Capital structure adjustments in private business group companies

Abstract

The literature on capital structure dynamics assumes that companies trade-off the advantages of a leverage adjustment and its costs. In general, private companies are assumed to face relatively large adjustment costs, and should have lower financing flexibility. However, we argue that an important class of private companies – business group affiliates – may face relatively low adjustment costs because of their access to both internal and external capital markets and the beneficial reputation effects of belonging to a group. Our empirical results show significant differences in the composition of the capital structure and the leverage adjustment process between affiliates of private Belgian business groups and comparable stand‐alone companies. Group affiliates have higher levels of leverage, and adjust their capital structure more frequently than stand‐alones. Our evidence suggests that the flexibility in group companies’ capital structure is not solely driven by the use of internal leverage: group affiliates more frequently adjust their external leverage as well, unless the group is in poor financial health, in which case the affiliates’ probability of attracting external leverage is severely reduced.

Keywords:

• business groups
• internal capital markets
• capital structure
• ownership structure

JEL Classification::

• G32
• L22

Acknowledgements

The authors gratefully acknowledge useful comments and suggestions on a previous version of this article by Frederiek Schoubben, Marc Deloof, Marie Dutordoir and participants at the 2009 EFMA Conference, the 2009 FMA European Conference and the 2010 Campus for Finance Research Conference.

Notes

¹ Empirical evidence on this issue is very limited: to the best of our knowledge, only one paper takes into account the impact of group‐affiliation on capital structure adjustment: for a sample of Korean listed manufacturing companies Kim et al. (2006) find that both the target level of leverage and the speed of leverage adjustment are higher for chaebol members.

² Under Belgian Accounting Law, companies are required to file complete (unconsolidated) accounts if they meet at least two of the following criteria: total assets exceed 3.125 million euro; operating revenue exceeds 6.25 million euro; more than 50 full time equivalent employees. Companies with more than 100 full time equivalent employees always have to file complete accounts. All other firms may file abbreviated accounts, which contain less information on issues which are relevant to our research questions (e.g. intra‐group financing).

³ Using the full stand‐alone sample would lead to important differences in size and industry distribution across samples. For instance, the total assets of the median group sample company are more than twice as large as those of the median firm in the unmatched stand‐alone sample. Size matching is based on total assets. Industry matching is based on a 2‐digit Statistical Classification of Economic Activities in the European Community (in French: Nomenclature statistique des Activités économiques dans la Communauté Européenne, NACE) classification code. As a robustness check, the tests were rerun on the unmatched stand‐alone sample, with very similar results. This implies that the matching procedure is unlikely to impact our conclusions.

⁴ Robustness checks show that although changing the cut‐off, e.g. 3% or 7% obviously affects the number of leverage increase and decrease firm years, the main results and conclusions of the univariate and multivariate tests remain unchanged.

⁵ Test statistics for Z‐test of equality of proportions of firm years with external leverage adjustments between stand‐alone sample and group sample: 6.10 (leverage increase firm years) and 6.27 (leverage decrease firm years).

⁶ Given the panel structure of our data set, the multinomial logistic regression's assumption of independence across observations cannot be fulfilled. Inferences are therefore based on Cluster Robust Standard Errors (CRSE) which allows the variance estimator to take into account within‐cluster correlation. Note that there are two levels of potential error clustering in the group sample: within company‐specific clusters and within group‐specific clusters. Given there is no transference of ownership of sample companies across groups in our sample, the two levels of clustering are nested, and the highest level of clustering (i.e. the group level) should be used to compute the CRSE (cf. Cameron et al., 2011).

⁷ Several types of robustness checks on the model specification were performed: (i) different target leverage ratios (based on model 1 (Table 3) or model 2 (Table 3)); (ii) allowing for a different coefficient for deviation to the target (LEV* – LEV_{t −  1}) across affiliates and stand‐alones and (iii) allowing for different coefficients for all explanatory variables across affiliates and stand‐alones. Findings remain qualitatively unchanged.

⁸ Alternative model specifications similar to those described in the previous footnote lead to comparable results.

⁹ For instance, data from 1999 to 2001 is used to estimate the leverage targets for 2002, data from 2000 to 2002 is used to estimate the leverage targets for 2003, and so on. The downside of this approach is a substantial loss of testable observations because out‐of‐sample targets cannot be estimated for 1999 and 2000.