Download presentation

Presentation is loading. Please wait.

Published byDaniella Dancer Modified over 2 years ago

1
Dr. Mo Vaziri, California State University San Bernardino, USA Dr. Rafiqul Bhuyan, California State University San Bernardino, USA Ponkala Anand Vaseekhar Manuel, California State University San Bernardino, USA Comparative Predictability of Failure of Financial Institutions Using Multiple Models

2
Introduction The impact of failure of financial institutions is beyond just the failure of a public corporation. The failure of financial institutions in the USA, is a clear evidence that the greater macro impact is beyond just the failure of few financial institutions. It can bring down the entire economy and can have global devastating impact. By realizing the grave systemic risk of the failure, US government is forced to intervene and bail out many institutions for greater macro-economic reasons. It raises the view that perhaps the current regulating policies and methods are lacking efficiency in predicting the possibility of failure ahead of time and hence not effective in preventing that to happen

3
Purpose of Research The objective of this paper is to investigate four different sets of variables that have been employed including, by two rating agencies, Moody (10 accounting ratios) and Standard and Poor (8 accounting ratios), one by Altman (4 accounting rations for non- manufacturing firms) and the one proposed by -Vaziri (17 accounting ratios). More than 200 banks over the period of 2000 to 2010 are considered in this study. Several methodologies including discriminant and logistic analyses are applied on each set of variables to develop four models one corresponding to each set. The models are then used to perform the in-sample and out-sample prediction of bank financial distress or failure. The attempt will be made to check which model will provide a higher accuracy of prediction and earlier warning to the potential banking crisis. In this paper, it also summarized how the stress testing could be used in both banking sector and insurance sector in practice. Furthermore, ten defaulted banks have been selected for the construction of three stress testing models (Moodys System, S&P System, and Vaziris Model) with the further statistic analysis of the testing result

4
Survey of the Literature. McAleer, Angel, Martin, and Amaral have discussed about choosing sensible model to calculate market risk. They also discuss about choosing optimal market risk models and combining alternative risk models. Volatility models like GARCH (Engle, 1982), GJR (Glosten, Jagannathan and Runkle, 1992) and EGARCH (Nelson, 1991) are used to calculate strategic market risk as these models are popular in estimating stochastic volatility. Also method like EWMA (Riskmetric, 1996; Zumbauch, 2007) is used in a unified framework and notation.

5
Survey of the Literature Cont. Beltratti, and Stulz (2009) discussed about the reasons behind the poor performance of banks during the financial crisis López-Iturriaga, López-de-Foronda, and Sanz(2007)used neural network model to predict bankruptcy of U.S. Banks before the occurrence of bankruptcy. Forty-one variables were extracted from the banks the went bankrupt between January 2010 to June Sample size of 82 default banks and 196 non-default banks were used. Results from the test shows that neural network model was able to predict 60% of bankruptcy six months before it occurred. Montgomery, Hanh, Santoso, and Besar (2005) used logistic regression analysis to find the usefulness of domestic bank failure prediction model and cross-country model using cross-correlation to study the reasons for failure of banks in Japan and Indonesia. Total of 17 financial variable were collected from banks income statement for period of which includes state-owned banks, private national foreign exchange and non-foreign exchange banks, regional development banks, joint-venture banks, and foreign banks from Indonesia and city banks, long-term credit banks, trust banks and regional I and II banks for fiscal years from Japan.

6
Survey of the Literature Cont. Kalkbrener and Jan Willing (2008) discusse the three-factor model that includes market rates, deposit rate and deposit volumes. they also includes liquidity risk and interest rate risk management for non-maturing liabilities Rob Jameson (2007) discusses how banks suffered from a liquidity crisis. In his article he argues how some of the current ratios and management strategies are irrelevant to the current market situation. Some strategies just assume that the market prices are going to behave as they would in the past

7
The Data and Methodology We chose 4 following methodologies to investigate financial distress and bankruptcy which are Moodys, Standard and Poors, Vaziris and Z-score Model Moodys system provides Coverage Interest Ratio (EBIT/Interest Expense, EBITDA/ Interest Expense) Coverage Asset Ratio, Leverage Ratios (Total Liability/Total Asset, Equity/Total Liability, Short Term debt/Equity Book Value) Liquidity Ratios (Current Asset/Total Liability, Intangible Asset/Total Asset, Cash/Net Sale, Working Capital/Total Asset, Cash/ Total Asset) Profitability Ratio (ROA)

8
The Data and Methodology Cont. Standard & Poors Financial Ratios S& P uses credit risk analysis to find the credit rating to analyze risk and provide credit rating. S&Ps system includes Coverage ratios (EBIT/ Interest Expense, EBITDA/Interest Expense, Net Operating Income/Total Debt) Profitability Ratios (ROE, Net Operating Income/Sale

9
The Data and Methodology Cont. Z-Score Model Developed by Edward I. Altman, expert on corporate bankruptcy, high yield bonds, distressed debt and credit risk analysis. The Model measures financial health of a company and can be used as a bankruptcy predictor Z-Model (Estimate for Private Firms) Z=0.71X X X X4*+0.998X5

10
The Data and Methodology Cont. Vaziri Model: 1. Vaziri Model Vaziris system includes all most all ratios in Moodys and S&P system and other financial ratios like efficiency ratio which could be used to determine if the bank is stable or at high risk. Vaziris Model identifies how heavily the banks are leveraged and if they can pay of their debt. It also helps to identify if the banks are efficient and if they are able to make profit with the money they borrowed. Vaziris model helps to identify the reason for the bankruptcy of banks in long run. Ratios for this method includes: Coverage Ratios (Current Ratio, Quick Ratio, Cash Velocity, Time Interest Earn), Leverage ratios (Total Liability/Total Asset, Short Term Debt/Equity Book Value, Equity/Total Liability), Profitability Ratios (ROE, Net Income/ Sale, Retained Earnings/ Total Asset, EBIT/Total Asset, Net Income/Total Asset), and Efficiency Ratios ( Asset Turnover, Fixed Asset Turnover, Inventory Turnover, Inventory to Net Working Capital

11
Empirical Results Moodys Model Failed Banks correctly Predicted Non-Failed Banks Correctly Predicted Type I Error Type II Error Incorrectly Predicted in Total Correctly Predicted in Total % of Failed Banks Correctly Predicted % % % % % % % 45.00% % % of Non-Failed Banks Correctly Predicted % % % % % % % 58.75% % of Total Incorrectly Predicted % % % % % % % % 44.00% % % of Total Correctly Predicted % % % % % % % % 56.00% % p value

12
Empirical Results. S & P Model Failed Banks correctly Predicted Non-Failed Banks Correctly Predicted Type I Error Type II Error Incorrectly Predicted in Total Correctly Predicted in Total % of Failed Banks Correctly Predicted 5.00% % % % % % % % of Non-Failed Banks Correctly Predicted % % % % % % % % % % of Total Incorrectly Predicted % % % % % % % % % % of Total Correctly Predicted % % % % % % % % % p value

13
Empirical Results. Vaziris Model Failed Banks correctly Predicted Non-Failed Banks Correctly Predicted Type I Error Type II Error Incorrectly Predicted in Total Correctly Predicted in Total % of Failed Banks Correctly Predicted % % % % % % % % % % of Non-Failed Banks Correctly Predicted % % % % % % % % % % % of Total Incorrectly Predicted % % % % % % % % % of Total Correctly Predicted % % % % % % % % p value

14
Empirical Results. Logit Model Failed Banks correctly Predicted Non-Failed Banks Correctly Predicted Type I Error Type II Error Incorrectly Predicted in Total Correctly Predicted in Total % of Failed Banks Correctly Predicted % % % % % % % % % % % of Non-Failed Banks Correctly Predicted % % % % % % % % of Total Incorrectly Predicted % % % % % % % % % % of Total Correctly Predicted % % % % % % % % % p value

15
Empirical Results Z Score Model Failed Banks correctly Predicted Non-Failed Banks Correctly Predicted Type I Error Type II Error Incorrectly Predicted in Total Correctly Predicted in Total % of Failed Banks Correctly Predicted % % % % % % % % of Non-Failed Banks Correctly Predicted % % % % % % % % % % % of Total Incorrectly Predicted 5.00% % % % % % % % % % of Total Correctly Predicted % % % % % % % % % p value0.05

16
Conclusion Moodys model predicted that 11 banks would be bankrupt a year before the bank filled for bankruptcy and 10 banks before two years of bankruptcy. Of 20 banks that filled bankruptcy 18 of them filled in 2010 and 2 of them in Studies suggest that models can predict perfect bankruptcy two years before the file for it. The percentage of correct prediction ranges from 69% to 76%. This shows that this model is 72.5% reliable on average. This model also shows that most of the banks had high leverage and has less liquidity.

17
Conclusion S & P model, though shows that percentage of correct prediction is 80% it lacks to predicts the failure of banks in advance. The correct prediction of failed banks is only 5% in 2010 and 2009 and only 35% in 2008 and 2007, which shows that it is not reliable in predicting bankruptcy Vaziris model predicts much better than Moodys and S&P model. Percentage of correct prediction is almost 80% for all years and percentage of correctly predicted failed banks is 45% before two years of actual failure

18
Conclusion Percentage of correct prediction of Logit is almost same as Vaziris model. Both the models show that they are 50% reliable Of all the models Z Score gives the best prediction. Its prediction percentage of failed banks is 80% and shows 75% correct prediction before two years. Shareholders can relay more in this model

19

Similar presentations

© 2016 SlidePlayer.com Inc.

All rights reserved.

Ads by Google