The impact of firm characteristics on speed of adjustment to target leverage: a UK study

ABSTRACT

Responding to the need to investigate heterogeneity in the speed of adjustment (SOA) to target leverage in a manner that reflects the fractional nature of leverage, we estimate SOA across sub-samples of UK firms using the Dynamic Panel Fractional (DPF) estimator. Using firm characteristics to identify firms subject to varying costs of deviation from and adjustment to target leverage, we find significant evidence of heterogeneity in the speeds at which UK firms adjust to target leverage. Our results show that small, high growth and low dividend paying firms adjust to target leverage faster than their large, low growth and high dividend paying counterparts. We also find some evidence to suggest openly held firms adjust faster than closely held firms, though our results are not robust to the categorizing criterion employed or target leverage specification.

KEYWORDS:

• Target leverage
• speed of adjustment
• heterogeneity
• dynamic panel fractional estimator

JEL CLASSIFICATION:

• C23
• C24
• G32

Disclosure statement

No potential conflict of interest was reported by the authors.

Notes

¹ We do not winsorize our R&D dummy variable as it is a binary variable.

² We use total assets as the denominator rather than net income, as zero and negative values of net income would significantly reduce the number of usable observations.

³ For example, if a firm’s value for the categorizing variable is below the within-period median value in 1996, 1997 and 1998, the firm is assigned a value of 1 in each of these years. If its value for the categorizing variable is above the within-period median value in 1999, 2000, 2001, 2002 and 2003, the firm is assigned a value of 2 in each of these years. The time-series average of the values assigned to the firm is 1.625, and this is then rounded up to 2. Thus, all observations for the firm are categorized into the ‘high’ sub-sample.

⁴ The results of our re-estimated model are available upon request.

⁵ The z test statistic is calculated as follows: $z=\frac{{\beta }_1-{\beta }_2}{\sqrt{se{\beta }_1^2+se{\beta }_2^2}}$, where ${\beta }_1$ and ${\beta }_2$ are the coefficients of the lagged dependent variable within each sub-sample pairing, and $se\beta$₁ and $se\beta$₂ are the associated standard errors.

⁶ Due to data limitations relating to some of the additional variables we construct, the number of firm-year observations before accounting for the reduction due to lagged explanatory variables decreases to 14,959.