Research Methodologies and Data Analysis for US Corporate Bankruptcy

This report outlines the research methodologies employed to predict corporate bankruptcy in the United States between 2008 and 2018. The study utilizes annual financial data acquired from the Bloomberg Terminal and SEC Edgar for 50 manufacturing companies, comprising 25 bankrupt and 25 non-bankrupt firms. The research employs a bootstrap test to enhance result accuracy given the sample size, alongside ANOVA and Multiple Discriminant Analysis. Ten financial ratios are calculated to evaluate operational efficiency, profitability, and debt levels. Multiple Discriminant Analysis is used to differentiate between bankrupt and healthy companies, assigning weights to variables to create a cut-off point for classification. The analysis aims to provide a robust framework for identifying potential corporate failures based on financial indicators.

Running head: RESEARCH METHODOLOGIES
RESEARCH METHODOLOGIES
Name of the Student
Name of the University
Author Note

1RESEARCH METHODOLOGIES
Chapter 3: Research Methodologies
3.1 Data
The data considered while preparing this research is the annual financial data of the
company for the year 2008-2018. The data is acquired from the Bloomberg Terminal and
SEC Edgar. Bloomberg Terminal is used because it is one of the best terminal of providing
real time financial data on every market (Wood et al. 2016). It has the best coverage of the
different securities and markets (Cong, Du and Vasarhelyi 2017). The SEC Edgar publishes
high value data on its online portal. A list of numerical data for the bankrupt companies in the
US as well as a few good companies of similar size has been acquired from the mentioned
portals. The data of about two years prior to the bankruptcy has also been considered. The
data are to be analysed in the Excel and the SPSS.
3.2 Sample of failed and non-failed companies
The research focuses on the period from 2008-2018. The period of 2008 underwent
the Global Financial Crisis. The sample has been acquired from the Bloomberg Terminal’s
Bankruptcy Dashboard. Only the data of 500+ sized companies are available on the
Bloomberg terminal and hence any other missing data were related to the smaller size
companies. 50 different manufacturing companies from the United States have been
considered for the data collection process, out of which 25 companies are bankrupted and
remaining 25 companies are healthy and profitable companies.
3.3 Bootstrap Test
The number of companies selected for the research paper is only 50 and hence the
bootstrap test has been used in order to maintain the accuracy of the result. The process of
bootstrapping assigns different measurements of accuracy to the estimates of the sample
(Hoang et al. 2017). The measurement of the accuracy can be defined in the terms of

2RESEARCH METHODOLOGIES
confidence levels, intervals, biasness, variances and prediction errors. It helps in measuring
the predictive power of the samples. The research considers the Confidence Interval Type of
‘Bias Corrected and Accelerated’ and a Confidence Interval level of 95%. The 95%
confidence level generates a trim of 5%, which is 2.5% from both the ends.
3.4 Group Statistics
The research paper has used several analytical tools while performing this research
paper. The statistical methods that has been used here Bootstrap test, ANOVA Test and
Multiple Discriminant Analysis. The financial measurement that has been used here is the
ratio analysis of the financial statement that has been obtained from the portal. The mean
score and the standard deviation has also been analysed. The bankrupt companies are grouped
as Group ‘0’ and the non-bankrupt companies are grouped as Group ‘1’. Ten financial ratios
have been calculated to measure the operational efficiency of the companies. The profitability
and the debt condition of the companies has also been measured in order to categorise it into
bankrupt and the non-bankrupt sectors.
3.5 Multiple Discriminant Analysis
Multiple Discriminant Analysis or the Canonical Discriminant Analysis helps in the
understanding of the different variations in between the bankrupt companies and the non-
bankrupt companies of the US by using the variable ratio means. In this case there are 25
bankrupt companies and 25 healthy companies. The analysis therefore maximally
discriminates between those bankrupt and the healthy companies. This analysis provides
certain coefficients after screening the variables and allocating certain weights to the
variables (Mihalovic 2016). Thereafter a cut-off point is identified and a score is allocated
based on which the grouping of the companies as bankrupt and non-bankrupt is being done.

3RESEARCH METHODOLOGIES
References:
Hoang, D.T., Chernomor, O., Von Haeseler, A., Minh, B.Q. and Vinh, L.S., 2017. UFBoot2:
improving the ultrafast bootstrap approximation. Molecular Biology and Evolution, 35(2),
pp.518-522.
Mihalovic, M., 2016. Performance comparison of multiple discriminant analysis and logit
models in bankruptcy prediction. Economics & Sociology, 9(4), p.101.
Wood, C., King, A., Catlow, R. and Scott, B., 2016. Terminal value: Building the alternative
Bloomberg. Finance and Society, 2(2), pp.138-50.
Cong, Y., Du, H. and Vasarhelyi, M.A., 2017. Are XBRL Files Being Accessed? Evidence
from the SEC EDGAR Log File Dataset. Journal of Information Systems, 32(3), pp.23-29.