Review of Quantitative Finance and Accounting, 13 (1999): 323±345
# 1999 Kluwer Academic Publishers, Boston. Manufactured in The Netherlands.
Predicting Corporate Financial Distress: A Time-Series
Associate Professor of Accounting, School of Business, Rutgers University,
Camden, NJ 08102
Professor of Finance, School of Business, Rutgers University, Camden, NJ 08102
Abstract. The ability to predict corporate ®nancial distress can be strengthened using models that account for
serial correlation in the data, incorporate information from more than one period and include stationary
explanatory variables. This paper develops a stationary ®nancial distress model for AMEX and NYSE
manufacturing and retailing ®rms based on the statistical methodology of time-series Cumulative Sums
(CUSUM). The model has the ability to distinguish between changes in the ®nancial variables of a ®rm that are
the result of serial correlation and changes that are the result of permanent shifts in the mean structure of the
variables due to ®nancial distress. Tests performed show that the model is robust over time and outperforms
similar models based on the popular statistical methods of Linear Discriminant Analysis and Logit.
Key words: ®nancial distress models, linear discriminant analysis logit model, non-stationary ®nancial ratios,
time-series CUSUM, vector autoregressive process
JEL Classi®cation: C32, G33, M41
Financial variables included in ®nancial distress models as explanatory variables generally
exhibit positive serial correlation over time, e.g., Theodossiou (1993).
As such, positive
deviations in these variables from their long-run equilibrium means in one period are
followed by positive deviations in subsequent periods while negative deviations are
followed by negative deviations. The magnitude of these deviations depend on the degree
of serial correlation inherent in the ®nancial variables as well as a random white noise
error term. The presence of serial correlation may be attributed to active attempts by the
management to align the variables with their population means and/or systematic micro-
and macroeconomics effects operating on the ®rm, e.g., Lee and Wu (1988). The random
error term may be attributed to ®rm-speci®c or economic random shocks.
Under stationarity of the time-series process, the variables have a tendency to revert
around their long-run equilibrium means over time. The latter implies that the deviations
of the variables from their means are transitory over time. Thus, the variables have a
tendency to return back to their mean values in the near future. The time needed for the
variables for such a return (i.e., the persistence of deviations) depends on the degree of
serial correlation inherent in the variables.
In addition to the serial correlation and white noise error components, both of which are