Nonlinear target adjustment in corporate liquidity management: an endogenous thresholds approach

Abstract

We provide new empirical evidence on nonlinear liquidity management in Dutch firms. Our results reveal that liquidity adjustment from below the target is significantly faster than from above. We find no evidence for bands of inaction around the target.

Acknowledgements

We thank Jaap Bos, Bertrand Candelon, Leo de Haan, Alain Hecq, Wilko Letterie, Michiel van Leuvensteijn, Jean-Pierre Urbain, Tom van Veen, an anonymous referee, as well as participants of the CAED 1999, the NAKE Research Day 1999, the European Monetary Forum 2002, the Maastricht University Economics Lustrum Conference 2004 and seminar participants at De Nederlandsche Bank for helpful comments. The research is supported by a grant from the Maastricht Research School of Economics of Technology and Organizations (METEOR) and was partially carried out at the Center for Research of Economic Micro-data (Cerem) of Statistics Netherlands. The views expressed are those of the individual authors and do not necessarily reflect official positions of Statistics Netherlands or Rabobank.

Notes

¹ See e.g. Huang et al. (2001) for an application of a similar conceptual framework in the case of macroeconomic money demand.

² We refer to the appendix for details on sample selection.

³ For a more extensive definition and motivation of this list of long-run determinants, we refer to BK04 and Opler et al. (1999). Originally, net working capital and investment were included in the list of long-run determinants as well. However, they lacked statistical significance in the long-run equation and were subsequently dropped from the analysis.

⁴ Note the resemblance of our approach with the way that debt targets have been computed in the capital structure literature (Auerbach, 1985; Shyam-Sunder and Myers, 1999; Drobetz and Wanzenried, 2006).

⁵ This conclusion is independent of the level of aggregation of the included sector dummies; i.e. even two-digit sector dummies do not adequately capture individual firm's idiosyncratic long-run liquidity levels.

⁶ The motivation for this simplification is that the endogenous threshold method applied later on is tailored to balanced panels and it is unsure whether its asymptotic properties extend to unbalanced panels. At the same time, the dominance of the firm-specific effects in the cross-sectional variation of liquidity targets limits the resulting loss in precision.

⁷ Since the likelihood ratio test statistic may not be asymptotically pivotal, a word of caution on the efficiency of the bootstrapped test statistic is warranted. The errors of bootstrap estimates of the distribution of statistics that are not asymptotically pivotal converge to zero at the same rate as the errors made by first-order asymptotic approximations (Horowitz, 2001).

⁸ Additionally, Shapiro–Francia W’ tests for normality of the residuals of the triple-threshold model reveal that at the 99% confidence level, normality is accepted for more than 90% of the firms in our sample.

⁹ The Hansen (1999) methodology is tailored to balanced panels, while it is unsure whether its asymptotic properties extend to unbalanced panels. We therefore aim to keep the panel balanced and remove 17 firms entirely from the sample rather than unbalancing our panel by removing individual observations.

¹⁰ We have also censored at the 2nd and 98th quantile. This hardly affects the outcomes.