2. Literature review, theoretical analysis and hypothesis development
RPE theory is logical and widely applicable. In the case of common risk, using the RPE method to choose a
peer group can effectively extract the individual effort of managers and mitigate agency problems. The theory
and models of RPE suggest that it is necessary to exclude the combined effect of peer groups (Baiman and
Demski, 1980; Diamond and Verrecchia, 1982; Holmstrom, 1979, 1982). Antle and Smith (1986) and Gibbons
and Murphy (1990) show that the benefit of adding relative performance considerations to executive compensation
contracts is greater than the cost.
However, theoretical expectations have not received consistent support from empirical studies. Antle and
Smith (1986) examine the relationship between CEO compensation and industry returns from 1947 to 1977.
Using different definitions of compensation, they only find weak support for RPE. Gibbons and Murphy
(1990) examine the relationship between change in compensation and firm stock returns for 1668 CEOs from
1974 to 1986. Their results show that changes in compensation are negatively related to industry and market
performance. They also find evidence that RPE is related to CEO turnover. Jensen and Murphy (1990) find
that changes in CEO compensation are positively related to changes in shareholder wealth. However, they find
no significant relationships between change in compensation and change in net-of-industry wealth or net-ofmarket
wealth. Janakiraman et al. (1992) examine 609 companies from 1970 to 1988. They find weak evidence
in support of weak-form RPE, but no evidence in support of strong-form RPE. Aggarwal and Samwick
(1999b) examine stock returns from 1993 to 1996 and use several methods, including median regressions
128 D. Chen et al. / China Journal of Accounting Research 5 (2012) 127–144
and OLS regressions, but find no systematic support for RPE. Aggarwal and Samwick (1999a) find some support
for RPE using short-term compensation, but long-term compensation increases with industry performance.
Garvey and Milbourn (2003) examine the relationship between CEO compensation and stock
returns and find no support for RPE, except in companies where managers are younger and have less financial
wealth.
To resolve such conflicts, researchers have offered possible explanations for the lack of RPE. On the one
hand, constraint conditions will limit the use of RPE. For example, Aggarwal and Samwick (1999a) point
out that the degree of competition might be an important factor in using RPE. On the other hand, the effi-
ciency of compensation contracts is also in doubt. For example, Bebchuk and Fried (2003) find that in the
case of poor corporate governance, CEOs can influence their own pay through their control over the design
of CEO compensation. This will reduce the sensitivity of compensation and also lead to a lack of RPE.
There have been few empirical studies of RPE in China. Xiao (2005) tests the strong- and weak-form RPE
in a sample of listed companies in 2002. Using the average ROE of peer groups in the same region and the
average ROE by 2-digit industry code, they find evidence of weak-form, but not strong-form RPE. Gao
(2006) uses a sample of all A-share listed companies from 2001 to 2004 to test the theory. He finds that
RPE exists when the peer group is comprised of similar industry-size firms or the same industry-ownership
firms. However, the article neither explains the theoretical base for the composition of peer groups, nor compares
companies with different ownership types. Zhou and Zhang (2010) examine the effect of RPE in listed
companies in China from 1999 to 2006 using their comprehensive index of performance. They find that RPE
exists when peer groups are based on area, but find opposite effects when peer groups are based on industry or
size. However, the comprehensive index of performance designed by the authors is a subjective measure.
In recent years, there has been substantial progress in the study of RPE. Albuquerque (2009) proposes a
new theory and method for selecting peer groups. He regards size as the most important factor affecting outside
risks at the company level. For example, large companies often face lower financing constraints because
they are more likely to survive in the event of negative shocks. Additionally, size is often related to diversity
and the degree of diversity can affect companies’ risk tolerance. His empirical results support RPE theory.
Carter et al. (2009) hand-collect information on peer groups in FTSE-listed companies in the UK. They find
that the probability of using RPE does not increase when systemic risk increases, whereas external monitoring
is an important factor for using RPE. Faulkender and Yang (2010) hand-collect information on peer groups in
compensation plans in the US. Their results support RPE theory and their direct method has advantages over
traditional methods, such as the industry-size peer group method. In addition, they find that companies prefer
to put companies with high pay into their peer groups. Gong et al. (2011) also use manually collected information
on peer groups in compensation plans in the US to test RPE theory. They find that 25% of listed companies
explicitly use peer groups and the choice of peer groups supports a mixture of effective contract theory
and rent-seeking theory.
In our opinion, it is essential to study RPE theory in China. First, studies on RPE in China are still rare and
those that exist are mainly normative studies. Second, the existing empirical studies have produced mixed
results (Xiao, 2005; Gao, 2006; Zhou and Zhang, 2010), thus improving RPE methodology is important.
Third, we believe that the nature of ownership is an important factor to be considered in studying RPE, which
is not generally considered either in China or elsewhere1
.
There are various differences in the way executive compensation is designed and evaluated for SOEs and
non-SOEs. First, there are regulations on cash compensation in Chinese SOEs. In 2002 and 2009, for example,
the SAC set multiple limits for executive cash compensation in SOEs.2 However, there is no limit in non-SOEs,
so executive compensation in non-SOEs is more market-oriented. Chen et al. (2005a) provide evidence that
executives’ relative pay in SOEs is far less than in non-SOEs.3 Second, there are various forms of incentivesin SOEs. For example, Chen et al. (2005a) find that perks were widely used as an incentive. Cao et al. (2010),
Chen et al. (2011a) and Wang and Xiao (2011) argue that the opportunity for promotion of executives in
SOEs is also an important incentive. However, such incentives are rare in non-SOEs and cash compensation
is dominant.4 Third, SOEs are affected by multiple tasks. Bai and Xu (2005) and Bai et al. (2006) find that
executives in SOEs undertake diverse tasks. Executives in SOEs not only need to improve performance, but
also need to consider other issues, such as the employment of workers. However, non-SOEs are often subject
to less government intervention (Chen et al., 2011b), so performance is likely to be more closely related to
executive evaluation.
The above differences may cause RPE to be applied differently in SOEs and non-SOEs. First, the regulation
of executive cash compensation is expected to reduce the effect of RPE. For example, when cash compensation
is close to the limit, even if relative performance is high, executives may not work harder, thus RPE is not
effective. Second, the various forms of incentives in SOEs will reduce the benefit of RPE, because cash compensation
is just one type of incentive in SOEs. Third, multiple tasks will obscure the relationship between firm
performance and executive effort, which will increase the implementation cost of RPE. For example, the performance
in company A’s financial statements is lower than in company B’s, but company A undertakes a lot
of redundancies. Is the performance of company A better or worse than that of company B? However, in nonSOEs,
there is no regulation of executive cash compensation, there is only one form of incentive and there is
less intervention from multiple governmental goals, thus implementation of RPE will be easier and the net
effects of RPE will be more obvious.
In summary, we believe that RPE is more likely to be applied in non-SOEs than in SOEs.