Essays on failure risk of firms using multivariate frailty models

Abstract

The post-2007 global financial crisis, characterised by huge firm losses, especially in the USA and Europe, initiated a new strand of literature, where default models are adjusted for unobserved risk factors, including measurement errors, missing firm specific and macroeconomic variables. These new models assume that default correlations are not only driven by observable firm-specific and macroeconomic factors, but also by unobserved risk factors. This thesis present three empirical essays. The first essay estimates and predicts the within-sector failure rate and dependence of firms on the London Stock Exchange. The study offers an additive lognormal frailty model that accounts for both unobserved factors and regime changes. The analysis reveals that during distressed market periods the sector-based failure rates and dependencies tend to be high. The second essay proposes a novel approach based on a bias-corrected estimator to investigate the impact of informative firm censoring and unobserved factors on hazard rates of US firms. The approach uses inverse probability of censoring weighted scheme that explicitly accounts for firm specific factors, economic cycles, industry-level dependence and market activities induced by unobservable factors. The analysis shows that during distressed market periods the effect of informative censoring averagely increases the hazards rates, and varies across industries. The third essay employs a mixed effects Cox model to estimate the failure dependence caused by firms’ exposure to country-based and group-level unobserved factors within the Eurozone. The empirical results show that a higher failure dependence among firms in groups of countries with similar economic and financial conditions than countries with different conditions. Overall, the thesis contributes to the empirical literature on firm default in the broad area of corporate finance by offering a different approach of capturing default dependence and its variations during unfavourable market conditions and adjusting for the effects of non-default firm exit on active firms.