 6.56
Total 381 100
Panel C: % of total connected firms
Food, medicine & chemical 59 15.49
Electrical manufacturing 51 13.39
Transportation & power manufacturing 47 12.34
Metallic manufacturing 25 6.56
Culture, office &education goods 21 5.51
Oil, rubber &fibre 17 4.46
Total 220 57.74
3.3.3.2 Matching procedure
In much of the analysis, this study compares the operating performance of PCFs to a set of matching non-connected peers. Faccio (2010) matches connected samples by non- connected companies only on the basis of market equity capitalisation (within ±40%). However, the matching approach in this study is far more stringent. It matches one non- connected firm with each connected firm by the following rules.
a) Industry: According to the CSMAR database, there are two types of industry classifications; industry A, and industry B. The industry A classification has 6 broad categories (Commerce, Conglomerates, Finance, Industrials, Properties and Public Utility), while industry B is subcategorised into 175 categories based on industry A. This study divides the set of political firms into 80 sub-sectors in terms of industry B. A potential match should first come from the same division in industry B. If no company meets this criterion, this study broadly selects a non-connected firm from industry A.
b) Firm size: Based on the first step, a potential match from the same sector is then selected if its book value of total assets is closest to that of the connected firm at the end of the first year after the IPO.
c) Listing year: This study selects a match if its listing year is as close as possible to the listing year of the connected firm.
d) Status: CSMAR categorises firms into 4 statuses. A represents normal trading. S is suspension; D means termination; and lastly, N is delisting. As all of the connected
firms have the status of normal trading, this study selects the matching firms from normal trading as well.
A few caveats are required. First, this study follows the usual practice of excluding financial companies (subcategorised into securities, banks and insurance) because of their unique financial structure, regulatory requirements, and accounting standards. This distinguishes the current study from Faccio (2006), who includes the banking sector in her samples to study the propensity of bailout. Second, matching was done without duplication, which is consistent with Faccio, et al. (Faccio et al., 2006).If more than one firm meets the above requirements at the same time, this study selects the one whose book value of total assets is closest to that of the connected firm.
Regardless of the best efforts to match connected firms with non-connected firms, the matching approach might have limitations in the quality of judgements. One potential problem that should be pointed out is that matching firms can be potentially connected with politicians in ways that escape detection by the search algorithm. On the other hand, it is possible that political connections may decrease or increase within the sample period given that this study cannot track changes in PCFs over time. These may diminish the strength of any findings of differences between the samples. Nevertheless, the results in this study show that the PCFs are systematically different in their accounting balances indicating business performance and wealth distribution in contrast to matching firms.