Prediction on early warning signs of financial distress: machine learning approach

Budhidharma, Valentino (2023) Prediction on early warning signs of financial distress: machine learning approach. Doctoral thesis, Universitas Pelita Harapan.

[img] Text (Title)
Title.pdf
Restricted to Registered users only
Available under License Creative Commons Attribution Non-commercial Share Alike.

Download (68kB)
[img] Text (Abstract)
Abstract.pdf
Restricted to Registered users only
Available under License Creative Commons Attribution Non-commercial Share Alike.

Download (470kB)
[img] Text (ToC)
ToC.pdf
Restricted to Registered users only
Available under License Creative Commons Attribution Non-commercial Share Alike.

Download (1MB)
[img] Text (Chapter 1)
Chapter 1.pdf
Restricted to Registered users only
Available under License Creative Commons Attribution Non-commercial Share Alike.

Download (752kB)
[img] Text (Chapter 2)
Chapter 2.pdf
Restricted to Registered users only
Available under License Creative Commons Attribution Non-commercial Share Alike.

Download (638kB)
[img] Text (Chapter 3)
Chapter 3.pdf
Restricted to Registered users only
Available under License Creative Commons Attribution Non-commercial Share Alike.

Download (750kB)
[img] Text (Chapter 4)
Chapter 4.pdf
Restricted to Registered users only
Available under License Creative Commons Attribution Non-commercial Share Alike.

Download (522kB)
[img] Text (Chapter 5)
Chapter 5.pdf
Restricted to Registered users only
Available under License Creative Commons Attribution Non-commercial Share Alike.

Download (222kB)
[img] Text (Bibliography)
Bibliography.pdf
Restricted to Registered users only
Available under License Creative Commons Attribution Non-commercial Share Alike.

Download (280kB)
[img] Text (Appendices)
Appendices (1)_merged.pdf
Restricted to Repository staff only
Available under License Creative Commons Attribution Non-commercial Share Alike.

Download (5MB)

Abstract

The purpose of this study is to determine factors/variables that can differentiate the characteristics of distressed and good firms and propose a new model to explain financial distress in Indonesia. There have been many theories, variables, and estimation methods used by previous studies about early warning signs of financial distress. Determining factors of good and distressed firms uses Beaver's (1968) methodology. There are few studies on financial distress using a combination of theories, Random Forests (RF) as the machine learning algorithm, and Logit as the statistical method, especially in Indonesia. By using the RF, it is expected the study can get an improved combination of Classification and Regression Tree (CART) and Bagging. The samples used are most sectors in Indonesia’s Stock Exchange from 2005 to 2020, excluding the financial sector. The characteristic results show that good firms have higher NITA, GPTA, CTA, QATA, CATA, WCTA, CCL, RETA, and EBTCL, while distressed firms are better on CFTS, CFTA, CFNW, CFTD, NITS, NINW, NITD, ROE, CLTA, LTLTA, CLLTLTA, QACL, CR, CTS, ITS, CATS, WCTS, NWTS, and TATS. The proposed model results show that CTA, RETA, QATA, EBTCL, TLTA, TS, BVPS, and MB have a negative significant association with the probability of firms in distress. While CATA, QACL, TLMTA, TA, and INTEREST have a positive significant association with the probability of firms in distress. In conclusion, to avoid financial distress firms must have good selling while maintaining enough cash flow to fulfill their short-term liabilities. Firms must also keep on growing to become bigger so they can withstand more crises. This condition must be supported by a conducive interest rate. The additional result shows that combining theories, Random Forests, and Logit can be used to build a new financial distress prediction model. The comparison result with Altman’s (1968) model shows that the new model has a smaller RMSE than Altman’s. This is a new enlightenment since this method can be used to develop many new financial study models, not only using Logit estimates but also other estimation methods. / Tujuan dari penelitian ini adalah untuk mengetahui faktor/variabel yang dapat membedakan karakteristik perusahaan yang tertekan dan baik serta mengusulkan model baru untuk menjelaskan kesulitan keuangan di Indonesia. Ada banyak teori, variabel, dan metode estimasi yang digunakan oleh penelitian sebelumnya tentang tanda-tanda peringatan dini kesulitan keuangan. Menentukan faktor-faktor perusahaan yang baik dan tertekan menggunakan metodologi Beaver (1968). Ada beberapa penelitian tentang kesulitan keuangan menggunakan kombinasi teori, Random Forests (RF) sebagai algoritma pembelajaran mesin, dan Logit sebagai metode statistik, terutama di Indonesia. Dengan menggunakan RF, diharapkan penelitian dapat memperoleh peningkatan kombinasi Classification and Regression Tree (CART) dan Bagging. Sampel yang digunakan adalah sebagian besar sektor di Bursa Efek Indonesia dari tahun 2005 hingga 2020, tidak termasuk sektor keuangan. Hasil karakteristik menunjukkan bahwa perusahaan yang baik memiliki NITA, GPTA, CTA, QATA, CATA, WCTA, CCL, RETA, dan EBTCL yang lebih tinggi, sementara perusahaan yang tertekan lebih baik pada CFTS, CFTA, CFNW, CFTD, NITS, NINW, NITD, ROE, CLTA, LTLTA, CLLTLTA, QACL, CR, CTS, ITS, CATS, WCTS, NWTS, dan TATS. Hasil model yang diusulkan menunjukkan bahwa CTA, RETA, QATA, EBTCL, TLTA, TS, BVPS, dan MB memiliki hubungan signifikan negatif dengan probabilitas perusahaan dalam kesulitan. Sementara CATA, QACL, TLMTA, TA, dan INTEREST memiliki hubungan signifikan positif dengan probabilitas perusahaan dalam kesulitan. Kesimpulannya, untuk menghindari kesulitan keuangan, perusahaan harus memiliki penjualan yang baik sambil mempertahankan arus kas yang cukup untuk memenuhi kewajiban jangka pendek mereka. Perusahaan juga harus terus tumbuh untuk menjadi lebih besar sehingga mereka dapat menahan lebih banyak krisis. Kondisi ini harus didukung dengan suku bunga yang kondusif. Hasil tambahan menunjukkan bahwa menggabungkan teori, Random Forests, dan Logit dapat digunakan untuk membangun model prediksi kesulitan keuangan baru. Hasil perbandingan dengan model Altman (1968) menunjukkan bahwa model baru ini memiliki RMSE yang lebih kecil daripada Altman. Ini adalah pencerahan baru karena metode ini dapat digunakan untuk mengembangkan banyak model studi keuangan baru, tidak hanya menggunakan estimasi Logit tetapi juga metode estimasi lainnya.

Item Type: Thesis (Doctoral)
Creators:
CreatorsNIMEmail
Budhidharma, ValentinoNIM01617180018valentino.budhi@gmail.com
Contributors:
ContributionContributorsNIDN/NIDKEmail
Thesis advisorSembel, RoyNIDN0310076405roy.sembel@gmail.com
Thesis advisorUgut, Gracia Shinta S.NIDN0310026504gracia.ugut@uph.edu
Thesis advisorHulu, EdisonNIDN0315085902edison.hulu@lecturer.uph.edu
Uncontrolled Keywords: financial distress ; machine learning ; random forests ; logit
Subjects: H Social Sciences > HD Industries. Land use. Labor > HD28 Management. Industrial Management
Divisions: University Subject > Current > Faculty/School - UPH Karawaci > Business School > Doctor of Management
Current > Faculty/School - UPH Karawaci > Business School > Doctor of Management
Depositing User: Valentino Budhidharma
Date Deposited: 02 Oct 2023 06:12
Last Modified: 04 Oct 2023 08:58
URI: http://repository.uph.edu/id/eprint/58347

Actions (login required)

View Item View Item