The Correct Use of Hazard Models: A Comparison Employing Different Definitions of SMEs Financial Distress

Abstract

This study aims to shed light on the debate concerning the choice between discrete-time and continuous-time hazard models in making bankruptcy or any binary prediction using interval censored data. Building on theoretical suggestions from various disciplines, we empirically compare widely used discrete-time hazard models (with logit and clog-log links) and continuous-time Cox Proportional Hazards (CPH) model in predicting bankruptcy and financial distress of the United States SMEs. In line with the theoretical arguments we report that discrete-time hazard models are superior to continuous-time CPH model in making binary predictions using interval censored data. Moreover, hazard models that employ failure definition of SMEs based on bankruptcy laws and firms' financial health exhibits superior goodness of fit and classification measures, in comparison to hazard models that employ failure definitions that are based either on bankruptcy laws or firms' financial health.

Expected cost of financial distress in small and medium-sized enterprises (SMEs): A German-Italian comparison

“…In literature we see a rich literary production engaged in the formulation of ex-ante models to estimate the financial distress likelihood (FDL). Particularly, Grice and Ingram (2001), Pindado et al (2008), Keasey et al (2015) and Gupta et al (2015) underline that the seminal models by Beaver (1966), Altman (1968Altman ( , 1984 and Ohlson (1980) are revealed suitable in the classification of the sector (better for the manufacturing enterprises) but poorly sensitive to the typology of financial distress of the enterprise. To analogous conclusions, the studies of Zmijewski (1984) are reached.…”

Time to default in credit scoring using survival analysis: a benchmark study

“…Note that this resulted in data sets with either only personal loans or data sets with a mix of personal and small enterprise loans, for banks C and D. As the SMEs (small and medium-sized enterprises) in our data sets were all sole proprietorships, their properties were nearly identical to those of personal loans. More information on the use of survival models in SMEs in the broader sense can be found in, among others (Fantazzini and Figini, 2009;Gupta et al, 2015;Holmes et al, 2010). For the banks with data covering several loan terms, the data were split in order to get only one loan term per data set, resulting in ten data sets.…”

Public vs Private SMEs: A Comparison of Distress Hazard

“…Considering the suggestion of Gupta et al (2015), we use discrete-time duration-dependent hazard model with logit link to perform univariate and multivariate one-year financial distress hazard analysis of listed and unlisted SMEs respectively. Financial ratios with established reputation of distress prediction in earlier studies are being used as covariates along with liquidity measure of Amihud (2002) and illiquidity measure of Florackis et al (2011).…”

“…However, in this study we concentrate on financial distress rather than legal bankruptcy with the presumption that it is the primary reason behind bankruptcy and always precedes the bankruptcy filing event. Further, filing for legal bankruptcy is the least efficient exit strategy for SMEs (Balcaen et al 2012) and distress definitions based on bankruptcy laws are inefficient in comparison to distress definition based firms' financial performance (see Gupta et al 2015). Thus, following Keasey et al (2014), an SME experiencing financial distress is defined as one that satisfies the following: (i) its expenses exceeds earnings during two consecutive years, (ii) its total debt exceeds net worth during those two years in (i), and (iii) it records negative growth in net worth during the same consecutive periods in (i) and (ii).…”

“…While these specifications will generally produce results that are quite similar, there are significant differences in terms of nonproportionality (see Sueyoshi (1995) for detiled discussion). Therefore, considering this discussion and the suggestion by Gupta et al (2015), we employ a discrete hazard model with logit link to develop our distress prediction models.…”