Robust and fragile firm-specific determinants of the capital structure of Chinese firms

Abstract

We demonstrate, using Extreme Bounds Analysis (EBA), that some of the firm-specific determinants of the capital structure of Chinese firms reported as important in previous studies may be fragile, in the sense that the sign and/or significance of the coefficients on these variables change depending on model specification (the variables included in or excluded from the model). For this purpose, data on 344 publicly listed shareholding companies are used, covering nine potential firm-specific determinants of capital structure. The robust variables, whose coefficients remain significant and of the same sign irrespective of model specification, turn out to be size, liquidity, profitability and growth opportunities. Although conventional cross-sectional analysis would lend support to the importance of tangibility and stock price performance, these two variables are indeed fragile. Other variables turn out to be insignificant.

Keywords:

• extreme bounds analysis
• capital structure
• trade-off theory
• pecking order theory

JEL Classification:

• C21
• G32

Notes

¹ It is a common practice, particularly in the field of corporate finance, that researchers experiment with hundreds of regressions with various specifications and variable definitions, then they only report the ones that they like for some reason.

² Chen (2004) suggests that the negative relation between size and leverage can be explained in terms of the better access of large firms to equity finance because of their reputation in the markets and the attraction of capital gains in the secondary market. For the special case of China, it is also suggested that the negative relation is due to the fact that bankruptcy costs are low since the legal system is incomplete. These explanations are, however, less convincing than those of a positive relation between size and leverage. The empirical evidence is also overwhelmingly supportive of a positive relation. For theory and evidence, see Marsh (1982), Friend and Lang (1988), Rajan and Zingales (1995), Wald (1999), Antoniou et al. (2008) and De Jong et al. (2008).

³ For the econometrics of variable addition and deletion tests, see Pesaran and Pesaran (2009).