See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/305033149

The portability of altman’s Z-score model to predicting

corporate financial distress of Slovak companies

Article  in  Technological and Economic Development of Economy · June 2016

DOI: 10.3846/20294913.2016.1197165

CITATIONS READS

18 1,403

2 authors:

Martin Boda Vladimir Uradnicek

Matej Bel University in Banská Bystrica / Faculty of Economics Univerzita Mateja Bela v Banskej Bystrici
45 PUBLICATIONS   120 CITATIONS    11 PUBLICATIONS   33 CITATIONS   

SEE PROFILE SEE PROFILE

Some of the authors of this publication are also working on these related projects:

The finance-growth nexus in the Visegrad Group countries View project

SMART model - decision-making support tool in corporate management View project

All content following this page was uploaded by Martin Boda on 08 November 2018.

The user has requested enhancement of the downloaded file.

 TECHNOLOGICAL AND ECONOMIC DEVELOPMENT OF ECONOMY
ISSN 2029-4913 / eISSN 2029-4921

2016 Volume 22(4): 532–553

doi:10.3846/20294913.2016.1197165

THE PORTABILITY OF ALTMAN’S Z-SCORE MODEL TO PREDICTING

CORPORATE FINANCIAL DISTRESS OF SLOVAK COMPANIES

Martin BOĎA, Vladimír ÚRADNÍČEK

Quantitative Methods and Information Systems Department, Faculty of Economics,
Matej Bel University in Banská Bystrica, Tajovského 10, 975 90 Banská Bystrica, Slovakia

Received 11 April 2016; accepted 31 May 2016

Abstract. he paper challenges the widespread use of Altman’s bankruptcy formula known as the
“Z-score model” in Slovak corporate practice and comes with the goal to verify its usability in the
Slovak economic environment. To this end, a deinition of inancial distress is adopted that sum-
marizes weaknesses of Slovak enterprises stemming particularly from liquidity drain and operating
losses. he veriication juxtaposes three variants of the Z-score model and assesses their prediction
ability using a data set of Slovak enterprises for the period from 2009 until 2013. Both the original
1968 Z-score model and the revised 1983 Z-score devised for the US economic environment are
compared with the Z-score model re-estimated to the Slovak data copying the methodological pro-
cedure of Altman. he results indicate that Altman’s bankruptcy formula is portable into the Slovak
economic conditions and useful for predicting inancial diiculties in view of the adopted deini-
tion of inancial distress. Altman’s original and (especially the) revised formulation of the Z-score
model are preferable if overall classiication accuracy is the main interest. Finally, it is advisable to
re-estimate the coeicients of the Z-score model if inancially distressed enterprises are the focus
and the goal is to classify distressed enterprises as best as possible.

Keywords: Altman’s bankruptcy formula, Slovak enterprises, inancial distress, classiication ac-
curacy, true positive rate.

JEL Classiication: G33, M40.

Introduction

his paper’s thematic orientation revolves around Altman’s bankruptcy formula. his no-
torious bankruptcy prediction model popularized as the “Z-score model” has found ap-
plication in various ields of inance where knowledge of an enterprise’s future inancial
condition is perceived vital. Although devised almost half a century ago, Altman’s Z-score
model continues to be encountered and used routinely in a multitude of business situa-
tions, which encompass corporate inancial analysis made by enterprises for the purpose

Corresponding author Martin Boďa

E-mail: martin.boda@umb.sk

Copyright © 2016 Vilnius Gediminas Technical University (VGTU) Press

http://www.tandfonline.com/TTED
Technological and Economic Development of Economy, 2016, 22(4): 532–553 533

of managerial decision-making. It is also made by investors for their investment decision,

credit assessment of corporate customers carried out by commercial banks, or the on-going
concern assessment of corporate clients undertaken by auditors. A peculiar trait of the
model is that it was developed for the dated environmental conditions of the US economy,
and yet the model is used on an international basis outside its domestic environment. In
such a way, Altman’s model is used also in Slovak corporate practice and is very popular
to both theorists and practitioners of corporate inance, which is striking and odd as there
are a number of reasons provoking caution or even outright opposition to its use.
he underpinning of Altman’s Z-score model is the belief that it is possible – on the ba-
sis of past information embedded and represented in inancial indicators (such as liquidity,
activity, capital structure, or proitability ratios) – to predict an enterprise’s future inancial
distress and in such an implicit way to measure its inancial health. Committed to this task,
Altman, in 1968, made use of 5 inancial indicators computed for a limited non-represen-
tative sample of US enterprises so as to develop a model in the framework of a multi dis-
criminant analysis (see the original paper: Altman 1968). Despite the methodological and
technical limits in the 1960’s and the limitations of the data samples available, the model
exhibited high accuracy rates for both the estimation and the hold-out sample and attained
a high level of prediction to the actual conditions of the US corporate practice of the 1970’s.
Altman is credited with being the irst to set up a rigorous and formal bankruptcy predic-
tion model in a multivariate framework (see e.g. Dimitras et al. 1996: 498; Delina, Packová
2013: 101). It is perhaps the main reason that the model proliferated and continues to be
a standard bankruptcy prediction model for US manufacturing companies and others as
well. As we consider the present period, its usability had been extended beyond the US
non-manufacturing companies to include the entire US economic environment and then
became internationalized. In this respect, a version of Altman’s Z-score model developed
again by Altman (1983) and known as the “revised Z-score model” is ot referred to in this
paper and contrasted with the “original” Z-score model.
Altman’s Z-score model has been applied to Slovak corporate practice without any re-
gard to the serious methodological limitations of its sensible use. Some examples are listed
here in this paper. In addition, the aspects of its use in application and misuse are not clear
as the formula was elaborated for the purpose of bankruptcy prediction, whereas it is usu-
ally interpreted in reference to inancial distress. hat being so, this problem appears to be
rather fortunate as there are views, some theoretical and others empirical, that this model
should be used for predicting inancial distress rather than bankruptcy itself (e.g. Gilbert
et al. 1990: 169; Grice, Ingram 2001; Grice, Dugan 2001).
Noting that bankruptcy models are better suited to the task of predicting inancial dis-
tress and building on this premise still leaves uncertainty as to whether Altman’s Z-score
model should be judiciously applied to Slovak economic conditions and as such, its ap-
propriateness must be veriied. his paper does not attempt to develop a new bankruptcy
model for predicting inancial distress in the Slovak economic environment; in this respect,
the initiative is let to others (see e.g. Chrastinová 1998; Zalai 2000; Gurčík 2002). With
regards to all the limitations that are associated with the current use of this model, the
goal of the paper is to verify whether it is useful for predicting inancial distress of Slovak
534 M. Boďa, V. Úradníček. he portability of Altman’s Z-score model to predicting corporate ...

enterprises. Using a data set of Slovak enterprises for the 2009–2013 period, the veriica-
tion proceeds on two tracks and is based on three alternative Z-score model representation.
One representation addressing performance predictions is the model with an original coef-
icients as was estimated from US data by Altman himself (Altman 1968), whilst another
representation is the revised model with updated coeicients estimated from the same data
set also made by Altman (1983). Eventually, the third representation is the model with
the coeicients estimated from Slovak data following Altman’s methodological procedure.
As is the case with the majority of studies of this sort (cf. e.g. Balcaen, Ooghe 2006: 75)
concerning Altman’s Z-score estimation, the sample of enterprise data for modelling does
not originate as a random draw, which is the reason that statistical testing is avoided and
not attempted here.
he results indicate that the Z-score models based on Altman’s methodology do have
some merit in predicting distress of Slovak enterprises. he Z-score models devised by Alt-
man (1968, 1983) for US manufacturing enterprises should be used in the Slovak economic
conditions only when overall accuracy is of central interest (i.e. when it is important to
classify enterprises correctly regardless of whether they are actually distressed or not). It
is particularly the revised Z-score model that may be recommended for use in this clas-
siication context. However, if the objective is to mainly identify distressed enterprises (and
misclassiications of non-distressed enterprises are not a nuisance), then it is appears as
advisable to re-estimate the Z-score model and to found the classiication procedure on
the new estimated classiication rule. hese indings conirm the fact that Altman’s Z-score
model is portable and robust for the Slovak economic environment.
Before going into updating the formula of the Z-score model with the use of Slovak data
and conducting the intended veriication, two stopovers are necessary. Some space must
be reserved for presenting the methodological elements of the model, and some attention
must be focused on discussing applicability of the model in the Slovak conditions and its
interpretation. he body of the paper is thus organized into three other sections: a section
expositing the model and particularizing the historical circumstances under which it was
developed; a section cautioning against its mechanical use in the Slovak economic condi-
tions whilst suggesting an appropriate interpretation that should be given to the model;
and a section presenting the results of the intended veriication. Finally, the last part of the
paper contains the concluding remarks an discussion.

1. Altman’s Z-score model and its alterations

he irst rigorous efort in the area of corporate bankruptcy prediction was attributed
to Altman (1968) who developed a ive-factor model using linear discriminant analysis
(LDA). His model continues to be used nowadays with non-diminishing popularity not
only it the US environment, but also in Slovak corporate practice. he model was origi-
nally developed for manufacturing enterprises with publicly traded shares, but was later
improved into a model for manufacturing enterprises with shares non-listed on the capital
market (see Altman 1983) and into other models including non-manufacturing companies
as well (see Altman et al. 1977; Altman 2013). Nonetheless, the original model of 1968,
Technological and Economic Development of Economy, 2016, 22(4): 532–553 535

known as the Z-score model or called Altman’s bankruptcy formula, still appears to be the
model of choice worldwide (see e.g. Altman et al. 2014: 2).
he original Z-score model assumes that the analysed enterprise is listed on a capital
market and integrates in a linear fashion 5 inancial indicators. heir list comprises: X1 –
working capital/total assets, X2 – retained earnings/total assets, X3 – earnings before inter-
est and taxes/total assets, X4 – market value of equity/book value of debt, and X5 – sales/
total assets. All of them are constructed as ratios of items reported in inancial statements
or computed from available market data. Usage of the market value of equity in X4 neces-
sitates that the entity is listed on the capital market (as it is usually equalled to the market
capitalization of the entity’s shares) or that this sort of information is available that goes
beyond the traditional informational content of inancial statements. All the other variables
appearing in the deinition of X1 – X5 are particular items declared in inancial statements.
At any rate, the values of these ive indicators are substituted into the linear function,

Z = 1.2X1 + 1.4X2 + 3.3X3 + 0.6X4 + 0.999X5, (1)

which gives a value of the so-called Z-score. High values of the Z-score point to good
inancial health and low values suggest an increased probability of bankruptcy. To distin-
guish between inancially healthy companies and companies with increased probability
of bankruptcy, two cut-of values were deined by Altman in the process of the model’s
veriication: if the Z-score is below 1.81, the analyzed entity is at risk of bankruptcy, but if
the Z-score exceeds 2.99, the analysed entity should be viewed as inancially healthy. he
area between 1.81 and 2.99 is interpreted as the grey area (or the “zone of ignorance”) with
diiculty to predict the inancial status.
It is instructive to study briely the methodological procedure that was utilized to de-
velop the Z-score model and estimate the coeicients of the function (1). he ive indica-
tors represented in (1) were selected from a list of 22 inancial ratios describing an entity’s
inancial status in terms of liquidity, solvency, activity, proitability and leverage. Whilst
modelling, Altman used LDA, and in obtaining the coeicients of (1), he used the paired
sample of 33 bankrupt and 33 non-bankrupt manufacturing companies with shares listed
on the capital market. Crucial in this respect, is not only the disturbingly small sample
size, but also the way this sample was constructed. Starting with the sub-sample of 33
bankrupt U.S. companies listed on the capital market, which between 1946 and 1965 iled
a petition for bankruptcy liquidation according to Chapter X of the National Bankruptcy
Act (addressed as the Chandler Act and efective in the U.S.A. between 1938 and 1978).
hese companies were all medium-sized manufacturing and were each paired with a com-
parable U.S. medium-sized manufacturing company listed on the capital market that had
not experienced bankruptcy diiculties in the investigated period of 20 years. he issue of
random sampling was not performed by Altman at all (understandably due to the limited
availability of data).
Later the model was modiied, re-estimated and published by Altman in his monograph
(Altman 1983: 120–124) in order to account for manufacturing entities non-listed on the
capital market. In the process, X4 was changed into X4’ deined as book value of equity/
book value of debt and under this deinition its calculation does not require market data.
536 M. Boďa, V. Úradníček. he portability of Altman’s Z-score model to predicting corporate ...

Using the same data set as in the 1968 paper, the re-estimated functional then took the
form:
Z¢ = 0.717X1 + 0.847X2 + 3.107X3 + 0.420X4’ + 0.998X5, (2)

in which Z¢ denotes the changed Z-score. he boundary points were subsequently changed
to 1.23 (from 1.81) and 2.9 (from 2.99), otherwise, the classiication rule remained intact.
his model is known as the revised Z-score model, which is the name assigned by Alt-
man himself (Altman 1983: 202). In the same monograph Altman (1983: 124) eventually
extended and adapted this model for non-manufacturing entities whose shares need not
be listed on the capital market. he model was named as the Z″-score model. here have
been a number of competing models proposed and estimated for the U.S. environment as
well as for diferent economic conditions using diferent methodologies (see e.g. Balcaen,
Ooghe 2006; Wu et al. 2010), in addition, there has been another model that was proposed
by Altman et al. (1977) along the established tradition that used 7 predictors and was for-
mulated for entities in manufacturing and retail. Implemented in the framework of LDA,
this model is proprietary and designated for the purpose of credit risk assessment. It came
with a trademark and was launched as the ZETA® model. Even so this second-generation
model has never received the level of immense popularity as the Z-score model in its
original variant.
Since then, a number of other models have been developed using sophisticated meth-
odology surpassing the limitations of discriminant analysis and specialized to a particular
country’s economic conditions, it seems that the original model still prevails in predicting
bankruptcy not only in the world (see Grice, Ingram 2001; Balcaen, Ooghe 2006; Atlman
et al. 2014) but also for Slovakia. he following section gives some examples of overconi-
dence in the use of the original Z-score model in the Slovak conditions and questions the
very possibility of such usage. It rather suggests an adequate interpretation of its indicative
capability.

2. Usage of the Z-score model in the Slovak conditions:

limitations and interpretation

It seems that Altman’s Z-score model given by (1) and (2) is utilized frequently in the
Slovak corporate practice notwithstanding the apparent methodological troubles. A num-
ber of methodological issues arise emanating from the fact that the model is even used
under Slovak economic conditions (being diferent from those for which it was formulated)
and the fact that its usage is extended to the contemporary period (irrespective of a histori-
cal gap of 50 to 70 years). Another speciic – enforced by the intentional specialization of
the original Z-score model in listed enterprises – is that it is implemented in mixed form
combining both formulas (1) and (2). Strictly speaking, the original 1968 discriminant
model in (1) is not workable without knowledge of non-market information (necessary
to compute the value of X4). Yet, practitioners prefer to use this original model, in which
they mechanically and nonchalantly substitute X4¢ for X4 keeping the coeicient values and
preserving the original boundaries of the classiication zones. In other words, the market
Technological and Economic Development of Economy, 2016, 22(4): 532–553 537

value of equity is simply replaced by the book value and the formula in (1) is otherwise
used in its fullness.
One more problem comes from the obvious incompatibility in the reporting standards
carried over in inancial statements that are the basis for computing the values X1 – X5.
his methodological note relates not only to the very diference between the US and Slovak
accounting systems but also to the temporal discrepancies resulting from past times to the
present. On the one hand, in the Slovak Republic, enterprises are required to report in
compliance with the national inancial accounting standards or in compliance with IFRS;
on the other hand, the data used by Altman (1968, 1983) were calculated from inancial
statements reported in conformity with US GAAP. Altman’s data pertained to the period
from 1946 until 1966 and fully absorbed methodological updates and modiications in the
US GAAP standards made efective occasionally from year to year. In efect, the methodi-
cal content of the inancial ratios used by Altman (1968, 1983) and the information they
convey difers from the content and informational value of the inancial ratios of present,
computable from Slovak inancial statements. herefore, a practitioner applying the original
Z-score model in (1) or the revised Z-score model in (2) in an otherwise “correct” manner,
makes at best an approximation and an educated guess to bankruptcy prediction.
All these points invalidate usage of Altman’s Z-score model in the Slovak conditions
although some of the deinitional uncertainties may be to some extent mended. Despite
these critical reservations, it is still possible that the model is portable and may be an aid
in predicting inancial health of enterprises, not only in the present period, but also in the
Slovak economic conditions. Such a possibility follows from the fact that the selected ive
indicators, X1 to X5 (with X4 changed to X4’), possess a universally good and a priori
of economic rationale in fulilling the task of distinguishing between enterprises that are
inancially sound and healthy and those that are inancially distressed. A justiicatory ac-
count for their use in a discriminatory context is provided in depth by Altman (1968:
594–596, 1983: 106–107). he generalization of Altman’s Z-score model across diferent
time periods has been extensively studied, albeit its application has not gone beyond the US
economic environment (cf. e.g. Begley et al. 1996; Grice, Ingram 2001; Wu et al. 2010; Li
2012). All the same, some researchers were interested in a possibility of applying Altman’s
Z-score model in the Czech economic environment (Pitrová 2011; Kopta 2009; Kalouda,
Vaníček 2013) or even in the Slovak economic conditions (Zalai 2000; Gurčík 2002; Delina,
Packová 2013). Notwithstanding the factors opposing this model or the research indings,
corporate theorists and practitioners are oblivious and do not seem discouraged from us-
ing it for Slovak enterprises. Some selective examples are provided that Altman’s model is
applied on a massive scale in real practice or amongst corporate professionals.
– Altman’s Z-score model is promoted within the professional community by virtue
of popularizing articles advising its use in the Slovak conditions. Examples include
Stolárik (1996), Zalai (1997: 21–22) or Štrangfeldová (2012: 14–16). Altman’s Z-score
model is further implemented (with some other bankruptcy prediction models) in
the functionalities of the Soina Microsot® Excel® spreadsheet application ofered on
a commercial basis as a complex sotware tool for inancial analysis (see Király et al.
2015) and its lite version complements a book on corporate inancial analysis by
Kotulič et al. (2010).
538 M. Boďa, V. Úradníček. he portability of Altman’s Z-score model to predicting corporate ...

– here are many instances in which Altman’s formula is used for the purpose of real
corporate analysis and is reported in inancial statements and annual reports, partly
in assessing the ability of an enterprise to continue as a going concern. his utilization
of the Z-score model can be found in many annual reports of manufacturing irms
(e.g. Púchovský mäsový priemysel 2014) or service providers (e.g. Dopravný podnik
mesta Prešov 2015; I.S.D.D. plus 2014; Kúpele Trenčianske Teplice 2012). Regrettably,
a number of examples of the sort can be found.
– An illustrious application of the revised Z-score model is encountered in the analyti-
cal materials of the Slovak Business Agency, which is a specialized non-proit organi-
zation founded by the Ministry of Economy of the Slovak Republic providing support
to small and medium-sized enterprises (SMEs). In the analyses of inancial perfor-
mance of SMEs issued by the Slovak Business Agency (2014: 45–51; 2015: 44–51),
Altman’s Z-score model is exploited to consider the inancial health of the segments
of Slovak SMEs and its trends are scrupulously assessed. It must be admitted, though,
that the Z-score model is used in these analyses in the correct format intended for
private enterprises (with the revised coeicients and boundary values of the grey
zone), which deies the usual practice of using the model in mixed form.
he listed applications of Altman’s Z-score model are made in the belief that they are
methodologically correct and that it is perfectly proper to predict bankruptcy in this way,
which naturally is not correct. It must be admitted though, that incentives for such an over-
use of Altman’s Z-score model come sometimes from respectable sources. For instance, in
their book on corporate inance, Hillier et al. (2010: 839–840) use the model for a publicly
traded European enterprise, upon which they voice a well-intentioned caution that one
should not take the outcome as “particularly precise”. Interestingly, it is the very Altman
(1983: 206) who discusses the Z-score model and concedes in his concluding notes that
the model should be applied with care in practical situations.
In exacerbation to incorrect usage of Altman’s Z-score model, there is an issue that it
is not completely clear what it really predicts and in what respect it is helpful. In part, this
relates to the diference that exists between inancial distress and bankruptcy as two difer-
ent notions of enterprise economic condition, although it is also connected to the empirical
behaviour of the model. Recognizing that bankruptcy is a term designating a legal status for
an enterprise while inancial distress is a term employed to capture its worsened economic
position, there is an opinion maintained e.g. by Taler (1984), Gilbert et al. (1990) or point-
ed out by Grice, Dugan (2001: 154) that bankruptcy models should be viewed in a broader
context of inancial distress prediction. On this point Gilbert et al. (1990: 169) clarify that
“bankruptcy model scores should be interpreted as descriptions of inancial distress rather
than as predictions of bankruptcy per se”. his comes with the empirical testimony focused
specially on Altman’s Z-score model and delivered by Grice, Ingram (2001) who make a
compelling case that “the model is useful for predicting inancial distress conditions other
than bankruptcy” (ibid.: 53). here are a number of occasions summarized e.g. by Altman
et al. (2014: 6) on which the Z-score model is utilized for measuring inancial distress or
inancial strength. In fact, as follows from a brief overview made by Begley et al. (1996: 283)
the model happens to be chosen frequently for predicting inancial distress (as opposed to
Technological and Economic Development of Economy, 2016, 22(4): 532–553 539

bankruptcy). he aspect of appropriate interpretation of Altman’s Z-score model is rooted

in the separation of the notions of bankruptcy and inancial distress from each other. To be
fair, it must be said that there are empirically established arguments that inancial distress
and bankruptcy are separate processes driven by diferent factors (see e.g. Platt, H. D.,
Platt, M. B. 2006). he immediate consequence is that bankruptcy models are not valid in
predicting inancial distress, and vice versa.
Bankruptcy describes a legal status of failure resulting from insolvency whose declara-
tion is initiated on request and sets of a formal procedure of corporate reorganization (in
an attempt at a recovery program) or liquidation. In contrast, inancial distress is usually
understood in a broader sense as it describes a inancial condition when an enterprise
experiences diiculties in fulilling inancial obligations on schedule to the fullest extent
or sufers from insuicient liquidity. Both bankruptcy and inancial distress are associ-
ated with insolvency, yet bankruptcy is a single-event issue and commences with a formal
declaration of insolvency, inancial distress results from various events and ends up with
declared on undeclared insolvency. Some useful discourse on the diferences and relation-
ship between inancial distress and bankruptcy can be found, e.g. in Gilbert et al. (1990:
161–162, 169–170), Balcaen, Ooghe (2006: 72–73), Sun et al. (2014: 42–43). As such, inan-
cial distress is characterized technically by a number of operational indicators that are used
in capturing negative events that may (and possibly) cause inancial instability eventually
manifested in insolvency. A conviction on suitability of these operational indicators to
detect inancial distress is then mirrored in the deinition that operationalizes the status of
inancial distress. Usually this deinition is adjusted to speciic corporate conditions under
investigation or is limited by data availability.
he sparse Slovak bankruptcy literature focuses upon bankruptcy rather than inancial
distress and therefore uses the judicial deinition of bankruptcy to this end (e.g. Delina,
Packová 2013), or tends to use a specialized deinition.
– In the former case, the legal deinition of inancial distress according to the current
bankruptcy law lays down the economic precondition for bankruptcy that a debtor is
in distress if he be either insolvent (i.e. incapable of paying at least two liabilities that
are 30 days overdue to more than one creditor) or overdebted (i.e. the value of liabili-
ties to more than one creditor is in excess of the value of assets). he legal precondi-
tion is that a petition for bankruptcy proceedings be iled by the debtor himself or by
any of his creditors. (Bankruptcy Act No. 7/2005 Coll., §§ 3, 11 (Zákon č. 7/2005…)).
– he second case is exempliied by the approach of Šnircová (1997: 16) who classiies
an enterprise as inancially distressed in a given year if it fails to report a positive EAT
(earnings ater taxes) and at the same time it shows a level of liquidity measured by
the current ratio lower than 1. he approach of Šnircová (1997) seeks to establish a
relationship between liquidity and inancial distress in a way similar to Hrdý, Šimek
(2012: 126) although these authors centred on the Czech economic environment.
Notwithstanding the territorial and economic scope of Altman’s Z-score model, it is
naturally questionable as to whether the model is outdated and still applies to the new
economic conditions, which has changed since 1968 when the original version of the model
was published. Although centred mostly on the US corporate environment, there is suf-
540 M. Boďa, V. Úradníček. he portability of Altman’s Z-score model to predicting corporate ...

icient evidence that the model delivers a decreased predictive accuracy with more recent
data and that the model’s coeicients seem to be unstable in diferent economic periods
(e.g. Begley et al. 1996; Wu et al. 2010: 35; Grice, Ingram 2001: 53; Balcaen, Ooghe 2006:
70). his observation substantiates the resulting advice to estimate the model’s coeicients
anew in order to relect the ever-present environmental changes.
Referring to the quoted testimony of the empirical research that the Z-score model is
more apt in identifying generally distressed enterprises than bankrupt ones, the described
circumstances against its use for Slovak enterprises are further borne in mind and the paper
proceeds in deiance of the logic of importability across time periods and economic envi-
ronments. Its aim is then to verify whether Altman’s Z-score model is, in the mixed-impure
form, useful to the task of discriminating between inancially distressed and inancially
non-distressed Slovak enterprises. What is not the goal of this paper is the development
of a completely new model, even one possibly based on the ive Altman’s ratio indicators.
he paper thus remains in proximity to the current Slovak corporate practice prejudiced
in favour of Altman’s Z-score model.

3. Veriication of the Z-score model for use in the Slovak conditions

In assessing the capability of the Z-score model to predict inancial distress of Slovak en-
terprises, the veriication procedure stood upon three variants of the Z-score model that
were compared in terms of their predictive performance and helpfulness in identifying
inancially distressed enterprises. he subject of comparison were thus
– the “original” 1968 Z-score model given in (1) with X4¢ forcibly substituted for X4,
– the 1983 revised Z-score model obeying in full the speciication in (2),
– the newly-estimated Z-score model using X1, X2, X3, X4¢ and X5.
Whilst the irst variant corresponds to the mixed impure form in which the model is
proliferating in the Slovak corporate practice, the second variant is perhaps more appropri-
ate and its usage is comparatively more valid if one omits a number of other methodologi-
cal issues. Lastly, the third variant of the model should expectably comply with the actual
environmental conditions in which Slovak enterprises operate and should therefore be
found outperforming the other two models. Its inclusion to the veriication conforms to the
views summarized before in the preceding section that the model should be re-estimated
with a new set of data. All the same, the classiication performance of this triplet of variant
adaptations of the Z-score model is reasonably anticipated to be less satisfactory than that
of the “true” Z-score models as initially ascertained by Altman (1968, 1983) for the US
economic environment or as found later by other researchers for the US environment in
diferent time periods (see e.g. Altman, Hotchkiss 2006: 244).
he depth of the veriication was limited by data availability and the analysis required
several methodological speciications which deserve both explanation and vindication.
Before anything else, a deinition of a inancially distressed enterprise had to be adopted
that would adequately relect inancial diiculties that distressed Slovak enterprises have
to face. A total of three criteria of inancial distress were formulated on the basis of the
Technological and Economic Development of Economy, 2016, 22(4): 532–553 541

deinitions of inancial distress as legalistically established in §§ 3 and 11 of Bankruptcy Act

No. 7/2005 Coll. (Zákon č. 7/2005…) and as formulated by Šnircová (1997: 16). For the
purpose of this paper, an enterprise is considered inancially distressed if (a) its equity is
negative, (b) its EAT is negative, and (c) its current ratio attains a value lower than 1. All
the three conditions must be satisied in order for an enterprise to be considered inan-
cial distressed. As to the irst condition, it is a simpler formulation of over indebtedness
stipulated by the legal deinition of inancial distress (save that for reasons that are merely
practical, it is not required that the liabilities in excess of assets must concern more than
one creditor). he second condition highlights that in a given year an enterprise is loss-
making and combined with the irst condition this criterion is instrumental in pinpoint-
ing enterprises that are in the long-run unproitable. A negative value of equity suggests
that the enterprise concerned has made over the past periods an unsustainable amount of
losses (beyond the level absorbable by its assets) and a negative value of EAT then reveals
that there are no sign of change and the enterprise continues to be on an adverse path of
(accounting) value destruction. he ex post inancial analyses elaborated by the Ministry
of Economy of the Slovak Republic and by the Slovak Business Agency bear testimony to
the fact that the ability to report a proit and maintain it is crucial and existential to many
Slovak enterprises (cf. Ministerstvo hospodárstva Slovenskej republiky 2014: 9–10; Slovak
Business Agency 2014: 11–13, 2015: 9–10). In addition, the third criterion is associated
with undercapitalization, a situation in which long-term assets are covered not only by
long-term liabilities but also by a fraction of current liabilities. Such a situation is in the
long run not sustainable and it also implies an accounting form of insolvency. By deini-
tion, an undercapitalized enterprise is unable to liquidate all its current assets to pay of or
settle all its current liabilities. A too low a level of liquidity is an impediment to honouring
short-term payables and implementing investment strategies. he signiicance of liquidity
for Slovak enterprises is discussed e.g. by Strachotová (2012: 3–4) who assesses the liquidity
of the whole entrepreneurial sector in Slovakia as well.
In respect of this particular choice employed in designating a inancially distressed
company, it need also be noted that the irst two criteria – in a manner of speaking – ex-
press the long-term aspect of inancial diiculties. It is customary to deine or identify
inancial distress on the basis of few technical indicators and many study have looked into
the past iscal periods of two or three years. Usually some variations of proitability ratios
screened for the two or three consecutive periods are therein employed (see e.g. Platt, H. D.,
Platt, M. B. 2006: 144, 155; Asquith et al. 1994: 628). he condition of negative equity and
negative EAT maps both the present situation and simultaneously provides a suicient in-
sight into the past underperformance. Alternatively, negative equity may be as well a result
of the present period’s immense loss, which is then a dependable indicator of sudden and
immediate inancial exposure.
he data set for the analysis was obtained from the leading Slovak corporate analytical
agency CRIF – Slovak Credit Bureau, s. r. o., and comprised detailed inancial statements
of a large proportion of Slovak enterprises with activities falling into all business sectors of
the Slovak economy. he data set involved all the four legal forms of enterprises common
in Slovakia (i.e. v.o.s. – general partnership, k.s. – limited partnership, s.r.o. – private limited
542 M. Boďa, V. Úradníček. he portability of Altman’s Z-score model to predicting corporate ...

company, a.s. – joint-stock company) and related to a range of 5 iscal periods: from 2009 to
2013. he fact that the data set did not originate as a random drawing from the population
of Slovak enterprises and constitutes just a selective sub-sample poses an obstacle to mak-
ing a generalization of the indings and statistical inference. Nonetheless, this is a troubling
complication of (almost) every study of this kind.
In the process of veriication, the classiication horizon of one year was applied in esti-
mating the coeicients of the discriminant function, similarly as was done in Altman (1968,
1983). his means that the classiication of enterprises into those in inancial distress and
those inancially non-distressed depended on their inancial statements released in the
next annual reporting period. hese inancial statements were screened for meeting all the
three speciied criteria of inancial distress and this resulted in separating enterprises into
two groups for the purpose of estimation. Furthermore, in the time period of 5 years from
2009 to 2013 there are a total of four sub-periods spanning two consecutive years. his
was permitted to perform estimations (or to re-estimate the Z-score model) sequentially
for the data derived from inancial statements dated to 2009, 2010, 2011 and 2012 (whilst
classiication of enterprises was based on inancial statements dated to 2010, 2011, 2012 and
2013, respectively). Only enterprises which were in the initial year found non-distressed
were included in the data sample used and their distress behaviour was tracked. It makes
sense only to predict if a non-distressed enterprise runs into inancial distress the next year
or later, and the situation that a distressed enterprise remains in inancial distress is not of
interest and is, to a degree, diicult to predict. he difusion of inancial distress over the
ive-year period for initially non-distressed enterprises in the individual years is described
in Table 1.

Table 1. Numbers of non-distressed and distressed enterprises used in the veriication analysis related
to the individual initial years 2009–2012

Year
Initial year Enterprises
2009 2010 2011 2012 2013
2009 Non-distressed: 2 414 2 193 2 102 2 043 1 989
total % (100%) (91.85%) (87.08%) (84.66%) (82.39%)
Distressed: NA 221 312 371 425
total % (9.15%) (12.92%) (15.37%) (17.61%)
2010 Non-distressed: 19 490 17 078 15 995 15 485
total % (100%) (87.62%) (82.07%) (79.45%)
Distressed: NA 2 412 3 495 4 006
total % (12.38%) (17.93%) (20.55%)
2010 Non-distressed: 27 920 24 293 23 086
total % (100%) (87.01%) (82.69%)
Distressed: NA 3 627 4 834
total % (12.99%) (17.31%)
2011 Non-distressed: 43 068 37 578
total % (100%) (87.25%)
Distressed: NA 5 490
total % (12.75%)
Technological and Economic Development of Economy, 2016, 22(4): 532–553 543

his table gives an insight into the distress behaviour of Slovak enterprises. For the
irst initial year 2009, the sample size of enterprises was rather low in comparison to the
other three initial years. Also the progression of inancial distress among the 2,414 enter-
prises marked in 2009 as non-distressed in the next few years appears somewhat delayed
as when compared to the propagation of inancial distress for non-distressed companies in
the initial years 2010, 2011 and 2012. he diferent pattern of an array of inancial distress
may be directly related to the global economic downturn of 2008–2009. In Table 1, the
entries in the irst two consecutive years for each initial year are highlighted to indicate a
data sample utilized in the re-estimation of the Z-score model (respecting the classiication
horizon of one year).
Despite the intention to emulate Altman’s methodological procedure as far as possible,
there were necessarily some unavoidable deviations following the preceding discussion.
hey encompass the redeinition of “failed” enterprises with respect to the formulated dei-
nition of inancial distress or the mismatch present between 1946–1965 US inancial items
and 2009–2013 Slovak inancial items. An important diversion from Altman’s methodol-
ogy is in using the entire dataset available for each initial year (as declared in Table 1) and
splitting it into a training sample and a test sample in the proportion of 75% to 25%. Whilst
the training sample served in the process of the re-estimation of the Z-score model, the as-
sessment of predictive accuracy stipulated use of both the training and test sample. Unless
clearly indicated and stated otherwise, the descriptive information and results presented in
the ensuing tables pertain to the full sample of enterprises for a given initial year. Altman
(1968, 1983) used the pooled sample of “failed” enterprises gathered over the entire 20-year
period (obviously in order to increase the number of efective observations).
Altman (1968, 1983) used the computational and applicational variation of discriminant
analysis as implemented in the FORTRAN codes of Cooley, Lohnes (1962, Chapters 6–7).
In order to ensure some degree of comparability with his results, this procedure was obeyed
to detail and implemented in program R using the codes compiled by one of the authors
(R Core Team 2013). he analysis was entirely afected by this program and the achieved
results are communicated in both tabular and graphical format. As in Altman (1968, 1983),
the discriminant function was speciied following the linear classiication rule (i.e. LDA)
in a usual way without an intercept, yet the cut-of value for discriminant scores that ac-
companies the classiication rule was determined additionally so that a suitably chosen
measure of training error was minimized. his goes in the wake of the procedure taken by
Altman himself. he analysis was undertaken with the use of a statistical technique of LDA
that depends upon a number of statistical assumptions (cf. e.g. Balcaen, Ooghe 2006: 67;
Sun et al. 2014: 43). Some discussion may be held on this topic (concerning their actual
necessity or validity), yet what matters in a classiication problem is the ultimate prediction
accuracy of the model built for classiication and, in consequence, whether these statistical
assumptions are indeed satisied is not relevant to the analysis and its results. his is one of
the two reasons why the validity of the statistical assumptions underlying the correct use
of LDA was not scrutinized. his is the other reason to investigate the capability of Alt-
man’s Z-score model to predict inancial distress of Slovak enterprises rather than building
a completely new model.
544 M. Boďa, V. Úradníček. he portability of Altman’s Z-score model to predicting corporate ...

he results of re-estimating the Z-score model are displayed in Table 2. For each of the
four initial years, Table 2 displays the coeicient values of the estimated discriminant func-
tion and the corresponding cut-of value allowing discrimination between inancially dis-
tressed and non-distressed enterprises. Together with the coeicients and cut-ofs answer-
ing to both the original and revised Altman’s Z-score model are displayed for convenience.
However, they difer from those reported with formulas (1) and (2) as all the discriminant
coeicients in Table 2 are normalized to unit norm. his warrants comparability, and the
cut-of values are re-calculated accordingly. It is questionable as to whether the discrimi-
nant coeicients should be required to possess the virtue of economic interpretability or
they should be treated as a black-box ingredient of a classiication model. Distractingly, the
discriminant coeicient on X4’ in each re-estimated Z-score model is very close to zero and
the discriminant coeicients on X5 for three re-estimated Z-score models have a negative
sign, though being close to zero. hough apparently being a concern in model building, it
is not troubling here as now only the Z-score model is reitted to a new setting. It is then
comprehensible that the data used with this re-estimation may have an internal structure
diferent from the one incorporated in Altman’s data. here are discernible diferences be-
tween the coeicients and cut-ofs of the individual models and there are also some trend-
ing features for the re-estimated models. One might establish that over the period of four
years, from 2009 to 2012, the predictor inancial indicators X1 and X3 rose in classiication
importance, whilst the classiication power of X2 diminished. he low coeicient values on
X4’ presumably result from an immensely high level of variability particular of this predic-
tion variable. In all probability, Altman (1968, 1983) operated with a “well-behaved” sample
of 66 US enterprises where the ratio of market/book value of equity to book value of debt
did not attain values dispersed over a large interval.
he cut-of values of the re-estimated Z-score models were not determined à la Altman
(1968, 1983) who minimized misclassiications in a rather trial-and-error way, but a more
general approach was employed instead. Following e.g. Begley et al. (1996) or Wu et al.
(2014), the cut-of value was chosen so that the sum of Type I error and Type II error rates
was minimal. Whilst Type I error comes from classifying a distressed enterprise as non-dis-
tressed, Type II error emerges with classifying a non-distressed enterprise as distressed. he

Table 2. Normalized discriminant coeicients and cut-of values of the re-estimated

and original Z-score models
Re-estimated Z-score model Altman’s model
Variable
2009 model 2010 model 2011 model 2012 model 1968 original†) 1983 revised†)
X1 0.345 0.459 0.588 0.701 0.303 0.206
X2 0.936 0.875 0.726 0.096 0.354 0.244
X3 0.016 0.156 0.356 0.706 0.834 0.895
X4¢ 0.000*) 0.000*) 0.000*) 0.000*) 0.152 0.121
X5 0.073 –0.005 –0.010 –0.005 0.253 0.287
Cut-of 0.349 0.358 0.225 0.131 0.458 0.354
Note: †) he coeicients are normalized to unit norm so as to induce comparability, and the cut-of
values are adjusted accordingly as well. *) hese coeicients are not nil.
Technological and Economic Development of Economy, 2016, 22(4): 532–553 545

error rates are then determined as the proportions of Type I error or Type II error events
to the number of distressed or non-distressed enterprises, respectively. For simpliication,
in line with Altman, Hotchkiss (2006: 244) a single value was considered as the cut-of,
discarding thus the role of the grey area. he classiication is simpliied into considering
an enterprise certainly distressed for a lower Z-score value and surely non-distressed for a
greater Z-score value than the speciied cut-of. For this reason, the cut-ofs 0.458 and 0.354
for both Altman’s Z-score models answer to the lower boundaries of the grey area, viz. 1.81
and 1.23. Enterprises falling otherwise with their Z-scores into the grey area are given the
beneit of the doubt, wherein they are considered as non-distressed.
A more detailed geometrical analysis of the normalized discriminant coeicients sug-
gests that great similarity is unsurprisingly observed for both the original and revised Alt-
man’s Z-score models and is similar to the re-estimated Z-score models corresponding to
the initial years 2009, 2010 and 2011. he re-estimated Z-score model related to the initial
year 2012 deviates from either of the two groups. his analysis goes for reasons of space
unreported in the paper.
For each initial year, a total of three Z-score models were inspected for their predictive
behaviour assessed in respect of a one-year classiication horizon (the outcome of which
is reported in Tables 3 and 4) and in respect to various classiication horizons (the results
of which are shown in Fig. 1).
Separately for the respective training and test sample in the four initial years, Table 3
summarizes the classiication accuracy of both the original and revised Altman’s Z-score
models and of the re-estimated Z-score model. Using the overall accuracy (the total pro-

Table 3. Classiication accuracy of the re-estimated and Altman’s Z-score models

1968 original 1983 revised Re-estimated

% of correctly Altman’s model Altman’s model Z-score model
Initial
classiied
year Training Test Training Test Training Test
enterprises
sample sample sample sample sample sample
Overall 77.89% 79.01% 82.42% 83.14% 61.53% 60.33%
2009 Non-distresses 81.69% 81.60% 87.71% 87.25% 60.04% 58.47%
Distresses 40.00% 53.57% 29.70% 42.86% 76.36% 78.57%
Overall 76.99% 78.21% 80.45% 80.55% 56.26% 56.93%
2010 Non-distresses 83.00% 84.05% 87.96% 88.03% 53.41% 54.03%
Distresses 34.44% 36.82% 27.31% 27.53% 76.45% 77.45%
Overall 76.81% 76.15% 79.45% 79.32% 66.07% 66.14%
2011 Non-distresses 82.64% 82.47% 87.08% 87.42% 66.04% 66.38%
Distresses 37.79% 33.85% 28.35% 25.03% 66.25% 64.50%
Overall 76.56% 76.83% 79.66% 79.51% 70.53% 70.54%
2012 Non-distresses 82.04% 82.38% 86.75% 86.84% 72.92% 72.78%
Distresses 39.08% 38.82% 31.07% 29.35% 54.19% 55.21%
Note: he best classiication performance in terms of accuracy is highlighted in bold for each initial
year sample and for each of the three accuracy rates presented.
546 M. Boďa, V. Úradníček. he portability of Altman’s Z-score model to predicting corporate ...

portion of correctly classiied enterprises), the true negative rate (the proportion of non-
distressed enterprises that are correctly classiied as non-distressed), and true positive rate
(the proportion of distressed enterprises that are assigned a correct classiication of being
distressed), the simplest, yet most informative, measures of predictions are made. As a mat-
ter of fact, these three proportions give a percentage of all correctly classiied enterprises,
non-distressed enterprises and distressed enterprises, respectively. It is worth noticing that
the true negative rate is a complement of the Type II error rate to unity whereas true posi-
tive rate complements Type I error to unity. he best accuracy delivered for both the train-
ing and test sample for each initial year and for each group (all enterprises, non-distressed
enterprises and distressed enterprises) is highlighted in bold. It transpires that the best
predictive performance in terms of one-year prediction accuracy is uniformly attained
by the revised Altman’s Z-score model, whereas the re-estimated model classiies most
accurately distressed enterprises. herefore, if the goal is to classify correctly an arbitrary
enterprise, the revised Altman’s Z-score model should be the choice, but if the emphasis
is placed on correct classiications of distressed enterprises, then the re-estimated Z-score
model should be given preference.
he global predictive performance of the Z-score models may be further assessed and
visualized by means of receiver-operating characteristic (ROC) curves. In order to conserve
space, the ROC curves are not included in this presentation, and the information that they
convey is a proxy for the area-under-curve (AUC) statistic that summarizes the predictive
performance generated by a ROC curve for various choices of the cut-of value. he AUC
statistics for the three Z-score models and for the four initial years are shown in Table 4
separately for the training and test sample. Again, the most favourable values are indicated
in bold face and they point to the re-estimated Z-score model delivering comparatively
the best predictive performance (which seems yet only slightly better than the predictive
performance of the other two Z-score models). Each Z-score model attains only fair pre-
dictive performance.
Both Table 3 and Table 4 show decreasing trends in terms of classiication accuracy
and predictive performance of the three Z-score models. Although all the three Z-score
models maintain almost identical classiication accuracy or predictive performance for both
the training and test sample (as usually for the test sample classiication models have a

Table 4. AUCs of the re-estimated and Altman’s Z-score models

Initial year 1968 original 1983 revised Re-estimated
Altman’s model Altman’s model Z-score model
Training Test Training Test Training Test
sample sample sample sample sample sample
2009 0.708 0.632 0.701 0.625 0.740 0.736
2010 0.649 0.626 0.639 0.616 0.710 0.703
2011 0.636 0.650 00.624 0.639 0.708 0.715
2011 0.642 0.640 0.631 0.632 0.684 0.673
Note: he best classiication performance in terms of AUC is highlighted in bold for each initial year
sample.
 Technological and Economic Development of Economy, 2016, 22(4): 532–553 547

tendency to deteriorate), their predictive quality somewhat descends over the four initial
years. his is especially seen for the true positive rates indicating that the ability of these
Z-score models to predict the distress condition of distressed enterprises diminishes. For
the re-estimated Z-score model this fall is about 22 percentage points.
Eventually, Figure 1 reports in a layout of six charts on the predictive behaviour of the
three Z-score models that they achieve for various classiication horizons. he models’
predictive behaviour is measured and visualized with respect to classiication accuracy of
all enterprises (overall accuracy) and distressed enterprises (true positive rate). While the
three charts in the irst row of the layout show overall accuracy, the other three charts in
the second row then display the true positive rate. he classiication accuracies of the three
Z-score models presented in Figure 1 are derived from the data sample pertaining to the
initial year 2009 (see the numbers of non-distressed and distressed enterprises over the
period from 2009–2013 in the irst row of Table 1). he predictive behaviour for the other
three initial years 2010, 2011 and 2012 is not reported here in the paper since it is very
much the same. Each of the six charts shows how the classiication accuracy of the Z-score
model behaves with respect to diferent timings of predictions and to diferent classiication
horizons. For the predictions starting in 2009, the data sample of 2,414 enterprises being
non-distressed in 2009 was used and their distress condition was predicted for 2010, 2011,
2012 and 2013, which implies the classiication horizon of one year, two years, three years
and four years, respectively. he predictions starting in 2010 related to the 2,193 enterprises
that were non-distressed in 2010. he distress condition of these 2,193 enterprises was
predicted for the next three years 2011, 2012 and 2013 implying a classiication horizons
of one, two and three years. his pattern was subsequently repeated for the starts in 2011
and 2012 and in each start predictions for the classiication accuracy of each of the three
Z-score models was evaluated and inally plotted in a respective chart in Figure 1. Not
only does this design reveal the ability of the Z-score models to retain their classiication
accuracy with respect to diferent classiication horizons, but it also shows how the clas-
siication accuracy of the Z-score models react to the changing economic conditions that

1968 original Altman's model 1983 revised Altman's model Re-estimated 2009 Z-score model
True positive rate

True positive rate

True positive rate

80 80 80
(% classified)

(% classified)

(% classified)

70 70 70

60 60 60
1 2 3 4 1 2 3 4 1 2 3 4
Classification horizon in years Classification horizon in years Classification horizon in years
True positive rate

True positive rate

True positive rate

70 70 70
(% classified)

(% classified)

(% classified)

50 50 50
30 30 30
1 2 3 4 1 2 3 4 1 2 3 4
Classification horizon in years Classification horizon in years Classification horizon in years

2009 2010 2011 2012

Fig. 1. Classiication accuracy of the re-estimated and Altman’s Z-score models

for various classiication horizons using the sample related to the initial year of 2009
548 M. Boďa, V. Úradníček. he portability of Altman’s Z-score model to predicting corporate ...

may have happened in the investigated time span of ive years. he global economic crisis
that demonstrated its efects mainly between 2007 and 2009 brought in some instability in
the Slovak economic environment and was followed by a slow stabilization accompanied
by a number of structural and legislative changes. Although the Slovak economy oicially
did not experience a recession, under scrutiny, all these factors revealed themselves in
the period from 2009–2013 and also afected the business environment and the distress
condition of Slovak enterprises. It is only understandable that this impacts the predictive
capability of the Z-score models.
hough the patterns displayed in Figure 1 apply to the initial year 2009, the patterns for
the other three initial years are almost identical. hey aford to make three fairly general
observations. First, the overall accuracy of both the original and revised Altman’s Z-score
model is for any classiication horizon higher than that of the re-estimated Z-score model.
Yet, the best classiication performance in terms of true positive rates (i.e. the classiication
accuracy of distressed enterprises) is globally achieved by the re-estimated Z-score model.
Second, the classiication horizon does not seem to be a major factor that afects the clas-
siication accuracy of the Z-score models. here are some oscillations or signs of a trending
behaviour in classiication accuracy displayed in the individual charts of Figure 1, but they
are virtually irrelevant and may not be given much attention. Lastly, the later the start of
predictions, the lower is in general the classiication accuracy of the Z-score models. his
is especially true for the overall accuracy of the original and revised Altman’s models and
for the true positive rate of the re-estimated Z-score model.

Conclusions and discussion

In spite of the methodological uncertainties associated with employing Altman’s Z-score

model in predicting inancial distress of Slovak enterprises, it looks that, to some degree, his
model works and yields satisfactory predictive performance. his statement is at least valid
for the sample of Slovak enterprises permitting the empirical veriication of the model that
was conducted in a setting of inancial distress prediction and focused on the Slovak eco-
nomic conditions. he warnings raised in the second section against the use of the Z score
model seem to be unfounded since both formulations proposed by Altman (1968, 1983)
or its re-estimated version function acceptably. Several reasons why the Z-score model
is usable for predicting inancial distress of Slovak enterprises may be identiied and are
compiled as follows:
– Firstly, the ive inancial indicators acting as the predictors in the model were chosen
favourably so that they include one liquidity ratio (X1), two proitability ratios (X2
and X3), one leverage ratio (X4 or rather X4’) and one activity ratio (X5). Since each
partial sphere of corporate inancial activity is represented by one or two inancial
indicators, inancial diiculties presenting themselves in at least one partial sphere
should be easily captured by the Z-score model and any sign of inancial distress
should, in theory, be easily mapped into a respective Z-score value. An enterprise
in inancial distress experiencing consistent operating losses faces the shrinking of
current assets in relation to total assets, which touches on liquidity (captured by X1)
Technological and Economic Development of Economy, 2016, 22(4): 532–553 549

and undergoes a decreasing earning power of its assets both in a given period and
also on a cumulative basis (which is seized by X2 and X3). A inancially distressed
company on the verge of insolvency necessarily has debt well in excess of its equity
regardless of whether equity is quantiied by the market value or by the book value
(this is embodied in X4 or X4’). Finally, the ability of such an enterprise to generate
sales through its assets is lowered and a fall in sales might expectably ensue (which is
then relected in X5). All these ive indicators have a sound economic justiication and
this quintet does represent a convenient instrument of identifying inancial diicul-
ties of enterprises in any economic environment and any period. It is not important
therewith as to whether inancial diiculties are understood in juridical terms and
equalled with bankruptcy, or they are given a purely economic treatment and inter-
preted in general as inancial distress.
– Second, the adopted deinition of inancial distress that is regarded as suitable for
the Slovak economic conditions is in close relation to the ive predictive inancial
indicators. he irst prerequisite for a inancially distressed enterprise that its equity
be negative should be forecast directly by X4’ and partly as well by X2. he second
prerequisite that the EAT be negative is tied closely with X3 and the third prerequisite
that the current ratio be lower than 1 is in a direct link with X1. When taken on a
univariate basis, these four inancial indicators should themselves be suicient to in-
dicate one aspect of the inancial condition of a distressed enterprise. herefore, with
respect to the chosen and vindicated deinition of inancial distress, it is unsurprising
that the model is fairly successful in predicting inancial distress.
– hird, under the maxim “the simpler a model, the better”, Altman’s Z-score model is
in the frame for theoretically good predictive performance. Balcaen, Oooghe (2006:
81) as well as Sun et  al. (2014: 53) highlight the popularity of simpler bankruptcy
prediction models thanks to their good classiication abilities. he Z-score model
works only on ive inancial indicators and is derived with the aid of LDA. Its con-
ceptual simplicity underpinned with a well thought out choice of predictive inancial
indicators justiied on economic grounds may be another reason for its comparative
acceptable predictive performance even outside its home environment in a diferent
time period.
One still may question another element of Altman’s Z-score model and its importabil-
ity to the Slovak economic conditions and this is the fact that it is employed to identify
inancial distress of enterprises using inancial statements one year preceding the distressed
condition. here is also the trend of including cash-low variables into bankruptcy predic-
tion models (Balcaen, Ooghe 2006: 80; Platt, H. D., Platt, M. B. 2006: 146), but these are not
represented in the Z-score model. However, it has been established through empirical veri-
ication that common ratio indicators (such as those made use in the Z-score mode) reveal
a good classiication capacity the year immediately prior to bankruptcy (which is consistent
with the classiication horizon of one year of the Z-score model), whilst cash-low variables
would need two or three years prior to bankruptcy to detect the distress condition (cf. Alt-
man, Hotchkiss 2006: 251). here is not much let to doubt about the potential of the ive
inancial ratios in the Z-score model to predict inancial distress.
550 M. Boďa, V. Úradníček. he portability of Altman’s Z-score model to predicting corporate ...

It further transpires that in terms of classiication accuracy the Z-score model applied
in the Slovak economic conditions can be trustworthy and compete with its application in
the home US environment. When transcribing the classiication accuracies of the Z-score
model portrayed in Figure 1 and the analogous igures for the initial years 2010–2012 into a
tabular format, the result is the summary presented in Table 5. he classiication accuracies
of the 1968 Altman’s original model, the 1983 Altman’s revised model and the re-estimated
Z-score models that were found for the data sample of Slovak enterprises utilized in the
paper are appended by the one-year classiication accuracies that the 1968 Altman’s original
model revealed for a sample of US enterprises. hese additional classiication accuracies
included for comparison purposes in Table 5 are those reported by Altman, Hotchkiss
(2006: 244).

Table 5. Classiication accuracy of the Z-score models for the sample of Slovak enterprises
and for the US economy
Years prior to distress
Z-score model and data
1 2 3 4
1968 Altman’s original 77.6–78.2% 74.7–76.6% 73.1–74.3% 73.9%
1983 Altman’s revised 79.4–82.6% 76.8–80.4% 75.3–78.6% 77.2%
Slovak 2009 re-estimated 61.2–65.2% 61.3–65.0% 60.2–64.8% 60.8%
2009–2013
data 2010 re-estimated 56.4–56.8% 56.7–57.5% 57.8% NA
2011 re-estimated 66.1–66.4% 65.6% NA NA
2012 re-estimated 70.5% NA NA NA
US 1968 data 88% NA NA NA
US 1969–1975 data 75% NA NA NA
1968 Altman’s original
US 1976–1995 data 78% NA NA NA
US 1997–1999 data 84% NA NA NA
Note: If applicable, the classiication accuracy is reported as a range of percentages. he percentages for
the US data are those reported by Altman, Hotchkiss (2006: 244).

It follows from Table 5 that both the 1968 Altman’s original model and its 1983 revi-
sion provide quite reliable prediction rules of the distress status of Slovak enterprises not
only one year preceding the distress condition but their prediction ability extends over a
longer classiication horizons. Understandably, accuracy tends to a decrease commensu-
rate with prolongation of the classiication horizon, yet the classiication accuracy of the
original and revised Altman’s model displayed for the Slovak data sample for three or four
years preceding the distress condition is comparable to the classiication accuracy of the
original Altman’s model shown for the US 1969–1975 or 1976–1995 data. he revised
Altman’s model especially manifests good prediction performance with respect to overall
classiication accuracy. he Z-score models with re-estimated coeicients are not capable
of classifying with a satisfactory level of accuracy, being positively outperformed by either
of Altman’s models with coeicients itted for the US economy. his is only suggestive that
the Z-score model is portable outside the domestic US environment and can be used under
Slovak conditions. Nevertheless, as argued in the third section of the paper, if the ambition
Technological and Economic Development of Economy, 2016, 22(4): 532–553 551

is to minimize the number of misclassiications of inancially distressed enterprises, which

is in fact a standard and usual aim of inancial distress prediction, then it is suitable to re-
estimate the coeicients of the Z-score model.
All in all, the conclusion is thus that Altman’s bankruptcy formula is suitable for Slovak
economic conditions and useful for predicting inancial distress of Slovak enterprises. his
ultimate statement naturally does not mean that Altman’s model cannot be improved on
e.g. by accommodating an additional set of predictive inancial indicators or by using a
diferent classiication technique and this quest goes beyond the intentional scope of the
paper. hat said, there still are a good many directions for further research in inancial
distress prediction of Slovak enterprises. One direction is connected to the presumable
non-existence of a universal deinition of inancial distress applicable to the Slovak eco-
nomic conditions. Such a deinition would have to capture and relect all the aspects of the
inancial diiculties that Slovak enterprises must face in their business activities. In addi-
tion, its introduction would also necessitate careful economic reasoning supported with a
detailed analysis of the distress condition of Slovak enterprises. It is not impossible that it
might be based on information reported in inancial statements from several iscal years.

Funding

his work was supported by Vedecká grantová agentúra Ministerstva školstva, vedy, výs-
kumu a športu Slovenskej republiky a Slovenskej akadémie vied under Grants [# 1/0647/14
and # 1/0554/16].

References
Altman, E. I. 1968. Financial ratios, discriminant analysis and the prediction of corporate bankruptcy,
he Journal of Finance 23(4): 583–609. http://doi.org/10.1111/j.1540-6261.1968.tb00843.x
Altman, E. I. 1983. Corporate inancial Distress: a complete guide to predicting, avoiding, and dealing
with bankruptcy. 1st ed. New York: Wiley.
Altman, E. I.; Hotchkiss, E. 2006. Corporate inancial distress and bankruptcy: predict and avoid bank-
ruptcy, analyze and invest in distressed debt. 3rd ed. New York: Wiley.
http://doi.org/10. 1002/9781118267806
Altman, E. I. 2013. Predicting inancial distress of companies: revisiting the Z-score and ZETA ® mod-
els, Chapter 17 in A. R. Bell, C. Brooks, M. Prokopczuk (Eds.). Handbook of research methods and
applications in empirical inance. Glos, UK: Edward Elgar Publishing.
http://doi.org/10.4337/ 9780857936097
Altman, E. I.; Haldeman, R. G.; Narayanan, P. 1977. ZETA™ analysis: a new model to identify bank-
ruptcy risk of corporations, Journal of Banking and Finance 1(1): 29–54.
http://doi.org/10. 1016/0378-4266(77)90017-6
Altman, E. I.; Iwanicz-Drozdowska, M.; Laitinen, E. K.; Suvas, A. 2014. Distressed irm and bankruptcy
prediction in an international context: a review and empirical analysis of Altman’s Z-Score model,
in International Risk Management Conference 2014, Seventh Edition “he Safety of the Financial
System: From Idiosyncratic to Systemic Risk”, 23–24 June 2014, Warsaw, Poland.
Asquith, P.; Gertner, R.; Scharfstein, D. 1994. Anatomy of inancial Distress: an examination of junk-
bond issuers, he Quarterly Journal of Economics 109(3): 625–658. http://doi.org/10.2307/ 2118416
552 M. Boďa, V. Úradníček. he portability of Altman’s Z-score model to predicting corporate ...

Balcaen, S.; Ooghe, H. 2006. 35 years of studies on business failure: an overview of the classic statistical
methodologies and their related problems, he British Accounting Review 38(1): 63–93.
http://doi.org/10.1016/j.bar.2005.09.001
Begley, J.; Ming, J.; Watts, S. 1996. Bankruptcy classiication errors in the 1980s: an empirical analysis
of Altman’s and Ohlson’s models, Review of Accounting Studies 1(4): 267–284.
http:// doi.org/10.1007/BF00570833
Chrastinová, Z. 1998. Metódy hodnotenia ekonomickej bonity a  predikcie finančnej situácie
poľnohospodárskych podnikov. Bratislava: Výskumný ústav ekonomiky poľnohospodárstva a  po-
travinárstva (in Slovak).
Cooley, W. W.; Lohnes, P. R. 1962. Multivariate procedures for the behavioral sciences. New York: Wiley.
Delina, R.; Packová, M. 2013. Validácia predikčných bankrotových modelov v podmienkach SR, E&M
Ekonomie a Management 16(3): 101–113 (in Slovak).
Dimitras, A. I.; Zanakis, S. H.; Zopounidis, C. 1996. A survey of business failures with an emphasis on
prediction methods and industrial applications, European Journal of Operational Research 90(3):
487–513. http://doi.org/10.1016/0377-2217(95)00070-4
Dopravný podnik mesta Prešov. 2015. Výročná správa za rok 2014. Annual Report. Ľubotice: Dopravný
podnik mesta Prešov, a. s. (in Slovak).
Gilbert, L. R.; Menon, K.; Schwartz, K. B. 1990. Predicting bankruptcy for irms in inancial distress,
Journal of Business Finance & Accounting 17(1): 161–171.
http://doi.org/10.1111/j.1468-5957. 1990.tb00555.x
Grice, J. S.; Dugan, M. T. 2001. he limitations of bankruptcy prediction models: some cautions for the
researcher, Review of Quantitative Finance and Accounting 17(2): 151–166.
http://doi.org/ 10.1023/A:1017973604789
Grice, J. S.; Ingram, R. W. 2001. Tests of the generalizability of Altman’s bankruptcy prediction model,
Journal of Business Research 54(1): 53–61. http://doi.org/10.1016/S0148-2963(00)00126-0
Gurčík, Ľ. 2002. G-index – metóda predikcie inančného stavu poľnohospodárskych podnikov, Agri-
cultural Economics (Zemědělská Ekonomika) 48(8): 373–378 (in Slovak).
Hillier, D.; Ross, S. A.; Westerield, R. A.; Jafe, J.; Jordan, B. D. 2010. Corporate inance. European ed.
Berkshire: McGraw-Hill.
Hrdý, M.; Šimek, B. 2012. Valuation of the company in inancial distress, E&M Ekonomie a Manage-
ment 15(4): 121–133.
I.S.D.D. plus 2014. Výročná správa za rok 2013. Annual Report. Bratislava: I.S.D.D. plus, s. r. o. (in
Slovak).
Kalouda, F.; Vaníček, R. 2013. Alternative bankruptcy models for CR conditions (concept and empiri-
cal veriication), in M. Čulík (Ed.). Financial Management of Firms and Financial Institutions. 9th
International Scientiic Conference Proceedings, Part I. Ostrava: VŠB –Technická univerzita Ostrava,
316–324.
Király, P.; Kotulič, R.; Smolko, M. 2015. Sotvér na inančno-ekonomickú analýzu irmy SOFINA Version
3.01. Sotware. (in Slovak).
Kopta, D. 2009. Possibilities of inancial health indicators used for prediction of future development of
agricultural enterprises, Agricultural Economics (Zemědělská Ekonomika) 55(3): 111–125.
Kotulič, R.; Király, P.; Rajčániová, M. 2010. Finančná analýza podniku. 2 ed. Bratislava: Iura Edition.
(in Slovak).
Kúpele Trenčianske Teplice. 2012. Výročná správa 2011. Annual Report. Kúpele Trenčianske Teplice, a. s,
Trenčianske Teplice (in Slovak).
Li, J. F. 2012. Prediction of corporate bankruptcy from 2008 through 2011, Journal of Accounting and
Finance 12(1): 31–41.
Technological and Economic Development of Economy, 2016, 22(4): 532–553 553

Ministerstvo hospodárstva Slovenskej republiky. 2014. Analýza stavu malého a stredného podnikania.
Analytical Document Bratislava. Ministerstvo hospodárstva Slovenskej republiky (in Slovak).
Platt, H. D.; Platt, M. B. 2006. Understanding diferences between inancial distress and bankruptcy,
Review of Applied Economics 2(2): 141–157.
Pitrová, K. 2011. Possibilities of the Altman Zeta model application to Czech irms; E&M Ekonomie
a Management 14(3): 66–76.
Púchovský mäsový priemysel. 2014. Výročná správa za rok 2013. Annual Report. Púchovský mäsový
priemysel, a. s., Púchov (in Slovak).
R Core Team. 2013. R: a language and environment for statistical computing (Version 3.0.2). (Sotware).
Slovak Business Agency. 2014. Analýza ukazovateľov inančnej výkonnosti malých a stredných podnikov
v rokoch 2008–2012. Analytical Document. Slovak Business Agency, Bratislava (in Slovak).
Slovak Business Agency. 2015. Analýza ukazovateľov inančnej výkonnosti malých a stredných podnikov
v roku 2013. Analytical Document. Slovak Business Agency, Bratislava (in Slovak).
Stolárik, M. 1996. Meranie inančného zdravia iriem, Hospodárske noviny, 3 March 1996, p.  19 (in
Slovak).
Strachotová, A. 2012. Finančná pozícia neinančných podnikov, Biatec 20(1): 2–9 (in Slovak).
Sun, J.; Li, H.; Huang, Q.-H.; He, K.-Y. 2014. Predicting inancial distress and corporate failure: a review
from the state-of-the-art deinitions, modeling, sampling, and featuring approaches, Knowledge-
Based Systems 57: 41–56. http://doi.org/10.1016/j.knosys.2013.12.006
Šnircová, J. 1997. Možnosti prognózovania inančnej situácie podnikov v slovenskej ekonomike, Biatec
5(4): 15–22 (in Slovak).
Štrangfeldová, J. 2012. Finančno-ekonomická analýza ex ante v lekárni, Poradca lekárnika 8(10): 13–16
(in Slovak).
Taler, R. J. 1984. Empirical models for the monitoring of UK corporations, Journal of Banking and
Finance 8(2): 199–227. http://doi.org/10.1016/0378-4266(84)90004-9
Wu, Y.; Gaunt, C.; Gray, S. 2010 A comparison of alternative bankruptcy prediction models, Journal
of Contemporary Accounting & Economics 6(1): 34–45. http://doi.org/10.1016/j.jcae.2010. 04.002
Zákon č. 7/2005 Z. z. o konkurze a reštrukturalizácii a o zmene a doplnení niektorých zákonov v zn. n. p.
[Bankruptcy Act No. 7/2005 Coll.] (in Slovak).
Zalai, K. 1997. Význam a metódy prognózovania inančnej situácie irmy, Biatec 5(3): 15–22 (in Slovak).
Zalai, K. 2000. Osobitosti prognózovania inančného vývoja slovenských podnikov, Biatec 8(1): 14–16
(in Slovak).

Martin BOĎA (Slovakia, 1984) orientates his scientiic work to applications of a wider spectrum of
quantitative methods in inance and economics. His scientiic preparation is founded in the area of
economics and inance (inance, banking and investment; tourism) and of quantitative methods (math-
ematical statistics and inancial mathematics; probability and mathematical statistics). His professional
orientation may be broken into three chief ields: (1) eiciency measurement in banking, (2) applica-
tions of data envelopment analysis in inance, (3) inancial risk measurement. Martin Boďa is employed
as a lecturer at the Faculty of Economics of Matej Bel University in Banská Bystrica, Slovakia.

Vladimír ÚRADNÍČEK (Slovakia, 1963) has his scientiic work continually directed towards applica-
tion of quantitative methods in the area of inance, corporate economics and management. His scien-
tiic and professional activities are targeted at four compact parts: (1) pricing of inancial derivatives, (2)
inancial mathematics and inancial analytics, (3) corporate inancial health prediction, and (4) inves-
tigation of economic sustainability and convergence issues. Vladimír Úradníček currently occupies the
positon of Vice-Dean at the Faculty of Economics of Matej Bel University in Banská Bystrica, Slovakia.

View publication stats