Syntax Literate: Jurnal Ilmiah Indonesia p–ISSN: 2541-0849
e-ISSN: 2548-1398
Vol. 9, No. 3, Maret 2024
EVALUATION AND
RE-ESTIMATION OF BANKRUPTCY PREDICTION MODELS IN FACING THE CRISIS PERIOD IN
INDONESIA (2017-2022)
Mentari M. Lubis*, Imo
Gandakusuma
Universitas Indonesia, Depok, West Java, Indonesia
Email: [email protected]*
Abstract
This research was conducted to see how suitable the existing bankruptcy
prediction models that have been used in other countries to be used during the
crisis in Indonesia. The data used in research are companies in Indonesia
registered in the Indonesia Stock Exchange (IDX). The re-estimate of the
coefficient of variables models is carried out and then the bankruptcy
prediction of the re-estimation model is re-calculated. The results of the
bankruptcy prediction of the re-estimate model are then compared with the
results of the bankruptcy prediction of the original model to see whether the
model can be used during the crisis in Indonesia. The results of the study is
that Springate original model is the most suitable model for the conditions in
Indonesia during the crisis caused by the COVID-19 pandemic. The Springate
model has the highest financial distress prediction accuracy, while the Altman
Emerging Market model produces the highest Error Type I.
Keywords: Financial distress, bankruptcy, re-estimate models, developing
countries, Altman, Ohlson, Springate
Introduction
The COVID-19 pandemic that is occurring globally has actually had a lot of impact on the global economy. The COVID-19 pandemic has triggered substantial risk across financial markets, which is also highly correlated with the level of investor panic. COVID-19 has been shown to significantly increase risks in global stock, bond, crude oil and foreign exchange markets respectively in the short term. The COVID-19 pandemic also increases risks that have a domino impact on global financial markets in the medium and long term, and the magnitude of the risk is highly correlated with the level of investor panic in financial markets (Fang et al., 2023).
The COVID-19 pandemic is not the first economic crisis the world has faced. There have been various economic crises. One of the crises experienced by Indonesia occurred in 1998. In fact, one of the impacts of the economic crisis that occurred in Indonesia in 1998 was the high bank interest rates in Indonesia. The average interest margin in Indonesia after the 1998 crisis reached 6.36%, which is the highest figure compared to other Asian countries (Lin et al., 2012).
During times of crisis, companies usually have large working capital due to the large inventory which is not accompanied by any liabilities. In general, working capital consists of accounts receivable, inventory, and accounts payable. During the global financial crisis, working capital levels increased due to reasons such as unexpectedly excessive inventory levels, many delayed payments for goods and services resulting in large trade receivables, and a decline in company sales resulting in a decrease in accounts payable. (Tsuruta & Uchida, 2019). With the decline in company sales during the crisis, purchases of goods also decreased, causing a decrease in accounts payable (Tsuruta & Uchida, 2013).
There are also various possibilities for credit constraints for companies that cause companies to have to use other financial sources. Companies must adjust working capital levels to the company's specific targets during and after the crisis period (Tsuruta, 2019). One of the efforts that companies can make to get through the crisis is restructuring. In fact, the restructuring coefficient is one of the variables in the recovery equation with a significance of 0.0578 (Koh et al., 2015). The various bankruptcy prediction models that have previously been studied by previous researchers may not necessarily be able to be used generally in various types of economic environments. This is because these models are used to predict bankruptcy in developed countries (Oz & Simga-Mugan, 2018). Apart from that, to test the generality of these models it is necessary to carry out a re-estimation test (Grice & Dugan, 2003).
From previous studies, it can be seen how previous studies that examined bankruptcy prediction models tended to focus on developed economies so they did not provide an understanding of the impact of the pandemic on companies in developing countries. Therefore, this research aims to fill this gap by evaluating existing bankruptcy prediction models in the context of the Indonesian economy during the pandemic period. As a basis, this research will test prediction models that have been proven effective in advanced economies, namely Altman 1968, Ohlson 1980, and Springate 1978 with a special focus on the applicability of these models to companies in Indonesia. To support this analysis, accurate and up-to-date economic data from the Indonesian Central Bureau of Statistics will be used as the main source of information to describe the impact of the pandemic on the financial condition of companies in Indonesia.
Company failure in financial difficulties leading to bankruptcy is
usually the result of financial difficulties and economic difficulties. Company
financial difficulties often arise from cash flow deficiencies required to meet
the company's debt obligations. Meanwhile, even in conditions of economic
difficulty, companies often have a sustainable business model. In practice,
corporate difficulties that lead to bankruptcy are often a combination of both (Altman,
2006). Altman formulated the bankruptcy prediction model to take into account
developments in emerging markets. While there are not many changes to the
variables, Altman adds a new constant to the bankruptcy prediction model
equation for emerging markets. The equations are:
NWC/TA = Net Working
Capital to Total Asset
RE/TA = Retained
Earning to Total Asset
EBIT/TA = Earning
Before Interest and Tax to Total Asset
BVE//TL = Book Value of Equity to Total Liabilities
The next model used
besides the Altman model is the Ohlson model which uses nine independent
indicators as depicted in the following equation (Ohlson, 1980):
…(2.2)
SIZE = log (total assets / GNP price
level index)
TL/TA = total liabilitas / total aset
WC/TA = working capital / total assets
CL/CA = current liabilities / current
assets
ONEGG = 1 (if total liabilities > total
asset) ; 0 (if total liabilities < total asset)
NI/TA = net income / total assets
FU/TL = funds provided by operations /
total liabilities
INTWO = 1 (if net income is negative for the
last 2 years) ; 0 (if else)
CHINN
= 
The calculation
of the Ohlson model value will also be considered based on the cut-off which is
also the result of Ohlson's research. If the calculated value exceeds 3.8%, or
0.038, it can indicate that the company is likely to experience bankruptcy
(Ohlson, 1980). From the results of the calculation of the Ohlson value, the
probability of bankruptcy of the company concerned can also be obtained using
the following logistic method (Utama & Lumondang, 2009):
…(2.3)
The next
bankruptcy prediction model is the model proposed by Gorgon L.V Springate in
1978, namely the Springate Model. This model was tested on 40 manufacturing
companies in Canada. The results of this modeling were that 20 of the 40
companies were predicted to experience bankruptcy with an accuracy rate of
92.5% (Ghodrati, 2012). The equation of the Springate Model uses 4 ratios as
follows:
…(2.4)
Where X1 is working capital / total assets, X2 is EBIT / total assets, X3 is EBT / current liabilities, and X4 is sales / total assets. The interpretation of the S-score value is to look at the value of the S-score of each company, where companies with an S-score value of more than 0.862 are predicted not to experience bankruptcy or be in good health, while on the contrary, companies with an S-score value below 0.862 are predicted to experience bankruptcy.
Multi
Discriminant Analysis (MDA) is a statistical technique as well as a branch of
discriminant analysis used in finance and investment to evaluate the potential
of various investments when the number of variables used is large. MDA is used
to group variables by reducing differences between variables. In the Altman
Model itself, for example, MDA helps identify variables that are considered to
be the most important variables in the equation (Ishmah et al., 2022). Multi
Discriminant Analysis is used in the Altman Model and Springate Model so that
in this research it is used to re-estimate the coefficients of the variables
from the two equations.
The next algorithm is Logit Regression Model, which is similar to Multi Discriminant Analysis, is also used for classification as well as being a tool for predictive analytics. The Logit Regression Model is used to estimate the possibility or probability of an event based on available data with the dependent variable expressed in binary form, namely 0 and 1. From this probability, an evaluation is then carried out on how well the model predicts the dependent variable. Logit Regression Analysis is used in the Ohlson Model so that in this research it is used to re-estimate the coefficients of the variables from the Ohlson Model equation itself. Moreover, this research was conducted to find out how suitable the existing bankruptcy prediction models that have been used in other countries to be used during the crisis in Indonesia
Research
Methods
This research was conducted using secondary data in the
form of financial reports from 60 companies in Indonesia which were recorded
from the closest period to the pre-crisis period, data recorded during the
crisis period itself, and data recorded for several years after the global
COVID pandemic crisis. -19, namely in 2017, 2018, 2019, 2020, 2021 and 2022.
This was done to see the effects of the global financial crisis which was the
effect of the COVID-19 pandemic on the financial condition of these companies.
The data is downloaded from Revinitif Eikon and must
contain the variables needed in the five bankruptcy prediction models used in
this research. In this research, there are 7 stages starting from data
collection, sorting data according to research criteria as well as outlier data
based on confidence levels, predicting company bankruptcy using the models that
have been chosen for this research which is also continued with re-estimating
the coefficients used. on these models, validity testing, accuracy testing and
error type analysis, and ending with drawing conclusions. These stages are
summarized in the research workflow below:
Figure 1. Research Flow Diagram
The data processed are financial reports of companies in
Indonesia registered on the Indonesian Stock Exchange that experienced
financial distress when hit by the pandemic in 2020. The financial reports used
as data are financial reports for 2016, 2017, 2018, 2019, 2020, 2021, 2022 and
2023 where financial reports in 2016, 2017, 2018, 2019 act as pre-crisis data,
financial reports in 2020 and 2021 as data during the crisis, and data in 2021
as post-crisis data. This was done to see how significant the changes
experienced by these companies were due to the crisis. The data taken must
contain the variables needed in the five bankruptcy prediction models used in
this research. The data used in the research was taken from Revinitif Eikon.
Once collected, the data is then sorted according to the
required criteria and outliers are eliminated. The criteria used as the basis
for sample selection consist of four criteria. First, company year data must
come from the fiscal year-end financial statements. Second, all firm-year
observations must include every variable required by the models tested in this
study. Third, company year data points must be continuous for the entire
observation period. Fourth, outliers are removed at a confidence level of 95%
to increase the robustness of the research results (Oz & Simga-Mugan,
2018).
After all the required variables have been collected,
financial distress indicators are calculated using the bankruptcy prediction
models selected in this research, namely the Altman Z Score, Beneish M Score,
Ohlson Score and Springate Score. The Altman model used in this research is the
Altman model for emerging markets, namely a model for calculating financial
distress indicators specifically for developing countries.
This calculation was carried out on data from the
pre-crisis year, the crisis period itself, and the post-crisis period from year
to year with the aim of seeing differences in the financial condition of
companies before and after the crisis and then comparing them. From the
comparison results, companies that experienced significant and insignificant
changes in financial conditions will also be analyzed.
The next step is to carry
out accuracy tests and Type I and Type II error analysis. Test errors include
two types of errors, namely Type I and Type II errors. Type I error is an error
condition where the calculation classifies a company experiencing financial
difficulties as a healthy company, while Type II error is an error where the
calculation results classify a healthy company as experiencing financial
difficulties). Type 1 errors are riskier than Type 2 errors because errors in
identifying companies that are actually experiencing financial difficulties as
healthy can have serious consequences, whereas Type 2 errors, although important,
are not as fatal as I errors.
Conclusions were drawn based on the research objectives,
namely finding out whether the Altman Z Score, Beneish M Score, Springate
Score, Ohlson O Score bankruptcy prediction models can be used to detect
bankruptcy during the crisis due to the pandemic in Indonesia, as well as
knowing the best bankruptcy prediction models. good for use in times of crisis
due to the pandemic in Indonesia.
Results and Discussion
The first stage carried out
was to sort the data to see the continuity and completeness of the data by
eliminating companies that were detected as having incomplete data. This is
done so that data distribution can be obtained properly at the next stage. In
addition, sorting was carried out to ensure that all companies processed had
all the variables needed in the research. From the results of this sorting, the
remaining number of companies was 185 companies, of which 11 companies in the
automotive sector were eliminated, 24 companies in the energy sector were
eliminated, 48 companies in the manufacturing sector were eliminated, 20 companies
in the health sector were eliminated, 50 companies in the food sector were
eliminated, and 42 companies in the real estate sector eliminated.
After eliminating companies
that did not have data continuity and completeness, the next thing to do was to
carry out descriptive statistics on the data for the 185 companies. Data that
is eliminated is data with a Z value below -3 and above 3 with reference to a
confidence level of 95% (Venkataanusha, et al., 2019). Z itself is a
standardization score used in statistics to measure the extent to which a value
in a data distribution differs from the average in standard deviation units. This
process is carried out using SPSS software. From the elimination of outliers,
there were 45 companies eliminated in the combined data of all sectors. The
coefficients resulting from the reestimation are as follows:
Table 1. Comparison of original coefficients (β) and re-estimated coefficients from
the Altman EM (Emerging Market), Ohlson and Springate models for combined data
for all sectors (
), automotive sector
(1),
energy sector (2),
industrial manufacture sector (3),
health sector (4),
food and beverage sector (5),
real estate sector (6)
|
X |
β |
|
|
|
|
|
||
|
6.52 |
*0.494 |
*5.906 |
*2.544 |
*-0.762 |
*4.806 |
*1.175 |
*-0.606 |
|
|
RE/TA |
3.26 |
*1.11 |
*1.405 |
*-0.045 |
*0.709 |
*0.428 |
*1.559 |
*1.201 |
|
OI/TA |
6.72 |
*16.782 |
*-3.129 |
*10.612 |
*19.173 |
*10.705 |
*20.199 |
*26.424 |
|
BE/TL |
1.05 |
*-0.087 |
*-0.008 |
*0.077 |
*-0.77 |
*-0.095 |
*-0.224 |
*-0.008 |
|
(Constant) |
3.25 |
*-1.111 |
*-0.773 |
-1 |
*-0.867 |
*-2.518 |
*-1.441 |
*-1.14 |
|
Ohlson |
||||||||
|
X |
β |
|
|
|
|
|
|
|
|
SIZE |
-0.407 |
-0.258 |
-2.921 |
-0.655 |
-0.725 |
6.245 |
0.648 |
-0.052 |
|
TLTA |
6.03 |
-0.029 |
21.095 |
3.879 |
0.703 |
-108.157 |
*-12.785 |
0.81 |
|
WCTA |
-1.43 |
0.222 |
-54.935 |
-3.101 |
-3.157 |
22.048 |
3.759 |
0.027 |
|
CLCA |
0.076 |
0.313 |
-20.117 |
-0.362 |
-3.427 |
126.12 |
4.111 |
-0.453 |
|
NITA |
-2.37 |
*-55.739 |
-104.085 |
*-26.486 |
*-50.293 |
120.846 |
*-101.11 |
*-72.011 |
|
OCFTL |
-1.83 |
-0.007 |
-0.041 |
*0.050 |
-0.005 |
-0.067 |
-0.086 |
-0.007 |
|
OINEG |
0.285 |
|||||||
|
CHIN |
-1.72 |
*-0.457 |
-0.281 |
-1.251 |
-0.708 |
-28.003 |
0.213 |
-0.391 |
|
INTWO |
-0.521 |
*1.048 |
-4.327 |
1.736 |
2.157 |
20.046 |
0.575 |
|
|
(Constant) |
-1.32 |
*-1.123 |
8.207 |
-4.28 |
1.015 |
-51.288 |
2.861 |
-0.934 |
|
Springate |
||||||||
|
X |
β |
|
|
|
|
|
|
|
|
WCTA |
1.03 |
*0.417 |
*6.602 |
*1.295 |
*-1.672 |
*5.281 |
*-0.794 |
*-0.637 |
|
EBITTA |
3.07 |
*13.022 |
*-2.066 |
*3.099 |
*9.380 |
*8.503 |
*18.798 |
*18.764 |
|
EBTCL |
0.66 |
*0.713 |
*-0.459 |
*1.015 |
*2.967 |
*-0.422 |
*0.404 |
*1.202 |
|
SALESTA |
0.4 |
*0.252 |
*0.930 |
*1.396 |
*0.333 |
*0.724 |
*0.273 |
*1.884 |
|
(Constant) |
*-1.181 |
*-1.143 |
*-1.306 |
*-0.702 |
*-3.616 |
*-1.783 |
*-1.104 |
|
*Represent Statistical significance at 5%
After the entire model was recalculated using the original
coefficients and re-estimations, interpretation results were obtained for
companies experiencing financial distress (0) and healthy companies (1). After
seeing the differences in results between models and re-estimating each model,
the next step is to test for type I and II errors. Type I error is an error
condition where the calculation results classify a company experiencing
financial difficulties as a healthy company, while Type II error is an error
where the calculation results classify a healthy company as a company
experiencing financial difficulties. The benchmark for the actual condition of
a company is a situation where a company experiencing financial distress is a
company that has a negative net profit for two consecutive years.
Type I errors are riskier than Type II errors because
errors in identifying companies that are truly experiencing financial
difficulties as healthy companies can have serious consequences, while Type 2
errors, although important, are not as serious as error I. And besides type
error tests, they are also carried out. accuracy test on each model in each
sector. Accuracy testing is important in evaluating the performance of a
predictive model or system after conducting type I and type II error testing.
Accuracy tests provide a comprehensive picture of the
extent to which a model can make correct and useful predictions. While type I
error and type II error testing provide insight into specific errors, accuracy
testing provides a general idea of the model's ability to make correct and
accurate predictions. Apart from that, accuracy tests also present data that is
easier to understand.
Apart from the overall accuracy test, the author also
carried out specific accuracy tests, namely accuracy tests carried out on each
model that succeeded in predicting non-financial distress and succeeded in
predicting financial distress. This was done with the aim of seeing more
clearly the accuracy of each model from a closer perspective, namely predicting
non-financial distress and financial distress respectively. Then, to facilitate
the analysis, the author carried out a comparison of each analysis tool, namely
the accuracy value, both total accuracy and specific accuracy, with type I
error and type II error. This is to be able to see how accuracy values and
error values can build or bring down each other as the model chosen to analyze
bankruptcy predictions during the COVID-19 crisis.
In the combined data of all sectors (table 2), the highest
error value was obtained from the original Altman EM model, which was 99% of
the type I error. As discussed previously, the type I error is a more dangerous
type of error than the type II error because it predicts a firm year that experiencing
financial distress as a healthy firm year. Therefore, for combined data for all
sectors, the original Altman EM model is not recommended. From a type I error
point of view, the most recommended model is the original Springate model
because it has the lowest type I error value, namely only 5%.
Meanwhile, from a specific accuracy perspective, financial
distress accuracy values are more important than non-financial distress
accuracy values. This is because financial distress accuracy describes how accurate
the model is in predicting the number of firm years with financial distress
conditions, which is different from non-financial distress accuracy which
describes how accurate the model is in predicting firm years with non-financial
distress conditions.
From the highest financial distress accuracy value, the
original Springate model is also the recommended model for analyzing bankruptcy
predictions during the COVID-19 crisis for all sectors as a whole. However,
this model has a specific accuracy value for firm years experiencing
non-financial distress of only 54%, while the highest specific accuracy value
for non-financial distress comes from the original Altman EM model. The highest
total accuracy was obtained from the Ohlson re-estimation model, although the
accuracy value for financial distress from this model was only 61%.
|
Error
Type I |
Error
Type II |
FD
Accuracy |
Non-FD
Accuracy |
Total
Accuracy |
Chi-square |
Prob |
|
||
|
Altman EM Ori |
99% |
0% |
1% |
100% |
75% |
|
|
||
|
Altman EM RE |
57% |
3% |
43% |
97% |
83% |
*332.22 |
0 |
||
|
Ohlson Ori |
45% |
21% |
55% |
79% |
73% |
|
|
||
|
Ohlson RE |
39% |
4% |
61% |
96% |
87% |
*21.49 |
0.006 |
||
|
Springate Ori |
5% |
46% |
95% |
54% |
65% |
|
|
||
|
Springate RE |
54% |
2% |
46% |
98% |
85% |
*337.34 |
0 |
*Represent
Statistical significance at 5%
Table 3. Summary of type I and II error
classification, FD and non-FD accuracy and total accuracy for automotive
sectors
|
Model |
Error Type I |
Error Type II |
FD Accuracy |
Non-FD Accuracy |
Total Accuracy |
Chi-square |
Prob |
|
|
|
Altman
EM Ori |
100% |
0% |
0% |
100% |
62% |
|
|
||
|
Altman
EM RE |
31% |
8% |
69% |
92% |
83% |
*17.440 |
0.002 |
||
|
Ohlson
Ori |
50% |
8% |
50% |
92% |
76% |
|
|
||
|
Ohlson
RE |
100% |
0% |
0% |
100% |
62% |
3.430 |
0.904 |
||
|
Springate
Ori |
0% |
58% |
100% |
42% |
64% |
|
|
||
|
Springate
RE |
38% |
12% |
63% |
88% |
79% |
*17.836 |
0.001 |
*Represent
Statistical significance at 5%
From the automotive sector
(table 3), when analyzing
bankruptcy prediction in the context of the COVID-19 crisis for the automotive
sector, caution should be exercised in relying on the original Altman EM value
and the re-estimated Ohlson model, as they exhibit the highest type I error
values. Conversely, the original Springate model stands out with the lowest
type I error value and a balanced 100% specific accuracy for financial
distress. Although the Altman EM re-estimation model achieves the highest total
accuracy at 83%, it is important to note that its specific accuracy for
financial distress is moderate at 69%. Therefore, considering both type I error
and specific accuracy, the original Springate model emerges as a more robust
choice for bankruptcy prediction in the challenging circumstances brought about
by the COVID-19 crisis in the automotive sector.
In the energy sector (table 4), despite the original Ohlson
model exhibiting the highest type I error value at 91% and a very low specific
accuracy of 9% for financial distress in the energy sector, the original
Springate model emerges as the recommended choice due to its lowest type I
error value and a relatively high specific accuracy of 97% for financial
distress in the context of bankruptcy prediction during the COVID-19 crisis. It
is noteworthy, however, that the Altman EM re-estimation model, despite
achieving the highest total accuracy at 81%, should be approached with caution,
as its specific accuracy for financial distress is only 44%. Therefore, while
the original Springate model offers a more balanced performance, the Altman EM
re-estimation model's higher total accuracy should be considered alongside its
lower specific accuracy for financial distress in the decision-making process
for bankruptcy prediction analysis in the energy sector during the COVID-19
crisis.
Table 4. Summary of type I and II error
classification, FD and non-FD accuracy and total accuracy for energy sectors
|
Model |
Error Type I |
Error Type II |
FD Accuracy |
Non-FD Accuracy |
Total Accuracy |
Chi-square |
Prob |
|
|
|
Altman
EM Ori |
65% |
5% |
35% |
95% |
80% |
|
|
||
|
Altman
EM RE |
56% |
6% |
44% |
94% |
81% |
*42.898 |
0.000 |
||
|
Ohlson
Ori |
91% |
12% |
9% |
88% |
67% |
|
|
||
|
Ohlson
RE |
3% |
40% |
97% |
60% |
70% |
2.170 |
0.976 |
||
|
Springate
Ori |
3% |
50% |
97% |
50% |
62% |
|
|
||
|
Springate
RE |
47% |
3% |
53% |
97% |
86% |
*53.495 |
0.000 |
*Represent Statistical significance at 5%
Table 5. Summary of type I and II error
classification, FD and non-FD accuracy and total accuracy for industrial
manufacture sectors
|
Model |
Error Type I |
Error Type II |
FD Accuracy |
Non-FD Accuracy |
Total Accuracy |
Chi-square |
Prob |
|
|
|
Altman
EM Ori |
100% |
0% |
0% |
100% |
77% |
|
|
||
|
Altman
EM RE |
34% |
27% |
66% |
73% |
71% |
*51.838 |
0.000 |
||
|
Ohlson
Ori |
49% |
29% |
51% |
71% |
67% |
|
|
||
|
Ohlson
RE |
37% |
5% |
63% |
95% |
87% |
*7.649 |
0.469 |
||
|
Springate
Ori |
9% |
53% |
91% |
47% |
57% |
|
|
||
|
Springate
RE |
63% |
2% |
37% |
98% |
84% |
*57.113 |
0.000 |
*Represent Statistical
significance at 5%
If you look at table 5 from
the perspective of the highest type I error value, then in the manufacturing
industry sector the original Altman EM model is the least recommended model
because the type I error value of this model reaches 100%, accompanied by a
specific financial distress accuracy value of 0%. The lowest type I error value
was obtained from the original Springate model, namely only 9%, accompanied by
a specific financial distress accuracy value of 91%. Therefore, the original
Springate model is the most recommended model for analyzing bankruptcy
predictions during the COVID-19 crisis for the manufacturing industrial sector.
The highest total accuracy was obtained from the Ohlson re-estimation model,
namely 87% even though the specific accuracy value for financial distress was
medium, namely only 63%.
In the health sector (Table 6), almost all models obtained
the highest type I error value, namely 100%. These models are the original
Altman EM model, the original Ohlson model, and the re-estimated Ohlson model.
These three models also have a low specific financial distress accuracy value,
namely 0%. With the type I error value being very high and the financial
distress accuracy value being very low, these three models are not recommended
for use in bankruptcy prediction analysis during the COVID-19 crisis for the
health sector. The recommended model is the original Springate model, with the
lowest type I error value, namely only 0%, with the highest specific accuracy value
for financial distress, namely 100%. If the analysis is carried out from a
total accuracy value perspective, then the model with the highest total
accuracy value is the Springate re-estimation model, which is 98%, even though
the specific accuracy value for financial distress is medium, which is only
50%.
Table 6. Summary of type I and II error
classification, FD and non-FD accuracy and total accuracy for health sectors
|
Model |
Error Type I |
Error Type II |
FD Accuracy |
Non-FD Accuracy |
Total Accuracy |
Chi-square |
Prob |
|
|
|
Altman
EM Ori |
100% |
0% |
0% |
100% |
96% |
|
|
||
|
Altman
EM RE |
50% |
2% |
50% |
98% |
96% |
*9.514 |
0.049 |
||
|
Ohlson
Ori |
100% |
13% |
0% |
87% |
83% |
|
|
||
|
Ohlson
RE |
100% |
0% |
0% |
100% |
96% |
0.000 |
1.000 |
||
|
Springate
Ori |
0% |
8% |
100% |
92% |
93% |
|
|
||
|
Springate
RE |
50% |
0% |
50% |
100% |
98% |
*9.90 |
0.042 |
*Represent Statistical significance at 5%
Table 7. Summary of type I and II error
classification, FD and non-FD accuracy and total accuracy for food and beverage
sectors
|
Model |
Error Type I |
Error Type II |
FD Accuracy |
Non-FD Accuracy |
Total Accuracy |
Chi-square |
Prob |
|
|
|
Altman
EM Ori |
100% |
0% |
0% |
100% |
82% |
|
|
||
|
Altman
EM RE |
59% |
1% |
41% |
99% |
89% |
*145.947 |
0.000 |
||
|
Ohlson
Ori |
46% |
10% |
54% |
90% |
83% |
|
|
||
|
Ohlson
RE |
81% |
0% |
19% |
100% |
86% |
0.674 |
1.000 |
||
|
Springate
Ori |
16% |
25% |
84% |
75% |
77% |
|
|
||
|
Springate
RE |
54% |
0% |
46% |
100% |
90% |
*109.223 |
0.000 |
*Represent Statistical significance at 5%
In assessing bankruptcy prediction during the COVID-19
crisis in the food sector (Table 7), it is crucial to note that the original
Altman EM model presents the highest type I error value at 100%, coupled with
an extremely low specific accuracy for financial distress at 0%. In contrast,
the original Springate model stands out with the lowest type I error value,
merely 16%, and concurrently boasts the highest specific accuracy for financial
distress at 84%. Consequently, the recommended model for robust bankruptcy
prediction analysis in the food sector during the COVID-19 crisis is the
original Springate model. Despite the Springate re-estimation model exhibiting
the highest total accuracy, reaching 46% for specific financial distress, its
lower specific accuracy compared to the original Springate model underscores
the latter's superior performance in this context.
In the real estate sector (Table 8), careful consideration
is crucial when selecting a bankruptcy prediction model for the COVID-19
crisis. The original Altman EM model, with the highest type I error value and a
minimal 0% specific accuracy for financial distress, is strongly discouraged
for use in this context. On the other hand, the original Springate model
emerges as the recommended choice, displaying an impressively low type I error
rate of 2% and a substantial specific accuracy for financial distress at 98%.
While the Altman EM re-estimation model attains the highest total accuracy
value at 83%, it is essential to note that its specific accuracy for financial
distress stands at 50%. Therefore, in the real estate sector during the
COVID-19 crisis, the original Springate model is preferred for its balanced
performance, striking a commendable trade-off between type I error and specific
accuracy for financial distress.
Table 8. Summary of type I and II error
classification, FD and non-FD accuracy and total accuracy for real estate
sectors
|
Model |
Error Type I |
Error Type II |
FD Accuracy |
Non-FD Accuracy |
Total Accuracy |
Chi-square |
Prob |
|
|
|
Altman
EM Ori |
100% |
0% |
0% |
100% |
75% |
|
|
||
|
Altman
EM RE |
50% |
6% |
50% |
94% |
83% |
*54.409 |
0 |
||
|
Ohlson
Ori |
39% |
41% |
61% |
59% |
60% |
|
|
||
|
Ohlson
RE |
23% |
6% |
77% |
94% |
90% |
14 |
0.072 |
||
|
Springate
Ori |
2% |
75% |
98% |
25% |
44% |
|
|
||
|
Springate
RE |
50% |
2% |
50% |
98% |
86% |
*70.112 |
0 |
*Represent Statistical significance at 5%
Conclusion
In the research conducted, it can be seen how the three models, namely Altman EM, Ohlson and Springate, show different responses to each sector that is the object of research, including when all sectors are combined into one large data set. From the 7 datasets that are the object of research, including when all sectors are combined into one large data set, the original Springate model always produces the lowest type I error value as well as the highest accuracy value in predicting firm years experiencing financial distress. The original Altman EM model always produces the highest type I error value, except in the energy sector, as well as the lowest accuracy value in predicting firm years experiencing financial distress. The original Altman EM, Ohlson re-estimation and Springate re-estimation models always produce the highest type II error values as well as the lowest accuracy values in predicting a healthy firm year. The Ohlson model is superior on 3 of the 7 datasets in obtaining the highest total accuracy value, while the Springate re-estimation model is also superior on 3 of the t datasets in obtaining the highest total accuracy value. The Altman EM re-estimation model was only superior in 1 of 7 data sets in obtaining the highest total accuracy value. Apart from that, it can be seen that the original Springate model is the model that produces the lowest type I error value so that this model is the most recommended model for predicting bankruptcy during the crisis due to the pandemic in Indonesia.
BIBLIOGRAPHY
Altman, E., & Hotchkiss, E. (2006). Corporate
financial distress and bankruptcy. Hoboken, N.J.: Wiley.
Fang, Y., Shao, Z., & Zhao, Y. (2023).
Risk spillovers in global financial markets: Evidence from the COVID-19 crisis.
International Review of Economics & Finance, 83, 821–840.
https://doi.org/10.1016/j.iref.2022.10.016
Ghodrati, Hasan (2012). A Study of the
Accuracy of Bankruptcy Prediction Models: Altman, Shirata, Ohlson, Zmijewsky,
CA Score, Fulmer, Springate, Farajzadeh Genetic, and McKee Genetic Models for
the Companies of the Stock Exchange of Tehran. American Journal of Scientific
Research, Issue 59 (2012), pp. 55-67
Grice, J. S., & Dugan, M. T. (2003).
Re-estimations of the Zmijewski and Ohlson bank- ruptcy prediction models.
Advances in Accounting, 20, 77–93.
Ishmah, H., Solimun, & Mitakda, M.
B. (2022). Multiple discriminant analysis Altman Z-score, multiple discriminant
analysis stepwise and K-means cluster for classification of financial distress
status in manufacturing companies listed on the Indonesia Stock Exchange in
2019. Advances in Computer Science Research. https://doi.org/10.2991/acsr.k.220202.035
Koh, S. K., Durand, R. B., Dai, L., &
Chang, M. (2015). Financial distress: Lifecycle and corporate restructuring.
Journal of Corporate Finance, 33, 19–33.
https://doi.org/10.1016/j.jcorpfin.2015.04.004
Lin, J.-R., Chung, H., Hsieh, M.-H., &
Wu, S. (2012). The determinants of interest margins and their effect on bank
diversification: Evidence from Asian Banks. Journal of Financial Stability,
8(2), 96–106. https://doi.org/10.1016/j.jfs.2011.08.001
Oz, I. O., & Simga-Mugan, C. (2018).
Bankruptcy prediction models’ generalizability: Evidence from emerging market
economies. Advances in Accounting, 41, 114–125.
https://doi.org/10.1016/j.adiac.2018.02.002
Tsuruta, D., & Uchida, H. (2019). The
real driver of trade credit. Pacific-Basin Finance Journal, 57, 101183.
https://doi.org/10.1016/j.pacfin.2019.101183
Tsuruta, D. (2019). Working Capital
Management during the Global Financial Crisis: Evidence from Japan. Japan and
the World Economy, 49, 206–219. https://doi.org/10.1016/j.japwor.2019.01.002
Utama, C. A., & Lumondang, A. (2009).
Pengaruh bankruptcy risk, size dan book-to-market perusahaan terhadap imbal
hasil. Jurnal akuntansi dan keuangan Indonesia, 6(2), 2.
Venkataanusha, P. et al. (2019) ‘Detecting
outliers in high dimensional data sets using Z-score methodology’,
International Journal of Innovative Technology and Exploring Engineering, 9(1),
pp. 48–53. doi:10.35940/ijitee.a3910.119119.
|
Copyright holder: Mentari M. Lubis, Imo Gandakusuma (2024) |
|
First publication right: Syntax Literate: Jurnal Ilmiah Indonesia |
|
This article is licensed under:
|
