The 2007–2009 financial crisis: changing market dynamics and the impact of credit supply and aggregate demand sensitivity

Abstract

This article singles out the determinants of changes in US firms’ systematic risk and idiosyncratic return induced by the 2007–2009 financial crisis. After establishing that systematic risk changes during the crisis, the results show that higher operational and financial leverage coincide with an increase in systematic risk, while high cash availability is associated with a decrease in systematic risk. The crisis-induced idiosyncratic return worsens with increasing financial leverage, higher sensitivity to aggregate demand shocks and banking sector problems, and lower operational leverage. Additional results show that the aforementioned variables have economically large effects on firm performance during the crisis.

Keywords:

• financial crisis
• time-varying systematic risk
• credit supply
• aggregate demand
• portfolio management

JEL Classification:

• G01
• G12
• G32

Acknowledgements

We thank Nicolas Boccard, Anthony Saunders, Carles Vergara-Alert, an anonymous referee and seminar participants at De Nederlandsche Bank, the XIX Finance Forum, Granada, Spain and the VIII Workshop in Public Policy Design, Girona, Spain, for useful comments and suggestions. This article presents the authors’ personal views and opinions. They do not necessarily coincide with those of de Nederlandsche Bank. All errors are our own.

Notes

¹ See, e.g., Andersen et al. (2006), Van Oordt and Zhou (2011) and references therein.

² See Grammatikos and Vermeulen (2012) for an application on the recent financial crisis. The authors find that European nonfinancials become more dependent on US market developments, compared to the pre-crisis period.

³ Lettau and Ludvigson (2001) and Petkova and Zhang (2005) report similar results.

⁴ Unreported results show that the magnitude and significance of the estimated coefficients are very similar when using a quantile regression on the median or when using bootstrapped SEs in the OLS estimation.

⁵ We are aware of Δα and Δβ being generated variables (in Equation 1). The coefficients in Equations 4 and 5 are however unbiased, and we correct the SEs for heteroscedasticity. Implicitly we assume that delta alpha and delta beta are independent by studying these variables in separate regressions. This assumption seems reasonable because the correlation between delta alpha and delta beta is low at 0.17.

⁶ Rajan and Zingales (1998) highlight the endogeneity problem when using contemporaneous data.

⁷ Note that Tong and Wei (2011) deviate from this definition by changing the sign on the return and rescaling the variable. This implies that a positive sign on the demand sensitivity variable in this article has the same interpretation as a negative sign on the demand variable in Tong and Wei (2011).

⁸ See Tong and Wei (2008) for a more extensive discussion on and justification of the demand sensitivity indicator.

⁹ Note that Rajan and Zingales (1998) only consider manufacturing industries, while this article also includes firms in service industries. Capital investments are less important for service industries, which may affect the usefulness of the external finance indicator.

¹⁰ The exclusion of new entrants and exiting companies in the S&P 1500 index can induce a survivorship bias. However, because of different reasons of exits, such as mergers, bankruptcy or delisting by acquiring investors, it is not possible to determine how this potential bias affects the results.

¹¹ For an extensive analysis on bank risk during the financial crisis, we refer to Altunbas et al. (2011).

¹² Robustness tests in Section V assess the sensitivity of the results to this approach.

¹³ Unreported results using raw returns show that the main conclusions when using log-returns remain unaffected. These results are available from the authors upon request.

¹⁴ These results are available from the authors upon request.

¹⁵ Tong and Wei (2011) use the cash conversion cycle to capture cash needs for working capital. The cash conversion cycle is appropriate for manufacturing industries, but less suitable for service industries, which dominate our sample.

¹⁶ Notice that increased sensitivity to demand or banking system problems implies a more negative value of these variables; the positive coefficients in Table 4 indicate an increasingly negative effect on delta alpha when sensitivity to demand shocks or banking system problems increases.

¹⁷ More detailed results at the industry level are available from the authors upon request.