Benefits of control, capital structure and company growth

Abstract

This article studies the influence of the benefits of control on the capital structure and the growth of private companies for a sample of 8964 UK companies with limited liability observed for up to 5 years. It is hypothesized that companies in which existing owners would lose more control if they expanded, have smaller equity increases, are more highly levered and grow more slowly. Potential loss of control is measured as the difference in the probability of winning a vote for the largest owner before and after a hypothetical equity increase. Evidence is found that is consistent with the hypotheses.

Acknowledgements

I would like to thank Ron Anderson, Christian Laux and Steve Nickell for helpful discussions, conference participants at ESEM 2004 in Madrid and seminar participants at the London School of Economics, the University of Manchester, the University of Mannheim and the Centre for European Economic Research (ZEW) for useful comments. All remaining errors are my own.

Notes

¹ For more recent analyses of the private benefits of control see, for example, Nicodano and Sembenelli (2004) and Dyck and Zingales (2004).

² Cubbin and Leech (1983) developed a formula to calculate approximate probabilities in the case of many owners with this voting model. For this analysis, it is possible to calculate the exact probabilities, because the number of owners is smaller in private companies. Felsenthal and Machover (1998, p. 36; p. 171f.) discuss the interpretation of measures of voting power. The Banzhaf measure reflects voting power as the degree to which an owner's vote is able to influence the outcome of a decision; and the Shapley–Shubik measure reflects voting power as the expected payoff that an owner gets from a fixed prize that is allocated to the winning coalition. Leech (2002) finds that the Banzhaf measure reflects variations in the power of shareholders of British listed companies better than the Shapley–Shubik measure. The measure used here is a linear function of the Banzhaf measure. It was, among others, also used by Nickell et al. (1997)

³ The relative weights are 0.71 for a 10% increase, 0.1 for a 30% increase, 0.07 for a 50% increase, 0.03 for a 70% increase and 0.09 for a 90% increase.

⁴ As a robustness check the potential loss of control was calculated for the case that equity is increased by an additional owner who enters the company. The results were very similar.

⁵ Restricted by the information available in the data set, a company is defined to be in family ownership, if two or more owners have the same last name. This is the case for 44% of companies in the sample.

⁶ The optimality of one-share one-vote was studied by Grossman and Hart (1988) and Harris and Raviv (1988) in the context of public companies faced by take-over threats.

⁷ The information on institutional details in this section are taken from the Companies Act 1985 and 1989 (Dey, 1994) and from the internet site of Companies House (www.companieshouse.gov.uk).

⁸ It should be noted that merger and acquisition (M&A) activity is not important for the results. With the use of the Zephyr database, distributed by Bureau van Dijk, the companies engaging in M&A have been identified. The results are not affected when those companies are excluded from the analysis.

⁹ Conditional on positive equity growth, the mean growth rate is 20.2% and the median is 4.8%.

¹⁰ In an exploratory specification, the ownership share was interacted with the dummy for the largest owner being a manager. Since the difference of the two effects was not significant, results are shown without the interaction term.

¹¹ The preference ratio, i.e. the ratio of preference capital to total equity capital, is quite high for companies that use preference capital. Its mean is 40.7% whereas its median is 35.7%.

¹² This specification contains more observations than the previous one explaining equity growth. Here all companies are included, whereas the equity growth specification excluded companies for the years in which they had negative equity growth.

¹³ This specification was also first estimated with the ownership share of the largest owner included up to its fourth power. The third power was significant, but since the cubic form showed a positive relationship over the relevant range, the specification with only the linear term is shown. This makes the results better comparable across the three subsections. The results of the other regressors were not affected.

¹⁴ There can also be an effect of reverse causality. If companies with high leverage need to pay higher interest rates, then profitability can be reduced. The results for the other regressors remain the same, when the first age of return on assets is excluded.

¹⁵ The results are qualitatively identical when the time period 1997 to 2000 is considered.

¹⁶ The company growth regression covers fewer companies than the previous ones because not all companies are observed over this period. An estimation with a Heckman correction for attrition bias has therefore been employed. But since the error terms of the selection and the main equation have not been correlated, the results using the standard OLS technique are shown.

¹⁷ It can be argued that leverage is endogenous in this regression. Companies in difficulties will have small growth rates and low profitability. Leverage can build up, if interest and capital cannot be serviced any more. This leads to an effect from small growth rates to high leverage. In this situation it would be appropriate to instrument leverage, but no good instruments are available. Therefore a specification without leverage was tested. The signs and significance levels of the other regressors remain the same.