The effects of China’s split-share reform on firms’ capital structure choice

ABSTRACT

China’s split-share structure reform in 2005–2006 mitigates agency conflicts between controlling shareholders and minority shareholders and thus may bring substantial changes to corporate financing behaviour. This article examines the impact of that reform on the capital structure decisions of firms by applying a variety of trade-off and pecking-order models. Using data from 1176 non-financial Chinese listed firms during the period 2000–2012, we present empirical evidence indicating that equity tracks the financing deficit better than debt in Chinese firms, a finding which is not consistent with pecking-order theory. This phenomenon is more prominent after 2006 as share reform increases trading activity in the secondary stock market and improves the transparency of financial markets. In addition, Chinese firms have an optimal leverage ratio and they adjust below-target leverage ratios faster than above-target leverage ratios after the implementation of share structure reform, although they make symmetric adjustments towards the target leverage ratio before 2007. Finally, recent share reform has prompted Chinese firms to more quickly address the divergence of actual leverage ratios from long-term target levels, but has slowed their response to short-term target leverage divergence.

KEYWORDS:

• Split-share reform
• capital structure
• trade-off theory
• pecking-order theory
• dynamic target adjustment models

JEL CLASSIFICATION:

• G30
• G32

Disclosure statement

No potential conflict of interest was reported by the authors.

Notes

¹ More details about split-share reform are introduced in Section II.

² A brief literature review on capital structures of Chinese companies is given in Section III.

³ Internal monitoring was not feasible since boards are nominated by dominant non-tradable shareholders and they did not perform well in terms of monitoring managerial behaviour in the best interest of minority tradable shareholders. Also, external monitoring through corporate takeovers was ineffective because of the non-transferability of controlling shares. Consequently, controlling shareholders issued new equity and raised money to seek rents through related-party transactions, corporate lending to their large shareholders, guarantees of loans for controlling shareholders, etc.

⁴ Consistently, as shown in Table 1, we find that the leverage ratio of Chinese listed firms grows by 9% after the reform.

⁵ Similar to previous Chinese empirical studies (e.g. Huang and Song 2006; Qian, Tian, and Wirjanto 2009), we do not use the market value of debt ratio to construct the measure of leverage ratio because there is a large amount of non-tradable shares in the Chinese stock markets before 2007 and this will lead to a relatively low market value of total debt ratio.

⁶ Liao, Liu, and Wang (2014) note that most of Chinese listed firms completed the share structure reform during 2005–2006.

⁷ We also ran the regression model (2) across different sub-samples grouped by firm size, leverage and growth and in unreported results, found that an asymmetric debt financing pattern remains across those sub-samples.

⁸ Adjustment costs refer to the costs that may occur when firms switch between debt financing and equity financing, including taxes, legal fee, registration fees, printing and accounting costs and other fees paid to the market dealers for placing the issue (Oliner and Rudebusch 1992; Bhanot and Guo 2011; 2012, Guo 2013).

⁹ The speed of adjustment should be between 0 and 1, if there are positive adjustment costs. A large coefficient would suggest low adjustment costs for the firm and hence, it can adjust toward its desired leverage level quickly. When the coefficient is 1, the firm makes the adjustment instantaneously.

¹⁰ The Hausman test is applied to compare the fixed-effects model with the random-effects model and it suggests to use the fixed-effects models in all of the panel data analysis applied in this study.

¹¹ Roodman (2009) claims that the Sargan statistic may be inconsistent and a theoretically superior over-identification test is based on the Hansen statistic (Hansen and Singleton 1982) from a two-step estimate (see, e.g. Guo 2015; Guo and Lien 2015).

¹² If AR2 test is rejected, there is an evidence of second-order autocorrelation, indicating that the instruments used in the estimation are not appropriate. The rejection of the Hansen test suggests that the GMM estimated results suffer from over-identification problems and therefore the results should be treated with caution. In terms of regression specifications used in this article, the AR2 tests show that no models suffer from the problem of second-order autocorrelation while the Hansen tests find no over-identification biases in our empirical results, implying that all the models are well-specified and the instruments used in the GMM estimations are valid.

¹³ The fixed-effects estimates of the adjustment speed seem to have a small upward bias compared to the GMM estimates. Therefore, we will focus our discussion on the GMM estimates.

¹⁴ Qian, Tian, and Wirjanto (2009) substitute (3) in (2) and estimate the resulting model in one stage. Also, the coefficients are estimated by Arellano and Bond’s (1991) difference GMM model.

¹⁵ Note that Equation 5 is reduced to the partial (symmetrical) adjustment model when $\phi =\theta$ and $\gamma$ =1.