International Review of Economics and Finance 37 (2015) 3341

Contents lists available at ScienceDirect

International Review of Economics and Finance

journal homepage: www.elsevier.com/locate/iref

An analysis of the determinants of nancial distress in Italy:

A competing risks approach
Alessandra Amendola a, Marialuisa Restaino a,, Luca Sensini b
a
b

Department of Economics and Statistics, University of Salerno, Via Giovanni Paolo II, 132 84084 Fisciano, SA, Italy
Department of Management and Information Technology, University of Salerno, Via Giovanni Paolo II, 132 84084 Fisciano, SA, Italy

a r t i c l e

i n f o

Article history:
Received 8 May 2012
Received in revised form 29 March 2014
Accepted 31 October 2014
Available online 8 November 2014
JEL classication:
C34
C40
G33
G34
Keywords:
Financial distress
Firm's exit
Competing risks model
Forecasting

a b s t r a c t
This paper investigates the inuence and the effect of micro-economic indicators and rm-specic
factors on different states of nancial distress. In particular, a competing risks model is estimated
taking into account the differences among variables leading rms to exit the market through
bankruptcy, liquidation and inactivity. The determinants of nancial distress for any exit route
are identied on the basis of the inuence on the hazard ratios of the signicant variables selected
for each state. Furthermore, the predictive performance of the competing-risks model over the
single-risk framework is evaluated, with respect to different time windows, by means of some
accuracy measures. The results reached on a sample of Italian rms provide support for the hypothesis that the factors inuencing rms' way out strongly depend on the exit routes and
highlighting the need to distinguish among them by means of a multiple-state approach.
2014 Elsevier Inc. All rights reserved.

1. Introduction
Modeling rm survival and studying the effect of factors that determine rm's exit are drawing an increasing attention from both
academics and practitioners over the last years. Most of the existing literature treats exit from the market as a homogeneous event or
focuses on only one form of exit, separately investigating any decision to leave the market. Starting from the seminal paper of Altman
(1968), researchers have essentially focused on the failing and non-failing dichotomous variable, examining the companies that actually went bankrupt by means of some models (logit, probit, discriminant analysis, survival analysis and so on) (Ohlson, 1980;
Zmijewski, 1984; Lennox, 1999; Shumway, 2001; Brabazon & Keenan, 2004; Figlewski, Frydman, & Liang, 2012, among theothers).
However, there are different exit options that may force a potentially distressed company to leave the business. Besides entering in
involuntary exit procedure (such as bankruptcy), a rm could choose for a merger or acquisition or decide for a voluntary liquidation.
Each type of exit is likely to be driven by different factors and can determine important implications for the stakeholders and, in general, for the whole economy (Harhoff, Stahl, & Woywode, 1998; Schary, 1991). Investigating the determinants leading to the different
forms of distressed rm's exit can, therefore, be particularly relevant. In order to examine the effects of explanatory variables across
the states of nancial distress, a multi-state approach can be used (Headd, 2003; Hensher & Jones, 2007; Jones & Hensher, 2004, 2007;
Rommer, 2005). Some studies analyze the different types of exit, whereas not much is said about the similarities or dissimilarities

Corresponding author.
E-mail addresses: alamendola@unisa.it (A. Amendola), mlrestaino@unisa.it (M. Restaino), lsensini@unisa.it (L. Sensini).

http://dx.doi.org/10.1016/j.iref.2014.10.012
1059-0560/ 2014 Elsevier Inc. All rights reserved.

34

A. Amendola et al. / International Review of Economics and Finance 37 (2015) 3341

among factors determining them (Chancharat, Tian, Davy, McCrae, & Lodh, 2010; Esteve-Prez, Sanchis-Llopis, & Sanchis-Llopis,
2010).
The aim of this paper is to give a contribution in this direction studying the determinants of the probability of alternative exit
routes, with particular attention to the differences in the factors driving rms out of the market. The effects of micro-economic indicators and rm-specic variables on different states are examined by a competing risks hazard model. This model is used for determining the probability and hazard ratio of three mutually exclusive ways of exit, namely bankruptcy, liquidation and inactivity. The
rst category includes those rms that involuntary exit the market; the second refers to rms that opt for a liquidation encouraged
by several possible reasons (avoid involuntary exit, restructuring, etc.). The last category includes those rms that exit the market
for reasons different from the previous ones. The active rms are selected as reference group. Unlike discrete outcome models
(logit, probit), hazard models allow to account for both whether and when an event occurs, by tracking the evolution of the risk
over time. Moreover, the competing risks model provides information regarding whether the effects of each variable change across
the multiple states of nancial distress.
Here we develop a four-state Cox proportional hazards competing risks model, in which the states are considered to be independent. In order to highlight the diverse role played by the explanatory factors, a single-risk model is also estimated in which all nancial
distress states are pooled together. The results obtained by the two model specications are compared not only in terms of the significance and sign of the selected variables, but also on the basis of hazard ratio and nancial meaning. A further comparison is made on
the forecasting accuracy evaluating the capability of the models to predict rms' exits by means of some accuracy measures. The analysis has been carried out on a sample of Italian rms. In particular we refer to the building sector underlining its relevance not only in
terms of relative weight of GNP, but also with respect to its role in the various objectives behind national development planning in
many European countries.
To perform these approaches and provide empirical evidence, a set of explanatory variables is considered from which selecting the
best-set of possible candidate indicators to be included in the estimated models. To this purpose, few considerations pointed out in the
literature have been taken into account. Some company characteristics, such as age and size, affect the probability of failure. In particular, the likelihood of a rm to go bankrupt decreases with size and age (Bhattacharjee, Higson, Holly, & Kattuman, 2009; Esteve-Prez
et al., 2010). Corporate governance, specically the structure of the rm's board of directors and ownership and the interaction among
them, may also affect the probability of failure. The agency problem between the owners of a rm (its shareholders) and the management leads to inefciency in case of ownership concentration (Zeitun, 2009). The legal form can be also considered as a potential indicator for risk measures. Private limited liability companies would face higher risk, as they would have less share capital to lose
compared with public limited liability companies (Esteve-Prez et al., 2010). From what concern the indicators of rm's nancial
performance, we consider the most relevant and effective indicators in highlighting current and prospective conditions of nancial
distress that refer to the different methodological proposals, from the pioneer works on the topic (Fitzpatrick, 1931, 1932; Smith &
Winakor, 1930) since the more recent contributions (Altman & Hochkiss, 2006; Amendola, Restaino, & Sensini, 2011; Balcaen &
Ooghe, 2006; Laitinen & Suvas, 2013; Ravi Kumar & Ravi, 2007; Xie, Shi, & Wu, 2008).
To anticipate the results, our ndings reveal several differences in the factors determining rms' way out with respect to the exit
routes. In particular, we nd out that some rm-specic characteristics, such as age, legal form and size, have inuence on the probability of being liquidated, inactive and bankrupted, thus conrming the empirical results available in literature. The protability ratios
also play a relevant role on the likelihood of going bankrupt. Then, for the single-risk model the variables selected show some similarities to those characterized by the inactive state and liquidation, whereas they are different for bankruptcy. Thus, our results corroborate the need to separately investigate the different forms of exit, and allow to better understand the effects of diverse
explanatory factors. The method developed in this paper can be easily applied to data collected from other industries or countries.
The paper is structured as follows. In the next section, the statistical method is briey reported. The predictors' dataset is introduced in Section 3. The results are discussed in Section 4, while Section 5 concludes.
2. Methodology
The Competing risks model is one of the most popular settings of the Multi-State Models (for details, see (Andersen, Borgan, Gill, &
Keiding, 1993; Hougaard, 2000; Andersen, Abildstrm, & Rosthj, 2002)). It extends the simple mortality model for survival data and
is based on one transient state (alive state) and a certain number of absorbing states, corresponding to death from different causes.
Thus, all transitions are from the state alive.


e C be the observed time and let = I(T C)
Let Te and C be the failure time and the censoring time, respectively. Let T min T;
be an indicator function, which is equal to 1 if the cause of failure is known, and zero otherwise. Thus, the observed data are given by
(T, ).
Let D be the cause of failure (event-causing failure). Assume that the possible causes are numbered from 1 to K. The main
feature of competing risks model is that from a given set of k causes, one and only one cause can be assigned to every failure.
Analyzing competing risks data means to get insight the joint distribution of T and D. The fundamental concept in competing
risks model is the cause-specic hazard function, i.e. the probability of failing due to a given cause k, after one has reached the
time point t:

k t lim

t0

P T t t; D kjT t 
;
t

k 1; ; K:

A. Amendola et al. / International Review of Economics and Finance 37 (2015) 3341

35

Then, the cumulative cause-specic hazard function is:

Z
k t

t
0

k sds;

and the cause-specic survival function is:

Sk t exp k t :

Finally, the overall hazard function can be written as:

t

K
X

k t ;

k1

Since we are sometimes interested in the probability to fail from cause k up to time point t, the probabilities as functions of t, called
the cause-specic cumulative incidence functions, can be considered:
Ik t P T t; D k:

Cause-specic hazard function and cumulative incidence function point to a single cause. If we are interested in the relative contribution of the different causes to the overall failure, we can transform these quantities to relative measures of the contribution.
Thus, the probability to have failed because of a given cause k and before time t is given by:
I t
I
k t P D kjT t Xk
:
I t
k k

Then, we can consider the probability to fail from cause k, given one will fail within a short time interval after reaching the time
point t, leading to:
t

k X k
;
t
k k

k 1; ; K:

Let's assume that 0 b t1 b t2 b b tu are ordered distinct time points at which failures of any causes occur. Let dkr (r = 1,, u) denote the number of rms failing from cause k at tr and let dr = Kk = 1 dkr denote the total number of failures (from any cause) at tr. In
the absence of ties, only one of the dkr equals 1 for a given r and dr = 1. Let nr be the number of rms at risk (i.e. rms that are still in
follow-up and have not failed from any cause) at time tr.
Consider a discretized version of the cause-specic hazard, i.e. the proportion of subjects at risk that fail from cause k:
k t r P T t r ; D kjT N t r1 :

It would be estimated by:

^ t dkr ;

k r
nr

for k = 1,, K and for r = 1,, u.

Since the cause-specic hazards are identiable, a regression on them is possible. In Proportional Hazards Regression, the causespecic hazard of cause k for a subject i with covariate vector Zi is given by:
n
o
T
ik tjZik k;0 t exp k Zik t ;

10

where k,0(t) is the baseline cause-specic hazard of cause k which does not need to be explicitly specied, Zik(t) is a vector of covariates for rm i specic to k-type hazard at time tt, and the vector k represents the covariate effects on cause k to be estimated. Since
the same variables could have different effects on the different risks, it is reasonable to assume that, for each k, k is independent of
each other.

36

A. Amendola et al. / International Review of Economics and Finance 37 (2015) 3341

In order to have an estimate of the coefcients' vector, we build the partial likelihood function for each specic hazard k using the
results available in univariate Cox Proportional Hazard model:
n
o
T
exp k Zik t
Lk k X
T Z t
i1
lRt k lk
nk

11

ik

where nk is the number of rms in specic hazard k, and R(tik) = {l|tlk tik} is the set of individuals at risk at time tik.
The overall partial likelihood function is given by:
n
o
T
exp k Zik t
L1 ; ; k X
T Z t
k1 i1
lRt k lk
K

nk

12

ik

3. The data
The data refer to the population of Italian rms that operate in the building sector1 in the period 20042009. The information on individual rms and their nancial characteristics have been obtained from the Amadeus database, provided by Bureau van Dijk. Our main
interest is in investigating the determinants of rms that end up in nancial distress and compare these determinants for each different
form of exit. The focus is on three mutually exclusive states of exit from the market: bankruptcy, liquidation and inactivity. The bankrupt
status includes those rms that have been legally declared not to be able to pay its creditors and are under a Court supervision. The second
status includes those companies that no longer exist because they have ceased their activities and are in the process of liquidation. The
last state includes those rms that exit the database, but it is unknown the reason of the exit. The reference group is given by active rms.
From the overall population of active and non-active rms, we select a cluster random sample of n = 1462 rms based on the geographical distribution of the industrial rms across the regions. The distribution of the nal sample consists of 221 companies that went
bankrupt, 129 that had entered voluntary liquidation, and 228 that were inactive. The companies in the active state are 884.
The predictor data-base for the years of interest (20042009) is elaborated starting from the nancial statements of each rm included in the sample, for a total of 8030 balance sheets. In particular, we compute nv = 24 indicators selected as potential predictors among
the most relevant in highlighting current and prospective conditions of nancial distress (Altman & Hochkiss, 2006; Dimitras, Zanakis, &
Zopoudinis, 1996). The selected indicators reect the main aspects of the rms' structure such as protability, solvency and liquidity. The
specications of the nancial indicators included in the analysis are shown in Table 1. Non-nancial information on the corporate governance such as region, legal form, number of shareholders, presence of auditors board, rm size and rm age, are also considered.
A pre-processing procedure is performed on the original data set. The results of exploratory data analysis indicate that there are
some accounting data observations which are severe outliers. These observations would seriously distort the estimation results if
they were included in the default risk model. Therefore, those rms showing values of the nancial predictors outside the 3th and
97th percentiles are excluded from the analysis. In order to achieve stability, we apply a modied logarithmic transformation, dened
for non-positive argument.
Finally, the sample is divided into two parts: in-sample set, used for the classication ability, in order to determine how accurately a
model classied businesses, and out-of-sample, used for prediction ability, in order to determine how accurately a model classied
new businesses. Two predictions' windows are considered: 1-year ahead and 2-years ahead.
4. Empirical ndings
4.1. The determinants of different exit routes
This section provides the empirical results obtained from the estimated single-risk and competing risks models. The effect of some
strategic factors on the likelihood of exiting the market for different reasons is investigated and the determinants of various exit routes
are compared.
To better understand the effects of the explanatory variables used in our analysis, we consider a non-parametric test, the Log-rank
test (Kalbeisch & Prentice, 2002), for testing the equality of hazard functions across groups of rms, according to some explanatory
variables. This test is considered as a starting point in order to check which variables have an effect on each exit route and whether
some differences arise not only between the single-risk (where all states of nancial distress are pooled together) and the competing
risks models, but also among the different exit routes.
Table 2 shows the p-values of the log-rank test, for single-risk model (column 5) and for the three competing events (columns 24)
for 1-year ahead predictions.2
1
Following part of the literature, the analysis is focused on a specic economic sector and, as state in the Introduction, we refer to the building sector given its relevance not only in terms of relative weight of GNP but also in terms of socio-economic policy implications.
2
Although the test is performed for both time-windows (1-year ahead and 2-year ahead), here we only report the results of the rst period considered.

A. Amendola et al. / International Review of Economics and Finance 37 (2015) 3341

37

Table 1
Specication of nancial predictors.
Area

nv

Protability

Operational

Structure

Firm-specic variables

Indicators
PM = (PBT / OR) 100
EBEBITDA = R E(EITDA)
EBIT = R E DA
CF / OR
ROE = EA / SE
ROCE = RBIT / CE
ROA = NI / TA
ROTA = (NI + IE + T) / TNA
NAT = OR / (SF + NCL)
IC = OP / IP
ST = OR / STO
COP = (D / OR) 360
CRP = (C / OR) 360
CR = CA / CL
LR = (CA STO) / CL
SLR = SF / NCL
SR = (SF / TA) 100
GE = ((NCL + LO) / SF) 100
Region, legal form, number of shareholders, auditor board, rm size and rm age.

C = Creditors; CA = Current assets; CE = Capital Employed; CF = Cash ow; CL = Current liabilities; COP = Collection period; CR = Current ratio; CRP = Credit
period; D = Debtors; DA = Depreciation & Amortization; E = Expense; EA = Earnings; EBIT = Earnings Before Interest and Tax; EBITDA = Earnings Before
Interest, Taxes, Depreciation and Amortization; EITDA = excluding interest, taxes, depreciation and amortization; GE = Gearing; IC = Interest cover; IE = Interest
Expense; IP = Interest paid; LO = Loans; LR = Liquidity ratio; NAT = Net assets turnover; NCL = Non current liabilities; NI = Net Income; OP = Operating
prot; OR = Operating revenue; PBT = Prot before tax; PM = Prot margin; R = Revenue; RBIT = Return (before Interest and Tax); ROA = Return on
Assets; ROCE = Return on Capital employed; ROE = Return on Equity; ROTA = Return on Total Assets; SE = Shareholders Equity; SF = Shareholders funds;
SLR = Shareholders liquidity ratio; SR = Solvency ratio; ST = Stock turnover; STO = Stock; T = Taxes; TA = Total Assets; TNA = Total Net Asset.

Regarding the risk of exiting the market, the test results show that there are signicant differences in the hazard rates between
groups for most of the variables, looking at both the competing risks and the single-risk models.
Considering the different competing events, some variables are signicant for all the three status. Namely, the legal form, the rm
size, the rst and the last categories of rm age, EBITDA, EBIT, credit period and gearing are signicant for all exit routes with a p-value
less than 0.05. The remaining variables have a different effect in bankruptcy, inactivity and liquidation. For example, the number of
shareholder and the ROCE are only signicant for the inactive state; the stock turnover and the liquidity ratio are important only
for the bankruptcy; the second and the fourth groups of the age and the rms being in Northern Italy are relevant only for the liquidation state. Therefore, it came up that there are some differences in variables inuencing each exit and we cannot neglect them when
predicting the probability of dropping out. Moreover, as expected, these results suggest that bankruptcy, inactive state and liquidation
are three rather different events.
Looking at the results for the risk of pooled exits, it can be noted that the variables relevant for some states in the competing risks
model are also signicant for the single-risk model. In particular, the results are in line with those regarding the risk of being inactive
and liquidated, but they diverge for the risk of being bankrupted.
The variables which are signicant by the log-rank test are considered as the initial set of explanatory variables in the model, in
order to assess their effect on the hazard rate of each exit route. The nal set of variables to be included in each state is further selected
by stepwise procedure.3 The results for the estimated competing-risks and single-risk models are shown in Tables 3 and 4. They respectively display the sign of the coefcients' estimates and the hazard ratios, obtained by computing the exponential of coefcients .
The hazard ratios attest the effect of the covariates on the hazard. A hazard ratio equal to one means that the variable has no effect on
survival, whereas a hazard ratio greater (less) than one indicates that the effect of the covariate is to increase (decrease) the hazard
rate.
In Table 3, columns 35 report the results for the three risks considered and the last column displays the results for the pooled
model. The regression results show some remarkable differences supporting the need to use the competing risk model over the single
risk one. Moreover, the variables are different in the determinants of the three exit routes and in their sign, not only between the competing risks and single-risk models, but also among the states. In particular, looking at the different exit routes, we notice that the legal
form has a positive effect on being bankrupted and liquidated. The medium size and the last category of age (more than 30 years)
have a negative effect on being inactive and liquidated. The rst category of age (up to 7 years) has a positive coefcient for bankruptcy and inactive states. Finally, gearing has a different sign for inactivity and bankruptcy. In fact, it has a negative impact on the probability of being inactive, while it has a positive impact on the probability of going bankrupt.
Some of the variables are selected as potential predictors of exit for only one state. Actually, the geographical division, the number
of shareholders, the prot margin and the collection period are relevant for being inactive. Then the size, the protability ratios, the

For a discussion on the variable selection problem, see Amendola et al. (2011).

38

A. Amendola et al. / International Review of Economics and Finance 37 (2015) 3341

Table 2
Log-rank test for the equality of hazard functions of different exit routes, by explanatory variables, for 1 year ahead.

Limited company
One shareholder
Micro rmsa
Small rms
Medium rms
Large rms
Age1b
Age2
Age3
Age4
Age5
North
Center
South
Return on total assets
Prot margin
EBITDA
EBIT
Cash ow/operating revenue
ROE
ROA
ROCE
Net assets turnover
Interest cover
Stock turnover
Collection period
Credit period
Current ratio
Liquidity ratio
Shareholders liquidity ratio
Solvency ratio
Gearing

Bankruptcy

Inactive

Liquidation

Single-risk

The levels of signicance are: 0.0001, 0.001, 0.01, and 0.05, 0.10, and 1.00.
a
For the rms' size, the European classication is used. Micro rms are those having a number of workers less than 10 and sales less than 2 millions; small rms are
those having a number of workers between 11 and 50 and sales between 2 and 10 millions; medium rms are those having a number of workers between 51 and 250
and sales between 10 and 50 millions; large rms are those having a number of workers greater than 350 and sales greater than 50 millions.
b
The age of rms is classied into 5 classes: 1) up to 7 years; 2) from 8 to 13 years; 3) from 14 to 21 years; 4) from 22 to 30 years; and 5) more than 30 years.

Table 3
Sign of coefcients' estimates for the competing risks and the single risk models, for 1 year ahead.
Variables

Areaa

Bankruptcy

Limited company
South Italy
Micro rms
Small rms
Medium rms
Large rms
One shareholder
Age1
Age4
Age5
Return on total assets
Prot margin
EBIT
Cash ow/operating revenue
ROE
ROCE
Net assets turnover
Interest cover
Collection period
Credit period
Current ratio
Shareholders liquidity ratio
Solvency ratio
Gearing

4
4
4
4
4
4
4
4
4
4
1
1
1
1
1
1
2
2
2
2
3
3
3
3

Positive

Inactive

Liquidation

Single-risk

Positive

Positive

Negative
Positive
Positive

Positive
Negative

Positive

Negative

Positive
Positive
Negative

Negative
Negative
Negative

Negative
Negative
Positive
Positive
Negative

Positive
Negative
Negative
Negative
Positive
Positive

Positive
Negative

Negative

Negative

Negative

Negative
Negative
Negative

Negative
Negative
Negative

Positive

Negative

The numbers from 1 to 4 refer to the protability ratios, operational ratios, structure ratios and rm-specic variables, respectively.

A. Amendola et al. / International Review of Economics and Finance 37 (2015) 3341

39

Table 4
Hazard ratios for the competing risks and the single risk model, for 1 year ahead.
Variables

Area

Limited company
South Italy
Micro rms
Small rms
Medium rms
Large rms
One shareholder
Age1
Age4
Age5
Return on total assets
Prot margin
EBIT
Cash ow/operating revenue
ROE
ROCE
Net assets turnover
Interest cover
Collection period
Credit period
Current ratio
Shareholders liquidity ratio
Solvency ratio
Gearing

4
4
4
4
4
4
4
4
4
4
1
1
1
1
1
1
2
2
2
2
3
3
3
3

Bankruptcy

Inactive

2.0158

Liquidation

Single-risk

3.5913

2.0340

0.5328
71.8679
7.0732

2.8020
0.1030

0.2193

2.2712
1.4888

1.6327

0.3312
0.3862
0.8643

0.4576

0.1421
0.2677
1.5049
1.4769
0.5702

1.0795
0.8752
0.7466
0.9164
1.1457
1.2245

1.1535
0.8942

0.8632

0.9475

0.7292

0.8986
0.8871
0.8994

0.8187
0.8036
0.4602

1.2187

0.9013

interest cover and the current ratio are selected as potential predictors of going bankrupt. Signicant variables for being liquidated are
the third group of age (from 22 to 30 years), return on total assets, net assets turnover, credit period and solvency ratio.
To sum up, the main difference among the competing risks is related to the role of the protability ratios, which are only selected as
predictors of bankruptcy. The reason may be related to their nature of assessing the rm's ability of generating earnings as compared
to its expenses and other relevant costs incurred during a specic period of time. In other words, they measure the company's use of its
assets and control of its expenses to generate an acceptable rate of return. Consequently, since the inactive state includes rms for
which the cause of exit is unknown, we do not have any information about their activity.
The effect of the nancial ratios and the rm-specic variables can be further explored by examining the hazard ratios (Table 4).
It can be noted that the limited companies have a greater probability of being liquidated and going bankrupt. In fact, there is an
increase in the risk of liquidation and bankruptcy, as compared to the rms having other legal forms. Regarding the dimension of
rms and controlling for other rms' characteristics, we note that the risk of being bankrupted is higher for small rms, while the
risk of being liquidated and inactive is lower for medium sized rms. Moreover, young rms have a higher probability of going bankrupt or becoming inactive. On the contrary, old rms have a lower probability of being liquidated and becoming inactive. Then, the
effect of the number of shareholder is to increase the hazard rate of being inactive.
As concerns the nancial indicators of protability area selected as potential predictors of bankruptcy state, they have a negative
effect on the hazard rate, except for the ROCE. The different sign may be related to the fact that the ROCE includes debt funds like loans
and preference capital.

Table 5
Accuracy measures for in-sample.
Bankruptcy

Inactive

Liquidation

Single-risk

20042008
Misclassication
Error II
Error I
AUC
AR

0.38683
0.39492
0.02484
0.91090
0.82180

0.42363
0.42881
0.16552
0.76163
0.52326

0.39823
0.40173
0.09524
0.84261
0.68522

0.39864
0.41319
0.13846
0.80605
0.61211

20042007
Misclassication
Error II
Error I
AUC
AR

0.33213
0.33596
0.14634
0.84980
0.69961

0.40703
0.40885
0.26316
0.74844
0.49689

0.37943
0.38130
0.05714
0.81609
0.63217

0.40719
0.41835
0.12821
0.81837
0.63674

40

A. Amendola et al. / International Review of Economics and Finance 37 (2015) 3341

Table 6
Accuracy measures for out-of-sample.
Bankruptcy

Inactive

Liquidation

Single-risk

2009
Misclassication
Error II
Error I
AUC
AR

0.25113
0.25227
0.00000
0.89930
0.79859

0.29624
0.29892
0.18750
0.79488
0.58975

0.21955
0.21782
0.23729
0.86435
0.72870

0.22105
0.22147
0.21795
0.83672
0.67344

20082009
Misclassication
Error II
Error I
AUC
AR

0.27970
0.27795
0.66667
0.58258
0.16516

0.25564
0.25116
0.43750
0.76714
0.53428

0.20000
0.19142
0.28814
0.83812
0.67623

0.21053
0.19761
0.30769
0.82606
0.65212

Finally, it results that the variables in single-risk framework are substantially different from those in competing risks model. In particular, very small companies have a faster hazard timing of exit the market, as opposite to big ones. Moreover, the young rms have a
higher probability of leaving the market, compared to the older ones. Then, other important characteristics that determine an increase
of the hazard rate are the legal form and the number of shareholders. Furthermore, it is important to notice that the protability ratios
are not selected as potential predictors in single-risk, as shown before. The results conrm most of the main theoretical hypothesis set
up in the default risk models and are mainly in line with those found in the empirical literature (Bellovary, Giacomino, & Akers, 2007;
Esteve-Prez et al., 2010; Rommer, 2005).
4.2. Predictive performance
After selecting the most relevant variables for the competing events and the single risk, we evaluate the predictive performance of
the developed models by means of some accuracy measures. Firstly, we compute the type I error, i.e. a failing rm is misclassied as a
non-failing rm, and the type II error, i.e. a non-failing rm is wrongly assigned to the failing group. Then, as total misclassication
error, we consider the Miss-Classication errors (type I and II), the Area under the ROC curve (AUC) and the Accuracy Ratio (AR)
(Engelmann, Hayden, & Tasche, 2003). These measures are displayed in Tables 5 and 6 for in-sample and out-of-sample sets with respect to the two time-windows (1-year ahead and 2-years ahead). Looking at the in-sample set, used for testing the classication ability of the models, it can be noted that the correct classication rate is slightly higher for bankruptcy and liquidation than the single-risk
and it does not increase at approaching the year of exit, while the inactive state's rate is lower than the one of both the other two states
and pooled model. Moreover, the AUC, which considers the impact of the I and II type errors, has a very discriminative power for bankruptcy and liquidation, compared to the single-risk framework, and its power increases as the failure time is approaching.
For evaluating the predictive power of the models, we also compute the above-mentioned measures on the out-of-sample set.
Looking at the results displayed in Table 6, we notice that even though the correct classication rate is higher for the single-risk
model than for the other three states, the competing risks model has a better performance than the single-risk framework in terms
of AUC. Therefore, we can conclude that the competing-risks model produce an improvement for predicting the failure of rm.
5. Conclusions
In this study the competing risks model was estimated for investigating the differences and the effects of potential predictors on
the probability of rms' exit the market for different reasons and it was compared to the single-risk model, in which all nancial distress states are pooled together. A dataset of nancial statements of a sample of Italian rms was analyzed and three different exit
routes were considered: bankruptcy, liquidation and inactivity. The starting point was to test the signicance level of the microeconomic and rm-specic variables on all the three states by means of the log-rank test. Then, the variables inuencing the exit
route were selected by the stepwise procedure. After that, we analyzed the sign and the inuence of the estimated coefcients on
each exit route and evaluated the predictive performance of the competing-risks model over the single-risk framework, at two
time horizons, by considering some accuracy measures.
The results discovered some differences in rm-specic variables and micro-economic factors determining rm exit with respect
to the different routes. In particular, we found out that some rm-specic characteristics, such as age, legal form and size inuence the
probability of being liquidated, inactive and bankrupted, conrming the empirical results available in literature. Moreover, it can be
noted the important role of the protability ratios on the likelihood of going bankrupt. Then, for the single-risk model the variables
selected are quite similar to those characterizing the inactive state and liquidation, while they are different for bankruptcy. Therefore,
our results gave evidence in favor of distinguishing among bankruptcy, inactivity and liquidation as three different forms of exit and
give ground for the use of multiple-state models. Further research could include the development of more appropriate procedure of
selecting the variables in the competing risks framework.

A. Amendola et al. / International Review of Economics and Finance 37 (2015) 3341

41

References
Altman, E. I. (1968). The prediction of corporate bankruptcy: a discriminant analysis. Journal of Finance, 23, 193194.
Altman, E. I., & Hochkiss, E. (2006). Corporate Financial Distress and Bankruptcy: Predict and Avoid Bankruptcy, Analyze and Invest in Distressed Debt. New York: John
Wiley & Sons.
Amendola, A., Restaino, M., & Sensini, L. (2011). Variable selection in default risk model. Journal of Risk Model Validation, 5, 319.
Andersen, P. K., Abildstrm, S. Z., & Rosthj, S. (2002). Competing risks as a multi-state model. Statistical Methods in Medical Research, 11, 203215.
Andersen, P. K., Borgan, ., Gill, R. D., & Keiding, N. (1993). Statistical Models based on Counting Processes. Berlin: Springer.
Balcaen, S., & Ooghe, H. (2006). 35 years of studies on business failure: an overview of the classical statistical methodologies and their related problems. The British
Accounting Review, 38, 6393.
Bellovary, J., Giacomino, D., & Akers, M. (2007). A review of bankruptcy prediction studies: 1930 to present. Journal of Financial Education, 33, 143.
Bhattacharjee, A., Higson, C., Holly, S., & Kattuman, P. (2009). Macroeconomic instability and business exit: determinants of failures and acquisitions of UK firms.
Economica, 76, 108131.
Brabazon, A., & Keenan, P. B. (2004). A hybrid genetic model for the prediction of corporate failure. Computational Management Science, 1, 293310.
Chancharat, N., Tian, G., Davy, P., McCrae, M., & Lodh, S. (2010). Multiples states of financially distressed companies: tests using a competing-risks model. Australasian
Accounting Business and Finance Journal, 4, 2744.
Dimitras, A., Zanakis, S., & Zopoudinis, C. (1996). A survey of business failures with an emphasis on failure prediction methods and industrial applications. European
Journal of Operational Research, 90(3), 487513.
Engelmann, B., Hayden, E., & Tasche, D. (2003). Testing rating accuracy. Risk, 16, 8286.
Esteve-Prez, S., Sanchis-Llopis, A., & Sanchis-Llopis, J. A. (2010). A competing risks analysis of firms' exit. Empirical Economics, 38, 281304.
Figlewski, S., Frydman, H., & Liang, W. (2012). Modeling the effect of macroeconomic factors on corporate default and credit rating transitions. International Review of
Economics and Finance, 21, 87105.
Fitzpatrick, P. J. (1931). Symptoms of Industrial Failures as Revealed by an Analysis of the Financial Statements of the Failed Companies. Washington, D.C.: Catholic
University of America.
Fitzpatrick, P. J. (1932). A comparison of the ratios successful industrial enterprises with those failed companies. The Certified Public Accountant, 12(598-605), 656662
(727-31).
Harhoff, D., Stahl, K., & Woywode, M. (1998). Legal form, growth and exit of West German firms: empirical results for manufacturing, construction, trade and service
industries. The Journal of Industrial Economics, 46, 453488.
Headd, B. (2003). Redefining business success: distinguishing between closure and failure. Small Business Economics, 21, 5161.
Hensher, D. A., & Jones, S. (2007). Forecasting corporate bankruptcy: optimizing the performance of the mixed logit model. Abacus, 43, 241264.
Hougaard, P. (2000). Analysis of Multivariate Survival Data. Statistics for Biology and Health. New York: Springer.
Jones, S., & Hensher, D. A. (2004). Predicting firm financial distress: a mixed logit model. The Accounting Review, 79, 10111038.
Jones, S., & Hensher, D. A. (2007). Modelling corporate failure: a multinomial nested logit analysis for unordered outcomes. The British Accounting Review, 39, 89107.
Kalbfleisch, J. D., & Prentice, R. L. (2002). The Statistical Analysis of Failure Time Data. New York: Wiley.
Laitinen, E. K., & Suvas, A. (2013). International applicability of corporate failure risk models based on financial statement information: comparisons across European
countries. Journal of Finance & Economics, 1, 126.
Lennox, C. (1999). Identifying failing companies: a re-evaluation of the logit, probit and DA approaches. Journal of Economics and Business, 51, 347364.
Ohlson, J. A. (1980). Financial ratios and the probabilistic prediction of bankruptcy. Journal of Accounting Research, 18, 109131.
Ravi Kumar, P., & Ravi, V. (2007). Bankruptcy prediction in banks and firms via statistical and intelligent techniquesa review. European Journal of Operational Research,
180, 128.
Rommer, A. D. (2005). A comparative analysis of the determinants of financial distress in French, Italian and Spanish firms. Working paper no.24. (pp. 176)Danmarks
Nationalbank.
Schary, M. (1991). The probability of exit. RAND Journal of Economics, 22, 339353.
Shumway, T. (2001). Forecasting bankruptcy more accurately: a simple hazard model. Journal of Business, 74, 101124.
Smith, R. F., & Winakor, A. H. (1930). A test of analysis of unsuccessful industrial companies. Bureau of Business Research, Bulletin no. 31Chicago: University of Illinois.
Xie, Y. A., Shi, J., & Wu, C. (2008). Do macroeconomic variables matter for pricing default risk. International Review of Economics and Finance, 17, 279291.
Zeitun, R. (2009). Ownership structure, corporate performance and failure: evidence from panel data of emerging market the case of Jordan. Corporate Ownership and
Control, 6(4), 96114.
Zmijewski, M. E. (1984). Methodological issues related to the estimation of financial distress prediction models. Journal of Accounting Research, 22, 5982.