Professional Documents
Culture Documents
Modele LR
Modele LR
r
m
s
W
e
a
k
r
m
s
C
o
n
t
r
o
l
r
m
s
T
o
t
a
l
r
m
s
S
t
r
o
n
g
r
m
s
W
e
a
k
r
m
s
C
o
n
t
r
o
l
r
m
s
T
o
t
a
l
r
m
s
C
o
n
s
t
a
n
t
2
1
5
.
3
2
2
1
1
.
3
4
*
2
1
7
.
5
0
2
1
0
.
9
8
*
2
6
.
8
7
2
2
.
4
2
2
2
2
.
1
3
*
2
1
0
.
8
1
*
(
1
1
.
4
0
)
(
5
.
5
6
)
(
1
7
.
1
0
)
(
2
.
4
8
)
(
9
.
4
9
)
(
3
.
3
0
)
(
1
2
.
9
4
)
(
3
.
2
5
)
C
A
S
H
1
7
.
4
7
1
9
.
3
4
1
5
.
6
5
2
7
.
4
7
2
0
.
4
3
2
5
.
5
8
2
2
2
.
3
9
2
3
.
6
3
(
2
5
.
0
6
)
(
1
2
.
1
6
)
(
4
1
.
0
6
)
(
7
.
4
9
)
(
2
0
.
0
6
)
(
9
.
8
0
)
(
1
8
.
1
9
)
(
7
.
6
4
)
N
I
N
2
7
1
.
3
4
6
.
3
8
2
1
4
2
.
0
7
*
1
.
6
3
2
6
.
9
6
9
.
0
1
2
0
.
9
8
2
1
5
.
6
4
(
5
1
.
7
5
)
(
9
.
7
0
)
(
8
0
.
0
5
)
(
9
.
2
3
)
(
2
7
.
9
1
)
(
1
3
.
0
8
)
(
2
2
.
6
8
)
(
1
0
.
7
4
)
L
E
V
1
5
.
9
2
1
1
.
7
8
*
2
0
.
1
9
6
.
3
1
*
1
2
.
1
8
5
.
9
9
2
1
.
5
5
8
.
4
9
*
(
1
2
.
2
6
)
(
5
.
8
5
)
(
1
9
.
0
0
)
(
2
.
6
8
)
(
7
.
8
7
)
(
3
.
7
2
)
(
5
.
6
1
)
(
2
.
7
1
)
W
C
2
1
.
3
2
2
4
.
4
9
2
0
.
6
1
2
3
.
6
2
8
.
2
9
5
.
6
1
2
3
0
.
7
2
1
.
7
4
(
1
0
.
7
8
)
(
4
.
6
1
)
(
1
2
.
6
6
)
(
2
.
7
2
)
(
1
4
.
7
6
)
(
4
.
6
7
)
(
2
0
.
4
5
)
(
4
.
1
8
)
L
I
Q
3
.
1
4
1
.
6
5
4
.
1
1
1
.
6
8
*
2
2
.
4
4
2
2
.
2
5
2
3
.
5
6
2
0
.
3
0
(
4
.
6
5
)
(
1
.
6
4
)
(
4
.
3
9
)
(
0
.
8
7
)
(
7
.
7
5
)
(
1
.
9
0
)
(
1
3
.
8
8
)
(
2
.
2
1
)
N
C
R
E
2
.
6
6
0
.
8
2
0
.
8
2
2
0
.
3
8
2
2
.
0
4
2
1
.
5
3
2
3
.
2
4
2
0
.
2
8
(
4
.
1
3
)
(
1
.
2
2
)
(
5
.
2
3
)
(
0
.
9
0
)
(
5
.
2
6
)
(
2
.
2
7
)
(
3
.
7
6
)
(
1
.
0
7
)
2
2
l
o
g
L
2
2
.
6
1
1
6
.
4
2
3
1
.
8
2
5
8
.
6
9
2
4
.
0
5
9
.
3
2
1
4
.
3
8
6
5
.
7
5
x
2
p
-
v
a
l
u
e
0
.
0
0
0
0
.
0
1
1
7
0
.
0
0
0
0
0
.
0
0
0
0
0
.
0
0
0
5
0
.
1
5
6
5
0
.
0
2
5
7
0
.
0
0
0
0
M
c
F
a
d
d
e
n
R
2
0
.
4
7
9
7
0
.
3
4
8
4
0
.
6
7
5
1
0
.
2
9
4
3
0
.
5
1
0
3
0
.
1
9
7
7
0
.
3
0
5
1
0
.
3
3
1
9
A
c
c
u
r
a
c
y
i
n
-
s
a
m
p
l
e
P
e
r
c
e
n
t
a
g
e
c
l
a
s
s
i
e
d
8
5
.
2
9
7
9
.
4
1
9
1
.
1
8
9
9
.
4
1
8
8
.
2
4
6
4
.
7
1
7
6
.
4
7
9
9
.
3
0
T
y
p
e
I
e
r
r
o
r
(
p
e
r
c
e
n
t
)
1
7
.
6
5
2
3
.
5
3
1
1
.
7
6
0
.
0
0
1
1
.
7
6
4
1
.
1
8
2
3
.
5
3
0
.
0
5
T
y
p
e
I
I
e
r
r
o
r
(
p
e
r
c
e
n
t
)
1
1
.
7
6
1
7
.
6
5
5
.
8
8
7
6
.
4
7
1
1
.
7
6
2
9
.
4
1
2
3
.
5
3
8
2
.
3
5
A
R
C
A
P
0
.
9
2
9
8
0
.
8
7
2
8
0
.
9
1
4
0
0
.
7
8
7
9
0
.
9
5
2
4
0
.
7
9
8
1
0
.
8
5
0
6
0
.
7
8
2
9
A
R
R
O
C
0
.
7
6
2
5
0
.
7
5
8
6
0
.
7
1
6
5
0
.
3
0
7
9
0
.
7
9
0
1
0
.
5
7
3
3
0
.
8
4
6
1
0
.
3
1
9
9
N
o
t
e
s
:
T
h
e
s
a
m
p
l
e
i
s
c
o
m
p
o
s
e
d
o
f
1
7
b
a
n
k
r
u
p
t
r
m
s
.
T
h
e
d
e
p
e
n
d
e
n
t
v
a
r
i
a
b
l
e
t
a
k
e
s
t
h
e
v
a
l
u
e
o
n
e
f
o
r
t
h
e
1
7
b
a
n
k
r
u
p
t
r
m
s
a
n
d
z
e
r
o
f
o
r
t
h
e
n
o
n
-
b
a
n
k
r
u
p
t
r
m
s
.
S
t
a
n
d
a
r
d
e
r
r
o
r
s
a
r
e
g
i
v
e
n
i
n
p
a
r
e
n
t
h
e
s
e
s
.
*
S
i
g
n
i
c
a
n
t
a
t
t
h
e
5
p
e
r
c
e
n
t
l
e
v
e
l
Table I.
Estimates of the logit
target prediction model
based on the Beaver
(1966) variables
The bias of
unhealthy SMEs
69
behaves likewise. The percentage of correctly classied companies during the
estimation period is relatively high using the strong rm and the control rm samples.
The results for the total rmsample are over-valued due to bankrupt rms having a very
small weight in the total sample. On xing a cut-off equal to 0.5, the weak rm sample
has high type I and II errors. In the total sample, the type I error is insignicant while the
type II error is very high. The CAPand ROCmeasures indicate that a strong rmsample
allows the model to be estimated with the best in-sample predictive capacity.
The results concerning the Altman (1968), Altman et al. (1977) and Garc a et al.
(1997) models are likewise summarised in Tables II, III, and IV, respectively.
In general, Tables II IV illustrate that the strong rm and control rm samples
have the highest McFaddens R
2
in 2001 while the strong rm sample does so in 2002.
The percentage of correctly classied companies during the estimation period is
relatively high using the strong rms, from 82.35 per cent to 88.24 per cent. The CAP
and ROC measures point out that the strong rm sample allows the model to be
estimated with the best in-sample predictive capacity with values up to 0.9751. The
total rm sample has the highest type II error, up to 99.41 per cent, which means that to
predict that no rm is going to be bankrupt minimises the total error, since it
represents a reduced percentage of the total number of rms used in the estimation
process. On the other hand, the control rm sample shows different results depending
on both the year and the model. In this sense, the percentage correctly classied goes
from 64.71 per cent to 91.18 per cent, the type II error goes from 5.88 per cent to 35.29
per cent and the type I error goes from 5.88 per cent to 41.18 per cent. The CAP and
ROC values of the control rm sample are lower than the strong rm sample and
higher than the weak rm sample.
The out-of-sample predictive capacity of the models is reported in Tables V and VI.
When the total sample is considered to estimate the model, out-of-sample results
cannot be computed in the estimation year.
Concerning the strong rm sample, Table V reports that, in 2001, the type I error
ranges from 24.38 per cent to 29.72 per cent, the type II error is equal to 11.76 per cent in
all models and the CAP values go from 0.871 to 0.893 while the ROC values go from
0.7423 to 0.8520. These values do not worsen when the model is applied one or two
years after the estimation. In the worst case, the type I error rises to 33.35 per cent, the
type II error rises to 42.86 per cent, the CAP decreases to 0.801 and the ROC rises to
0.662. In general, the weak rm sample gives higher type I and II errors and lower ROC
and CAP values than the strong rm and the control rm samples. The control rm
sample has higher values of CAP and ROC as well as lower type I and II errors than the
weak rm sample. In contrast, the total rm sample has the highest type II error.
Table VI shows that, as in 2001, in 2002 the strong rm sample shows the most
accuracy. Type I error in 2002 goes from 28.16 per cent to 30.76 per cent, type II error
ranges from 5.88 per cent to 11.76 per cent in all the models and the CAP values go from
0.872 to 0.88 while ROC values go from 0.79 to 0.814. The results remain similar in 2003
and 2004. In the worst case, the type I error is 28.85 per cent, the type II error rises to 28.57
per cent, the CAP decreases to 0.785 and the ROC rises to 0.712. In general, the weak rm
sample gives higher type I and II errors and lower ROC and CAP values than the strong
rmandthe control rmsamples. It must be highlighted that the control rmsample has
higher values of CAPand ROCandlower type I and II errors than the weak rmsamples.
In contrast, the total rm sample has the highest type II error.
JSBED
17,1
70
2
0
0
1
2
0
0
2
S
t
r
o
n
g
r
m
s
W
e
a
k
r
m
s
C
o
n
t
r
o
l
r
m
s
T
o
t
a
l
r
m
s
S
t
r
o
n
g
r
m
s
W
e
a
k
r
m
s
C
o
n
t
r
o
l
r
m
s
T
o
t
a
l
r
m
s
C
o
n
s
t
a
n
t
2
.
2
4
*
0
.
7
5
5
.
2
2
*
2
3
.
0
6
*
1
.
6
0
2
1
.
5
1
0
.
1
9
2
2
.
7
6
*
(
1
.
2
8
)
(
0
.
9
2
)
(
2
.
2
1
)
(
2
.
2
1
)
(
1
.
5
8
)
(
1
.
1
0
)
(
1
.
1
6
)
(
0
.
6
4
)
W
C
1
.
8
2
2
1
.
6
2
1
.
6
2
2
2
.
1
6
4
.
2
0
5
.
3
2
3
.
2
1
1
.
0
0
(
3
.
0
7
)
(
2
.
1
4
)
(
7
.
0
2
)
(
1
.
4
4
)
(
4
.
4
8
)
(
3
.
4
7
)
(
2
.
6
4
)
(
1
.
8
1
)
C
P
2
0
.
3
5
1
.
7
7
2
2
.
9
7
2
.
3
4
2
9
.
8
7
9
.
8
2
*
0
.
6
3
1
.
1
5
(
7
.
1
5
)
(
3
.
6
7
)
(
8
.
1
5
)
(
2
.
7
2
)
(
9
.
1
3
)
(
5
.
9
0
)
(
6
.
5
3
)
(
3
.
5
6
)
R
O
A
2
1
1
.
3
0
4
.
4
3
2
3
4
.
8
3
*
2
8
.
1
4
*
2
1
0
.
5
8
2
0
.
1
2
2
7
.
6
9
2
1
2
.
2
3
*
(
9
.
9
6
)
(
3
.
0
8
)
(
1
7
.
6
0
)
(
3
.
7
0
)
(
7
.
8
8
)
(
4
.
9
4
)
(
5
.
9
7
)
(
4
.
1
6
)
C
A
P
2
6
.
0
2
*
2
3
.
3
5
2
6
.
0
4
2
3
.
9
0
*
2
5
.
0
1
2
1
0
.
1
8
*
2
2
.
5
9
2
5
.
9
7
*
(
3
.
3
0
)
(
2
.
8
0
)
(
5
.
0
6
)
(
1
.
2
4
)
(
3
.
0
8
)
(
4
.
2
9
)
(
2
.
3
9
)
(
2
.
2
5
)
A
C
T
0
.
0
4
2
0
.
0
3
2
0
.
6
2
2
0
.
1
7
1
.
0
9
2
.
0
6
*
0
.
3
1
0
.
1
1
(
0
.
4
5
)
(
0
.
3
6
)
(
0
.
4
7
)
(
0
.
3
0
)
(
0
.
8
2
)
(
0
.
8
9
)
(
0
.
5
1
)
(
0
.
2
6
)
2
2
l
o
g
L
2
0
.
2
9
7
.
8
0
2
7
.
4
2
5
4
.
8
2
2
5
.
3
7
1
4
.
1
2
7
.
2
2
5
2
.
5
1
x
2
p
-
v
a
l
u
e
0
.
0
0
1
1
0
.
1
6
7
4
0
.
0
0
0
0
0
.
0
0
0
0
0
.
0
0
0
1
0
.
0
1
4
8
0
.
2
0
4
6
0
.
0
0
0
0
M
c
F
a
d
d
e
n
R
2
0
.
4
4
3
0
5
0
.
1
6
5
5
0
.
5
8
1
7
0
.
2
7
4
9
0
.
5
3
8
3
0
.
2
9
9
7
0
.
1
5
3
2
0
.
2
6
5
0
A
c
c
u
r
a
c
y
i
n
-
s
a
m
p
l
e
P
e
r
c
e
n
t
a
g
e
c
l
a
s
s
i
e
d
8
2
.
3
5
7
0
.
5
9
8
8
.
2
4
9
9
.
3
2
8
8
.
2
4
7
6
.
4
7
6
4
.
7
1
9
9
.
3
0
T
y
p
e
I
e
r
r
o
r
(
p
e
r
c
e
n
t
)
2
3
.
5
3
3
5
.
2
9
1
7
.
6
5
0
.
6
4
1
7
.
6
5
2
3
.
5
3
4
1
.
1
8
0
.
0
0
T
y
p
e
I
I
e
r
r
o
r
(
p
e
r
c
e
n
t
)
1
1
.
7
6
2
3
.
5
3
5
.
8
8
8
3
.
0
3
5
.
8
8
2
3
.
5
3
2
9
.
4
1
8
8
.
2
4
A
R
C
A
P
0
.
9
2
3
1
0
.
8
4
6
1
0
.
8
6
8
3
0
.
7
8
1
0
0
.
9
5
0
2
0
.
7
3
0
3
0
.
8
9
8
0
0
.
7
6
8
7
A
R
R
O
C
0
.
7
4
2
4
0
.
6
0
0
7
0
.
7
1
7
1
0
.
2
7
0
2
0
.
8
0
2
2
0
.
4
9
5
9
0
.
6
7
2
2
0
.
2
4
6
9
N
o
t
e
s
:
T
h
e
s
a
m
p
l
e
w
a
s
c
o
m
p
o
s
e
d
o
f
1
7
b
a
n
k
r
u
p
t
r
m
s
.
T
h
e
d
e
p
e
n
d
e
n
t
v
a
r
i
a
b
l
e
t
a
k
e
s
t
h
e
v
a
l
u
e
1
f
o
r
t
h
e
1
7
b
a
n
k
r
u
p
t
r
m
s
a
n
d
0
f
o
r
t
h
e
n
o
n
-
b
a
n
k
r
u
p
t
r
m
s
.
S
t
a
n
d
a
r
d
e
r
r
o
r
s
a
r
e
s
h
o
w
n
i
n
p
a
r
e
n
t
h
e
s
e
s
.
*
S
i
g
n
i
c
a
n
t
a
t
t
h
e
5
p
e
r
c
e
n
t
l
e
v
e
l
Table II.
Estimates of the logit
target prediction model
based on the Altman
(1968) variables
The bias of
unhealthy SMEs
71
2
0
0
1
2
0
0
2
S
t
r
o
n
g
r
m
s
W
e
a
k
r
m
s
C
o
n
t
r
o
l
r
m
s
T
o
t
a
l
r
m
s
S
t
r
o
n
g
r
m
s
W
e
a
k
r
m
s
C
o
n
t
r
o
l
r
m
s
T
o
t
a
l
r
m
s
C
o
n
s
t
a
n
t
1
6
.
7
7
2
0
.
6
5
7
.
2
9
2
1
.
5
4
4
.
3
6
1
3
.
7
1
1
2
.
9
6
*
0
.
2
6
(
1
1
.
6
9
)
(
6
.
8
3
)
(
9
.
4
6
)
(
3
.
4
6
)
(
8
.
1
5
)
(
8
.
3
5
)
(
7
.
0
8
)
(
3
.
3
5
)
R
O
A
2
4
.
9
5
2
2
.
8
3
*
2
3
9
.
0
9
*
2
5
.
4
5
2
2
.
8
2
1
7
.
1
2
2
6
.
8
7
2
9
.
0
3
*
(
2
0
.
0
3
)
(
1
0
.
3
6
)
(
1
9
.
4
2
)
(
3
.
9
8
)
(
8
.
9
0
)
(
1
0
.
6
2
)
(
7
.
0
8
)
(
4
.
5
1
)
S
R
O
A
1
7
4
.
3
1
3
0
.
0
8
2
2
3
.
5
3
7
.
0
9
3
6
.
2
8
5
3
.
5
1
*
1
2
.
0
4
1
8
.
0
4
*
(
1
1
3
.
6
1
)
(
2
1
.
0
6
)
(
2
4
.
4
6
)
(
4
.
8
8
)
(
2
9
.
7
0
)
(
2
2
.
7
8
)
(
1
4
.
4
5
)
(
7
.
4
2
)
D
S
1
.
2
3
0
.
2
2
*
0
.
1
3
0
.
0
3
0
.
3
3
0
.
0
1
0
.
4
0
0
.
0
2
(
1
.
0
4
)
(
0
.
1
2
)
(
0
.
3
2
)
(
0
.
0
8
)
(
0
.
4
1
)
(
0
.
1
3
)
(
0
.
3
1
)
(
0
.
0
9
)
C
P
9
.
7
5
5
.
4
0
2
7
.
3
7
5
.
4
2
2
1
0
.
2
7
2
.
7
8
2
1
.
6
4
0
.
3
0
(
1
1
.
2
3
)
(
5
.
7
7
)
(
8
.
9
0
)
(
3
.
3
1
)
(
9
.
8
7
)
(
7
.
1
8
)
(
7
.
7
0
)
(
3
.
6
0
)
L
I
Q
2
4
.
8
1
2
0
.
6
3
0
.
0
4
0
.
1
4
6
.
1
5
2
.
0
3
6
.
3
2
1
.
7
9
(
1
6
.
4
5
)
(
2
.
6
8
)
(
9
.
5
1
)
(
1
.
8
0
)
(
5
.
6
2
)
(
4
.
3
4
)
(
3
.
7
4
)
(
1
.
8
6
)
C
A
P
2
5
6
.
3
1
2
1
3
.
1
4
*
2
8
.
3
1
2
1
2
.
4
2
*
2
1
2
.
0
4
*
2
1
2
.
1
4
*
2
7
.
0
9
2
9
.
8
8
*
(
3
5
.
1
1
)
(
6
.
3
4
)
(
1
1
.
6
1
)
(
3
.
4
7
)
(
6
.
8
7
)
(
7
.
3
3
)
(
5
.
1
6
)
(
3
.
3
4
)
S
I
Z
E
2
1
.
4
7
0
.
1
2
2
0
.
1
7
2
0
.
1
5
2
0
.
1
9
2
1
.
4
5
*
2
1
.
3
2
*
2
0
.
3
6
(
1
.
1
7
)
(
0
.
7
1
)
(
1
.
0
1
)
(
0
.
3
6
)
(
0
.
9
1
)
(
0
.
8
6
)
(
0
.
7
5
)
(
0
.
3
7
)
2
2
l
o
g
L
3
2
.
2
1
1
7
.
9
3
2
7
.
5
7
6
2
.
2
2
2
7
.
1
2
2
2
.
0
3
1
3
.
6
6
6
1
.
8
6
x
2
p
-
v
a
l
u
e
0
.
0
0
0
0
0
.
0
1
2
0
.
0
0
0
3
0
.
0
0
0
0
0
.
0
0
0
3
0
.
0
0
2
5
0
.
0
5
7
5
0
.
0
0
0
0
M
c
F
a
d
d
e
n
R
2
0
.
6
8
3
4
0
.
3
8
0
5
0
.
5
8
5
0
0
.
3
1
2
0
0
.
5
7
5
4
0
.
4
6
7
3
0
.
2
8
9
9
0
.
3
1
2
2
A
c
c
u
r
a
c
y
i
n
-
s
a
m
p
l
e
P
e
r
c
e
n
t
a
g
e
c
l
a
s
s
i
e
d
8
5
.
2
9
7
6
.
4
7
8
5
.
2
9
9
9
.
4
1
8
5
.
2
9
8
2
.
3
5
7
3
.
5
3
9
9
.
3
4
T
y
p
e
I
e
r
r
o
r
(
p
e
r
c
e
n
t
)
1
7
.
6
5
2
9
.
4
1
1
7
.
6
5
0
.
0
0
1
7
.
6
5
1
7
.
6
5
2
9
.
4
1
0
.
0
0
T
y
p
e
I
I
e
r
r
o
r
(
p
e
r
c
e
n
t
)
1
1
.
7
6
1
7
.
6
5
1
1
.
7
6
7
6
.
4
7
1
1
.
7
6
1
7
.
6
5
2
3
.
5
3
8
2
.
3
5
A
R
C
A
P
0
.
9
7
5
1
0
.
8
6
8
2
0
.
9
1
6
2
0
.
7
9
6
7
0
.
9
6
6
1
0
.
7
9
3
6
0
.
9
3
8
9
0
.
7
8
5
3
A
R
R
O
C
0
.
7
4
2
4
0
.
7
1
5
4
0
.
7
1
6
6
0
.
3
1
3
0
0
.
8
1
5
4
0
.
5
7
0
6
0
.
7
5
5
5
0
.
2
9
7
6
N
o
t
e
s
:
T
h
e
s
a
m
p
l
e
w
a
s
c
o
m
p
o
s
e
d
o
f
1
7
b
a
n
k
r
u
p
t
r
m
s
.
T
h
e
d
e
p
e
n
d
e
n
t
v
a
r
i
a
b
l
e
t
a
k
e
s
t
h
e
v
a
l
u
e
1
f
o
r
t
h
e
1
7
b
a
n
k
r
u
p
t
r
m
s
a
n
d
0
f
o
r
t
h
e
n
o
n
-
b
a
n
k
r
u
p
t
r
m
s
.
S
t
a
n
d
a
r
d
e
r
r
o
r
s
a
r
e
s
h
o
w
n
i
n
p
a
r
e
n
t
h
e
s
e
s
.
*
S
i
g
n
i
c
a
n
t
a
t
t
h
e
5
p
e
r
c
e
n
t
l
e
v
e
l
Table III.
Estimates of the logit
target prediction model
based on the Altman et al.
(1977) variables
JSBED
17,1
72
2
0
0
1
2
0
0
2
S
t
r
o
n
g
r
m
s
W
e
a
k
r
m
s
C
o
n
t
r
o
l
r
m
s
T
o
t
a
l
r
m
s
S
t
r
o
n
g
r
m
s
W
e
a
k
r
m
s
C
o
n
t
r
o
l
r
m
s
T
o
t
a
l
r
m
s
C
o
n
s
t
a
n
t
6
.
9
7
*
5
.
9
7
*
1
0
.
0
5
3
.
3
8
*
6
.
4
6
*
4
.
7
2
*
2
.
4
8
1
.
5
6
(
4
.
2
2
)
(
2
.
7
6
)
(
6
.
7
8
)
(
1
.
9
4
)
(
3
.
1
8
)
(
2
.
3
7
)
(
2
.
3
1
)
(
1
.
5
5
)
Q
L
I
Q
1
.
8
5
2
0
.
1
3
4
.
9
8
2
0
.
6
3
1
.
7
7
0
.
4
5
2
2
.
1
6
2
1
.
1
4
(
2
.
2
7
)
(
1
.
4
3
)
(
3
.
9
2
)
(
0
.
9
9
)
(
1
.
9
8
)
(
1
.
5
3
)
(
1
.
5
1
)
(
0
.
9
8
)
A
T
P
2
8
.
7
8
*
2
3
.
9
4
2
1
0
.
8
5
2
5
.
9
6
2
6
.
7
7
*
2
2
.
1
5
2
2
.
7
6
2
3
.
9
3
*
(
4
.
3
8
)
(
2
.
4
2
)
(
6
.
8
0
)
(
1
.
7
1
)
(
2
.
7
1
)
(
1
.
7
6
)
(
2
.
0
0
)
(
1
.
1
1
)
I
N
T
4
0
.
1
4
2
4
.
8
9
1
1
3
.
7
2
*
1
6
.
1
2
*
9
3
.
7
5
2
2
1
.
3
4
6
.
9
1
1
8
.
0
5
*
(
3
5
.
3
8
)
(
8
.
2
4
)
(
6
3
.
2
5
)
(
6
.
2
6
)
(
5
8
.
4
0
)
(
2
0
.
0
9
)
(
2
0
.
7
1
)
(
1
0
.
1
5
)
A
M
O
2
6
.
5
5
2
4
.
5
0
2
.
1
8
2
0
.
0
6
2
0
.
4
1
2
.
8
8
4
.
2
9
2
0
.
8
8
(
1
7
.
9
7
)
(
5
.
2
1
)
(
1
2
.
5
6
)
(
3
.
1
8
)
(
6
.
5
7
)
(
4
.
2
7
)
(
5
.
2
9
)
(
2
.
9
6
)
E
A
R
2
1
6
.
1
5
8
.
1
1
*
2
2
8
.
6
3
2
6
.
8
2
*
2
1
.
1
9
2
.
8
5
2
1
.
6
6
2
9
.
7
4
*
(
9
.
8
6
)
(
4
.
1
3
)
(
1
6
.
4
4
)
(
3
.
0
7
)
(
5
.
6
1
)
(
3
.
3
0
)
(
4
.
4
9
)
(
3
.
1
3
)
2
2
l
o
g
L
2
6
.
4
9
1
0
.
5
0
3
1
.
8
5
6
1
.
0
4
2
5
.
9
4
9
.
5
5
5
.
6
9
5
2
.
7
7
x
2
p
-
v
a
l
u
e
0
.
0
0
0
0
0
.
0
6
2
0
.
0
0
0
0
0
.
0
0
0
0
0
.
0
0
0
0
0
0
.
0
8
9
1
0
.
3
3
7
2
0
.
0
0
0
0
M
c
F
a
d
e
n
R
2
0
.
5
6
2
0
0
.
2
2
2
9
0
.
6
7
5
7
0
.
3
0
6
1
0
.
5
5
0
4
0
.
2
0
2
5
0
.
1
2
0
8
0
.
2
6
6
4
A
c
c
u
r
a
c
y
i
n
-
s
a
m
p
l
e
P
e
r
c
e
n
t
a
g
e
c
l
a
s
s
i
e
d
8
8
.
2
4
6
1
.
7
6
9
1
.
1
8
9
9
.
2
8
8
8
.
2
4
6
7
.
6
5
6
4
.
7
1
9
9
.
3
0
T
y
p
e
I
e
r
r
o
r
(
p
e
r
c
e
n
t
)
1
1
.
7
6
3
5
.
2
9
5
.
8
8
0
.
0
5
1
7
.
6
5
3
5
.
2
9
3
5
.
2
9
0
.
0
0
T
y
p
e
I
I
e
r
r
o
r
(
p
e
r
c
e
n
t
)
1
1
.
7
6
4
1
.
1
8
1
1
.
7
6
8
8
.
2
4
5
.
8
8
2
9
.
4
1
3
5
.
2
9
8
8
.
2
4
A
R
C
A
P
0
.
9
5
9
2
0
.
8
5
5
2
0
.
8
7
5
1
0
.
8
1
4
2
0
.
9
6
8
3
0
.
7
1
9
0
0
.
8
4
6
1
0
.
7
7
4
4
A
R
R
O
C
0
.
7
9
8
7
0
.
6
2
9
5
0
.
7
5
9
9
0
.
2
6
0
1
0
.
8
1
4
7
0
.
4
7
6
4
0
.
6
1
8
3
0
.
2
3
5
3
N
o
t
e
s
:
T
h
e
s
a
m
p
l
e
w
a
s
c
o
m
p
o
s
e
d
o
f
1
7
b
a
n
k
r
u
p
t
r
m
s
.
T
h
e
d
e
p
e
n
d
e
n
t
v
a
r
i
a
b
l
e
t
a
k
e
s
t
h
e
v
a
l
u
e
1
f
o
r
t
h
e
1
7
b
a
n
k
r
u
p
t
r
m
s
a
n
d
0
f
o
r
t
h
e
n
o
n
-
b
a
n
k
r
u
p
t
r
m
s
.
S
t
a
n
d
a
r
d
e
r
r
o
r
s
a
r
e
s
h
o
w
n
i
n
p
a
r
e
n
t
h
e
s
e
s
.
*
S
i
g
n
i
c
a
n
t
a
t
t
h
e
5
p
e
r
c
e
n
t
l
e
v
e
l
Table IV.
Estimates of the logit
target prediction model
based on the Garc a et al.
(1997) variables
The bias of
unhealthy SMEs
73
2
0
0
1
2
0
0
2
2
0
0
3
S
t
r
o
n
g
r
m
s
W
e
a
k
r
m
s
C
o
n
t
r
o
l
r
m
s
T
o
t
a
l
r
m
s
S
t
r
o
n
g
r
m
s
W
e
a
k
r
m
s
C
o
n
t
r
o
l
r
m
s
T
o
t
a
l
r
m
s
S
t
r
o
n
g
r
m
s
W
e
a
k
r
m
s
C
o
n
t
r
o
l
r
m
s
T
o
t
a
l
r
m
s
B
e
a
v
e
r
(
1
9
6
8
)
T
y
p
e
I
e
r
r
o
r
2
9
.
7
2
7
0
.
6
5
3
2
.
6
3
2
9
.
1
1
6
7
.
1
6
3
3
.
4
6
0
.
0
0
3
3
.
3
5
6
5
.
3
9
3
7
.
6
3
0
.
4
8
T
y
p
e
I
I
e
r
r
o
r
1
1
.
7
6
1
7
.
6
5
5
.
8
8
2
3
.
5
3
3
5
.
2
9
2
3
.
5
3
9
4
.
1
2
1
4
.
2
9
1
4
.
2
9
1
4
.
2
9
8
5
.
7
1
A
R
C
A
P
0
.
8
3
9
8
0
.
5
3
1
6
0
.
7
4
8
4
0
.
8
0
1
0
.
4
7
3
0
.
7
1
3
0
.
7
8
6
0
.
8
3
4
0
.
5
1
2
0
.
7
7
1
0
.
7
8
6
A
R
R
O
C
0
.
7
6
2
5
0
.
7
0
0
6
0
.
7
5
8
6
0
.
7
3
3
0
.
5
9
8
0
.
6
7
9
0
.
2
1
2
0
.
8
1
6
0
.
6
5
2
0
.
7
7
6
0
.
3
2
1
A
l
t
m
a
n
(
1
9
6
8
)
T
y
p
e
I
e
r
r
o
r
2
4
.
3
8
2
6
.
3
9
2
8
.
0
8
2
2
.
9
1
2
4
.
1
4
2
7
.
4
5
0
.
0
0
2
3
.
4
8
2
3
.
4
8
2
8
.
9
4
0
.
4
8
T
y
p
e
I
I
e
r
r
o
r
1
1
.
7
6
2
3
.
5
3
5
.
8
8
2
3
.
5
3
2
9
.
4
1
1
7
.
6
5
9
4
.
1
2
1
4
.
2
9
2
8
.
5
7
1
4
.
2
9
8
5
.
7
1
A
R
C
A
P
0
.
8
9
3
7
0
.
7
8
4
3
0
.
8
3
0
5
0
.
8
7
3
0
.
7
3
8
0
.
8
1
3
0
.
7
4
7
0
.
8
7
3
0
.
7
7
5
0
.
8
3
8
0
.
7
5
9
A
R
R
O
C
0
.
7
4
2
3
0
.
6
0
0
7
0
.
7
1
7
1
0
.
7
2
1
0
.
5
7
1
0
.
6
4
6
0
.
2
2
1
0
.
7
3
7
0
.
5
7
8
0
.
7
1
0
0
.
3
0
0
A
l
t
m
a
n
e
t
a
l
.
(
1
9
7
7
)
T
y
p
e
I
e
r
r
o
r
2
6
.
1
6
4
4
.
3
5
2
7
.
5
3
2
6
.
3
1
4
1
.
6
5
2
7
.
5
4
0
.
0
0
2
2
.
7
4
3
6
.
5
3
2
8
.
3
3
0
.
8
3
T
y
p
e
I
I
e
r
r
o
r
1
1
.
7
6
1
7
.
6
5
1
1
.
7
6
3
5
.
2
9
4
1
.
1
8
1
7
.
6
5
8
8
.
2
4
2
8
.
5
7
2
8
.
5
7
1
4
.
2
9
7
1
.
4
3
A
R
C
A
P
0
.
8
8
1
8
0
.
7
5
0
4
0
.
7
7
4
4
0
.
8
0
1
0
.
6
7
4
0
.
7
4
0
0
.
7
4
9
0
.
8
4
4
0
.
7
7
5
0
.
7
8
6
0
.
7
8
1
A
R
R
O
C
0
.
8
5
2
0
0
.
7
1
6
6
0
.
7
1
5
4
0
.
6
6
2
0
.
6
2
9
0
.
6
5
4
0
.
2
6
3
0
.
7
2
5
0
.
7
1
4
0
.
7
3
6
0
.
3
6
1
G
a
r
c
a
e
t
a
l
.
(
1
9
9
7
)
T
y
p
e
I
e
r
r
o
r
2
4
.
4
8
4
7
.
9
0
2
6
.
7
1
2
4
.
9
4
4
6
.
1
0
2
5
.
9
3
0
.
0
0
5
2
5
.
6
7
4
1
.
3
4
2
5
.
0
1
0
.
4
8
T
y
p
e
I
I
e
r
r
o
r
1
1
.
7
6
4
7
.
0
6
1
1
.
7
6
3
5
.
2
9
2
9
.
4
1
1
7
.
6
5
9
4
.
1
2
1
4
.
2
9
4
2
.
8
6
1
4
.
2
9
8
5
.
7
1
A
R
C
A
P
0
.
8
7
1
1
0
.
7
0
0
7
0
.
8
0
5
8
0
.
8
0
9
0
.
7
4
2
0
.
7
4
3
0
.
7
8
5
0
.
8
4
6
0
.
6
9
8
0
.
7
6
5
0
.
7
8
3
A
R
R
O
C
0
.
7
9
8
7
0
.
6
2
9
5
0
.
7
5
9
9
0
.
6
8
1
0
.
6
4
4
0
.
6
8
0
0
.
2
2
4
0
.
7
9
9
0
.
5
7
7
0
.
7
0
6
0
.
3
1
3
N
o
t
e
:
T
h
e
o
u
t
-
o
f
-
s
a
m
p
l
e
p
r
e
d
i
c
t
i
v
e
a
b
i
l
i
t
y
o
f
t
h
e
m
o
d
e
l
s
i
s
m
e
a
s
u
r
e
d
o
n
t
h
e
w
h
o
l
e
p
o
p
u
l
a
t
i
o
n
o
f
r
m
s
i
n
2
0
0
1
,
2
0
0
2
a
n
d
2
0
0
3
Table V.
Out-of-sample accuracy
of the models estimated
in 2001
JSBED
17,1
74
2
0
0
2
2
0
0
3
2
0
0
4
S
t
r
o
n
g
r
m
s
W
e
a
k
r
m
s
C
o
n
t
r
o
l
r
m
s
T
o
t
a
l
r
m
s
S
t
r
o
n
g
r
m
s
W
e
a
k
r
m
s
C
o
n
t
r
o
l
r
m
s
T
o
t
a
l
r
m
s
S
t
r
o
n
g
r
m
s
W
e
a
k
r
m
s
C
o
n
t
r
o
l
r
m
s
T
o
t
a
l
r
m
s
B
e
a
v
e
r
(
1
9
6
8
)
T
y
p
e
I
e
r
r
o
r
2
8
.
9
6
2
4
.
8
9
6
1
.
2
4
2
8
.
8
5
2
3
.
8
3
6
3
.
3
8
1
.
1
4
2
8
.
1
8
2
3
.
2
7
6
3
.
7
4
1
.
3
1
T
y
p
e
I
I
e
r
r
o
r
1
1
.
7
6
2
9
.
4
1
2
3
.
5
3
1
1
.
7
6
2
3
.
5
3
1
1
.
7
6
2
9
.
4
1
1
4
.
2
9
5
7
.
1
4
2
8
.
5
7
4
2
.
8
6
A
R
C
A
P
0
.
8
8
0
0
.
4
0
6
0
.
8
3
5
0
.
8
7
9
0
.
4
2
9
0
.
7
3
3
0
.
7
9
6
0
.
8
5
6
0
.
4
7
3
0
.
6
2
2
0
.
7
5
4
A
R
R
O
C
0
.
7
9
0
0
.
5
7
3
0
.
6
1
6
0
.
7
9
0
0
.
5
1
5
0
.
6
3
0
0
.
4
0
8
0
.
8
3
1
0
.
4
4
7
0
.
6
8
1
0
.
5
7
7
A
l
t
m
a
n
(
1
9
6
8
)
T
y
p
e
I
e
r
r
o
r
2
8
.
1
6
3
1
.
5
2
2
0
.
7
8
2
8
.
1
1
3
0
.
9
9
2
3
.
0
9
0
.
7
1
2
6
.
5
3
3
0
.
2
6
2
2
.
9
3
0
.
6
5
T
y
p
e
I
I
e
r
r
o
r
5
.
8
8
1
7
.
6
5
2
9
.
4
1
5
.
8
8
5
.
8
8
1
1
.
7
6
3
5
.
2
9
2
8
.
5
7
2
8
.
5
7
1
4
.
2
9
4
2
.
8
6
A
R
C
A
P
0
.
8
7
3
0
.
7
4
0
0
.
7
7
6
0
.
8
4
9
0
.
7
4
4
0
.
8
1
3
0
.
7
7
4
0
.
7
8
5
0
.
7
1
4
0
.
7
5
9
0
.
7
6
3
A
R
R
O
C
0
.
8
0
2
0
.
4
9
5
0
.
6
7
2
0
.
7
7
5
0
.
5
4
0
0
.
7
4
5
0
.
3
3
3
0
.
7
1
2
0
.
6
6
9
0
.
6
8
0
0
.
5
2
8
A
l
t
m
a
n
e
t
a
l
.
(
1
9
7
7
)
T
y
p
e
I
e
r
r
o
r
2
8
.
5
8
4
0
.
5
1
2
6
.
9
3
2
3
.
8
3
3
9
.
0
2
2
1
.
3
0
0
.
9
5
2
3
.
3
2
3
4
.
0
4
2
0
.
6
7
0
.
8
7
T
y
p
e
I
I
e
r
r
o
r
1
1
.
7
6
1
7
.
6
5
2
3
.
5
3
5
.
8
8
1
1
.
7
6
1
1
.
7
6
2
9
.
4
1
2
8
.
5
7
0
.
0
0
1
4
.
2
9
2
8
.
5
7
A
R
C
A
P
0
.
8
7
7
0
.
7
1
6
0
.
7
7
9
0
.
8
5
6
0
.
7
5
8
0
.
7
8
7
0
.
7
7
3
0
.
8
1
5
0
.
8
1
4
0
.
8
1
2
0
.
7
6
3
A
R
R
O
C
0
.
8
1
5
0
.
5
7
0
0
.
7
7
5
0
.
7
7
4
0
.
6
4
5
0
.
7
2
1
0
.
3
6
7
0
.
7
3
6
0
.
7
2
6
0
.
7
5
6
0
.
6
0
6
G
a
r
c
a
e
t
a
l
.
(
1
9
9
7
)
T
y
p
e
I
e
r
r
o
r
3
0
.
7
6
3
8
.
8
1
1
6
.
3
3
2
7
.
7
2
3
7
.
8
4
1
7
.
5
0
0
.
5
2
2
5
.
7
1
3
6
.
0
4
1
7
.
7
6
0
.
4
8
T
y
p
e
I
I
e
r
r
o
r
5
.
8
8
2
9
.
4
1
3
5
.
2
9
1
1
.
7
6
5
.
8
8
1
1
.
7
6
3
5
.
2
9
1
4
.
2
9
4
2
.
8
6
1
4
.
2
9
5
7
.
1
4
A
R
C
A
P
0
.
8
7
2
0
.
7
4
3
0
.
7
4
6
0
.
8
5
3
0
.
7
5
4
0
.
8
0
6
0
.
7
8
8
0
.
8
5
0
0
.
6
0
5
0
.
7
5
8
0
.
7
4
5
A
R
R
O
C
0
.
8
1
4
0
.
4
7
6
0
.
6
1
8
0
.
7
5
4
0
.
5
2
4
0
.
6
7
8
0
.
3
1
4
0
.
8
2
7
0
.
4
8
7
0
.
6
1
5
0
.
3
9
8
N
o
t
e
:
T
h
e
o
u
t
-
o
f
-
s
a
m
p
l
e
p
r
e
d
i
c
t
i
v
e
a
b
i
l
i
t
y
o
f
t
h
e
m
o
d
e
l
s
i
s
m
e
a
s
u
r
e
d
o
n
t
h
e
w
h
o
l
e
p
o
p
u
l
a
t
i
o
n
o
f
r
m
s
i
n
2
0
0
1
,
2
0
0
2
a
n
d
2
0
0
3
Table VI.
Out-of-sample accuracy
of the models estimated
in 2002
The bias of
unhealthy SMEs
75
Conclusions
Many efforts have been devoted to developing models to predict bankruptcy
probability and banks, investors and rms are interested in using these models.
However, since the models have been developed in the aim to estimate the bankruptcy
probability of quoted rms, less attention has been paid to non-quoted SMEs. Usually,
the bankruptcy probability of non-quoted SMEs is studied by analysing nancial
ratios using discriminating techniques. These models are based on constructing an
estimation sample using bankrupt and non-bankrupt rms, controlling by size and
industry or unbalancing the sample using all rms.
In our article, we propose an alternative to control the heterogeneity of healthy
rms. In doing this, we construct a nancial health indicator to dene the degree of
nancial health. Our results are obtained under four different models of credit scoring.
We show in-sample and out-of-sample bankruptcy predictions. We apply measures
proposed by the Basel Committee on Banking Supervision to evaluate the credit
scoring of the models depending on the criterion considered to select the rms.
Focusing on the non-bankrupt rms, the nancial health indicator allows rms to
be classied before the estimation process. This procedure permits the heterogeneity of
the rms to be reduced as well as identifying a strong rm sample to estimate the
bankruptcy probability accurately. The in-sample and out-of-sample evaluation based
on the CAP and ROC indicators, proposed by the BCBS, leads to the conclusion that the
models estimated under the strong rm sample are much more accurate.
References
Altman, E.I. (1968), Financial ratios, discriminant analysis and the prediction of corporate
bankruptcy, Journal of Finance, Vol. 23 No. 4, pp. 589-609.
Altman, E.I. and Saunders, A. (1998), Credit risk measurement: developments over the last
20 years, Journal of Banking and Finance, Vol. 21, pp. 1721-42.
Altman, E.I., Haldeman, E. and Narayanan, P. (1977), Z analysis: a new model to identify
bankruptcy risk of corporations, Journal of Banking and Finance, Vol. 10, pp. 29-54.
Basel Committee on Banking Supervision (2005), Studies on the validation of internal rating
systems, Working Paper No. 14, Basel Committee on Banking Supervision, Basel.
Beaver, W.H. (1966), Financial ratios as predictor of failure, Empirical Research in Accounting:
Selected Studies, Supplement to Journal of Accounting Research, Vol. 5, pp. 71-111.
Beaver, W.H. (1968), Alternative accounting measures as predictor of failure, The Accounting
Review, Autumn, pp. 112-22.
Berryman, J. (1983), Small business failure and bankruptcy: a survey of the literature,
European Small Business Journal, Vol. 1 No. 4, pp. 47-59.
Carter, R. and Van Auken, H. (2006), Small rm bankruptcy, Journal of Small Business
Management, Vol. 44 No. 4, pp. 493-512.
Cochran, B. (1986), Small business mortality rates: a review of the literature, Journal of Small
Business Management, Vol. 19 No. 4, pp. 50-9.
Craig, B., Jackson, W. and Thomson, J. (2003), On SBA-guaranteed lending and economic
growth, working paper, Federal Reserve Bank of Cleveland, Cleveland, OH, April.
Edmister, R. (1972), An empirical test of nancial ratio analysis for small business failure
prediction, Journal of Financial and Quantitative Analysis, Vol. 7 No. 2, pp. 1477-93.
JSBED
17,1
76
Engelmann, B., Hayden, E. and Tasche, D. (2003), Testing rating accuracy, Risk, January,
pp. 82-6.
Fredland, J. and Morris, E. (1976), A cross section analysis of small business failure, American
Journal of Small Business, Vol. 1, July, pp. 7-18.
Garc a, D., Arques, A. and Calvo-Flores, A. (1997), Un modelo discriminante para evaluar el
riesgo bancario en los creditos a empresas, Revista Espan ola de Financiacion y
Contabilidad, Vol. 24 No. 82, pp. 175-200.
Hall, G. and Young, B. (1991), Factors associated with insolvency amongst small rms,
International Small Business Journal, Vol. 9 No. 2, pp. 54-63.
Headd, B. (2003), Redening business success: distinguishing between closure and failure,
Small Business Economics, Vol. 21, pp. 51-61.
Thedodossious, P., Kahya, E., Saidi, R. and Philippatos, G. (1996), Financial distress and
corporate acquisitions: further empirical evidence, Journal of Business Finance and
Accounting, Vol. 23, pp. 699-719.
Watson, J. and Everett, J. (1996), Do small business have high failure rates?, Journal of Small
Business Management, Vol. 34 No. 4, pp. 45-52.
Watson, J. and Everett, J. (1999), Small business failure rates: choice of denition and industry
effects, International Small Business Journal, Vol. 17 No. 2, pp. 31-47.
Zingales, L. (2000), In search of new foundations, Journal of Finance, Vol. 55 No. 4, pp. 1623-53.
About the authors
J. Samuel Baixauli is an Associate Professor at the Department of Management and Finance at
the University of Murcia, Spain. He received his PhD in Business Administration for work on
nancial economics. He has been research visitor at the Department of Economics at the
University of York, UK. His main research areas include nancial modeling and estimation of
market and credit risk. He has published on these topics in international journals such as
European Journal of Operational Research, Journal of Financial Research, Review of Quantitative
Finance and Accounting and European Journal of Finance. J. Samuel Baixauli is the
corresponding author and can be contacted at: sbaixaul@um.es
Antonina Modica-Milo is an Associate Professor at the Department of Management and
Finance at the University of Oriente, Venezuela. She received her PhD in Business Economics
from the Department of Management and Finance at the University of Murcia, Spain. Her main
research areas include nancial modeling and estimation of credit risk. She has experience
working as a nancial advisor for relevant private companies.
The bias of
unhealthy SMEs
77
To purchase reprints of this article please e-mail: reprints@emeraldinsight.com
Or visit our web site for further details: www.emeraldinsight.com/reprints