Assessment of Altman’s Z-Score and Modifications for Predicting Bankruptcy in Germany

Research Methods
An assessment of Altman’s z-score and
modifications as a technique for predicting
bankruptcy: Empirical evidence in Germany.
Word Count: 2,493
Deadline: Thursday 8th December 2016
For the attention of Neil Lancastle
Executive Summary
This proposal will outlay the details of a potential research project, investigating the
effectiveness of Altman’s (1968) Z-score model for predicting bankruptcy. The main
contribution of this paper is to provide additional evidence that the Z-score remains
effective. The work of Almamy, Aston and Ngwa (2016) will be mimicked here, and their
J-UK model will also be assessed. This longitudinal study will test these models on a
sample of approximately 200 German firms. A risk analysis was conducted and ethical
consideration has been discussed in order to determine the feasibility of this research
project.

Abbreviations
Beaver (1966) A highly popular univariate method of predicting bankruptcy
using a series of financial ratios.
Bloomberg Online financial data source.
DMU De Montfort University.
EBIT Earnings Before Interest and Tax, used by Almamy, Aston and
Ngwa (2016) to establish a cash flow variable.
FMEA Failure Mode and Effects Analysis used to provide a risk
assessment for a foreseeable project.
J-UK Model A modified version of Altman’s Z-score, incorporating a cash
flow variable, established by Almamy, Aston and Ngwa (2016).
MDA Multiple Discriminant Analysis, used by Altman to develop the
first multivariate bankruptcy prediction model, namely Z-Score.
Multivariate A method using multiple factors within a model.
PRN Priority Risk Number (PRN), calculated as part of FMEA.
O-Score A popular multivariate method: Ohlson’s (1980) O-Score for
bankruptcy prediction.
Turnitin Online software used to check student’s work for plagiarism
and bad academic practice.
U.K. The United Kingdom.
U.S. United States (of America).
Univariate A method using a single factor model.
Z-Score A highly popular multivariate method: Altman’s (1968) Z-Score
for bankruptcy prediction.
Table of Contents
Introduction to the Research Topic ................................................................................. 4
Literature Review ............................................................................................................ 5
Key Theories ............................................................................................................... 5

Research on Bankruptcy Prediction Models ................................................................ 5
Research on Altman’s Z-Score .................................................................................... 6
Intended Contribution, Research Design & Research Questions .................................... 7
Contribution ................................................................................................................. 7
Research Design ......................................................................................................... 7
Research Questions .................................................................................................... 9
Ethical Considerations ................................................................................................... 10
Research Plan ............................................................................................................... 11
Conclusion .................................................................................................................... 12
References ............................................................................................................... 13
Bibliography ............................................................................................................. 16
Appendices ................................................................................................................... 18
Appendix 1 – German Economy ................................................................................ 18
Appendix 2 – Gantt Chart .......................................................................................... 20
Preliminary Research Plan ..................................................................................... 20
Research Plan........................................................................................................ 21
Appendix 3 – FMEA Risk Analysis............................................................................. 22
Rating Scale ........................................................................................................... 22
FMEA Analysis ....................................................................................................... 23

Introduction to the Research Topic
When looking into the history of bankruptcy prediction models, “research up to the
mid1960’s focused on univariate (single factor/ratio) analysis. The most widely
recognised univariate study is that of Beaver [1966]” (Bellovary, Giacomino and Akers,
2007).
Moving forward in time, “univariate studies...laid the groundwork for multivariate
bankruptcy prediction models” (Bellovary, Giacomino and Akers, 2007). Altman (1968)
argued; “traditional ratio analysis is no longer an important analytical technique...due to
the relatively unsophisticated manner in which it has been presented.” The work of
Altman was the first multivariate study on bankruptcy prediction and remains
creditworthy almost fifty years later. “The Multiple Discriminant Analysis (MDA)
technique has the advantage of considering an entire profile of characteristics common
to the relevant firms, as well as the interaction of these properties” (Altman, 1968).
However, this research analysed a small, restricted sample over a short time-period; “91
manufacturing firms from the 19581961 period” (Grice and Dugan, 2001).
This research will assess the German economy as according to Bloomberg, Germany is
among the top three most innovative economies in the world (Jamriskom and Lu, 2016).
Yet since the 1970s, the number of companies going bankrupt continues to rise (see
Appendix 1). The researcher here will somewhat replicate the work of Almamy, Aston &
Ngwa (2016), assessing Altman’s Z-Score within the German economy. This will also
influence the credibility of their modified Z-score model to broader countries and add to
its generalisability within today’s globalised economy.
Literature Review
The Chartered Association of Business Schools provides an academic guide to the most
credible journals. By using the 2015 guide, filtering to the field of finance, the researcher
used journals rated three or above to ensure a high quality of literature was obtained.
The researcher then used De Montfort University’s databases only to search for
keywords to find relevant literature. Keywords included; Bankruptcy prediction; Altman’s
Z-Score; Zscore; Financial ratios; Bankruptcy models; Business failure prediction.
Key Theories
In Beaver’s study, “for every set of financial statements available, 30 ratios were
computed”, therefore this method had the ability to provide detailed analysis and
Research Proposal - Part 1
Page 4

became the basis for future bankruptcy prediction models. (Beaver, 1966). However,
computing all ratios can prove complex and is arguably a simplistic method.
Additionally, the depth of analysis in univariate studies is capped, as one ratio fails to
provide a complex insight into a firm’s stability and disregards additional risk factors. To
contradict this, “Beaver [1966] was able to predict bankruptcy with 92% accuracy using
only one ratio. Jo et al’s [1997] model that considered 57 factors yielded only an 86%
accuracy rate” (Bellovary, Giacomino and Akers, 2007). As the first multivariate model,
Altman (1968) showed that discriminant analysis held greater accuracy when predicting
bankruptcy, and therefore gained popularity immediately. If a company scores below
1.81, this indicates severe risk of bankruptcy (Bandyopadhyay, 2006). A drawback was
that only U.S. manufacturing firms were analysed, therefore the generalisability of this
study is limited. Ohlson (1980) developed another popular multivariate model, this time
incorporating nine variables, arguably more accurate than the Z-score yet somewhat
more complex to compute.
Research on Bankruptcy Prediction Models
Lawrence and Bear (1986) tested the influence of leases on failure, based on Beaver
(1966). They found that this additional variable had no influence on the model’s
predictive ability. However, this method tested a sample that was not selected randomly,
therefore it may not be fully representative of the population. Beaver, Correia and
Mcnichols (2012) found that the accuracy of methods based on accounting figures was
slowly declining, suggesting the use of financial ratios as a bankruptcy prediction tool
was somewhat outdated. Conversely, this research did find that ratios assessed
independently proved significant.
Grice and Dugan (2001) tested another bankruptcy model: Zmijewski (1984) and Ohlson
(1980). While both methods proved highly accurate in their original studies, the
conclusions here highlighted that the level of accuracy deteriorated when applied over
time. This study used different sample sizes and time periods for the two categories,
which could have created bias or skewed the results. Yet Kwak, Shi and Kou (2012)
explained that the Z-score and O-score were among the most accurate bankruptcy
models. This is limiting as the research cannot be generalised to different time periods.
Research on Altman’s Z-Score
Katz, Lilien and Nelson (1985) found results consistent with Altman’s when they tested a
larger sample size within the U.S. market. However, one of their categories of firms was
not significant at a 5% confidence level, somewhat hindering their conclusions. Altman,
Danovi and Falini (2013) assessed 370 Italian manufacturing companies using the
Research Proposal - Part 1
Page 5

ZScore model. They found this method was highly effective, however, this paper also
suffered from limited generalisability as they focused on one industry. Wang and
Campbell (2010) suffered the same limitations when they tested manufacturing
companies in China.
On the contrary, Mossman, Bell and Turtle (1998) used the U.S. market to test multiple
bankruptcy prediction models, including Altman’s. They concluded that over time, the
accuracy of using just one model decreased, and specific models were better suited to
particular scenarios. Grice and Ingram (2001) found the same limitations; Altman’s
model was easily swayed by industry influences which significantly reduced its
accuracy. These findings were also consistent with Hillegeist et al (2004), as their model
proved a better predictor than the accounting-based methods of Altman (1968) and
Ohlson (1980). Agarwal and Taffler (2007) and Almamy, Aston and Ngwa (2016) both
tested the U.K. market using the Z-score and their own modified version of the model.
Both studies found their own model to be more superior at predicting bankruptcy than
the original Z-score method, however, both suffered from Type I and Type II errors.
Additionally, while the JUK model proved more accurate than the original Z-score
model, Almamy, Aston and Ngwa (2016) did not provide justification for why this specific
variable would aid the predictive ability of this model.
“One of the criticisms of failure prediction models in general, is that they have not been
tested on an entire underlying population” (Wang and Campbell, 2010). Additionally, this
literature is heavily based on research within the U.S. market. Therefore, this research
will consist of a sample across numerous industries in Germany. There seems to be
some uncertainty about the predictive ability of the Z-score, therefore further research
will provide a deeper insight. The researcher will also test the J-UK model on German
companies, developed by Almamy, Aston and Ngwa (2016).
Research Proposal - Part 1
Page 6