finally final report

7/29/2019 Finally Final Report

1/46

1

Estimation of Loss Given Default for Indian Banks

By

NITIKA GUPTA, R1001025

Guide

Dr. Arindam Bandyopadhyay

Project Undertaken in TERM-VI AT NIBM

Report submitted in partial fulfillment of the requirements

For the award of

Post-Graduate Diploma in Banking and Finance

NATIONAL INSTITUTE OF BANK MANAGEMENT 2010-12


2/46

2

INDEXChapter No. Contents Page

No.

Chapter I Introduction 4

Chapter II Review of literature 17

Chapter III Description of the Data,

Variables and Summary

Statistics

22

Chapter IV Methodology 30

Chapter V Analysis/Results 32

Chapter VI Suggestions for Future

research

40

Chapter VII Executive Summary 42

Chapter VIII Appendices 44


3/46

3

ACKNOWLEDGEMENTS

I would like to thank Dr.Allen C Parrera Director National Institute of Bank Management and

Prof Kalyan Swarup, Dean National Institute of Bank Management for giving us an opportunity

to undertake a project in our term-VI.

I also take immense pleasure in thanking Dr. Arindam Bandyopadhyay, Faculty- Risk

Management ,National Institute of Bank Management for having permitted me to carry out

this project work under his supervision. Without his able guidance and encouragement it would

have been impossible to complete the project work, in time.

Last but not the least, I would thank the faculty and staff members of National Institute of Bank

Management, who have extended their helping hand by sharing their knowledge and

experiences.

Finally, yet importantly, I would like to express my heartfelt thanks to my beloved parents for

their blessings, my friends/classmates for their help and wishes for the successful completion of

this project. I hope that I have done justice to their expectations.

Date: 17th Feb, 2012

NITIKAGUPTA,

R1001025


4/46

4

Chapter I

Introduction

The etymology of the word Risk can be traced to the Latin word Rescum meaning Risk atSea or that which cuts. Risk is associated with uncertainty and reflected by way of charge on the

fundamental/ basic i.e. in the case of business it is the Capital, which is the cushion that protects

the liability holders of an institution. These risks are inter-dependent and events affecting one


5/46

5

area of risk can have ramifications and penetrations for a range of other categories of risks. Each

transaction that the bank undertakes changes the risk profile of the bank.

Business grows mainly by taking risk. Greater the risk, higher the profit and hence the business

unit must strike a tradeoff between the two. The essential functions of risk management are to

identify measure and more importantly monitor the profile of the bank. While Non-PerformingAssets are the legacy of the past in the present, Risk Management system is the pro-active action

in the present for the future. Managing risk is nothing but managing the change before the risk

manages. While new avenues for the bank has opened up they have brought with them new risksas well, which the banks will have to handle and overcome.

Moving towards the Basel II Framework, the RBI has adopted a three-track approach to capitaladequacy regulation in India, with the norms stipulated at varying degrees of stringency for

different categories of banks. Similar differentiated approach has been adopted in some other

jurisdictions also. This has been a deliberate choice for RBI having regard to the size, nature and

complexity of operations and relevance of different types of banks to the Indian financial sector,the need to achieve greater financial inclusion and to provide an efficient credit delivery

mechanism. Thus, the commercial banks, which account for the lions share in the total assets ofthe banking system, will be on Basel II standards while the co-operative banks will remain onBasel I norms for credit risk with surrogate measures for market risk. The Regional Rural Banks,

on the other hand, which have limited operations in rural areas, will be on non-Basel standards.

RBI has already issued the guidelines for the new capital adequacy framework in regard to Pillar

1 and Pillar 3 on April 27, 2007. Accordingly, the foreign banks operating in India and the Indian

banks having operational presence outside India were required to migrate to the StandardizedApproach for credit risk and the Basic Indicator Approach for operational risk with effect from

March 31, 2008. All other Scheduled commercial banks were encouraged to migrate to these

approaches under Basel II in alignment with them, but, in any case, not later than March 31,

2009. It has been a conscious decision to begin with the simpler approaches available under theframework, having regard to the preparedness of the banking system. As regards the market risk,

under Basel II also, the banks will continue to follow the Standardized-Duration Method as

already adopted under the Basel I framework. For migration to the advanced approachesavailable under the framework, prior approval of the RBI would be required.

History of Basel accord:

The attempts at harmonizing the capital adequacy standards internationally date back to 1988,

when the Basel Committee on Banking Regulations and Supervisory Practices, as it was then


6/46

6

named, released a capital adequacy framework, now known as Basel I. This initiative set out the

first internationally accepted framework for measuring capital adequacy and a minimum ratio to

be achieved by the banks. This norm was widely adopted in over 100 countries, and in India, itwas implemented in 1992. Over the years, however, the Basel I framework was found to have

several limitations such as its broad-brush approach to credit risk, its narrow coverage confined

to only credit and market risks, and non-recognition of credit risk mitigates, which encouragedcapital arbitrage through structured transactions. Moreover, the rapid advances in riskmanagement, information technology, banking markets and products, and banks internal

processes, during the last decade, had far outpaced the simple approach of Basel I to measuring

capital. A need was, therefore, felt to replace this Accord with a more risk-sensitive framework,which would address these shortcomings. . Accordingly, after a wide-ranging global consultative

process, the Basel Committee on Banking Supervision (BCBS) released on June 26, 2004 the

document International Convergence of Capital Measurement and Capital Standards: A Revised

Framework, which was supplemented in November 2005 by an update of the Market RiskAmendment. This document, popularly known as Basel II Framework, offers a new set of

international standards for establishing minimum capital requirements for the banking

organizations. It capitalizes on the modern risk management techniques and seeks to establish amore risk-responsive linkage between the banks operations and their capital requirements. It

also provides a strong incentive to banks for improving their risk management systems. The risk

sensitiveness is sought to be achieved through the now-familiar three mutually reinforcing.

The Importance of Risk Management in Banks is extremely important. Banking is a dynamicbusiness in which new opportunities and threats are constantly emerging. Banks are in the

business of incurring, transforming and managing risk. They are also highly leveraged. Former

US Federal Reserve Chairman Mr. Alan Greenspan in April 2004 had made an interesting

comment before the Senate Committee that Only through steady and continued progress inmeasuring and understanding risk will our banking institutions remain vibrant, healthy and

competitive in meeting the growing financial demands of the nation. The need to react tomarket developments including venturing in to new business or launching new products andservices, business continuity issues, and meeting the changing regulatory requirements make risk

management a dynamic exercise. Risk management is first and foremost a science and then an

art. Given the appetite for risk, if one uses accurate and relevant data, reliable financial models

and best analytical tools, one can minimize risk and make the odds work in ones favor.


7/46

7

Risk is the volatility of unexpected outcomes. Risk Management is the identification and

evaluation of risks to an organization including risks to its existence, profits and reputation(solvency) and the acceptance, elimination, controlling or mitigation of the risks and the effects

of the risks. Researchers and risk management practitioners worldwide have constantly tried toimprove on techniques in measuring and managing key risks: credit risk, market risk andoperational risk. Enormous strides have been made in the art and science of risk measurement

and management for the last two decades in line with the developments in the Banking

regulatory regime worldwide through Basel I & Basel II capital accord.

Credit risk is the possibility of losses associated with changes in the credit profile of borrowers

or counterparties. These losses, associated with changes in portfolio value, could arise due todefault (single or joint) or due to deterioration in credit quality.

Default risk- obligor fails to service debt obligations

Recovery riskrecovery post default is uncertain

Spread riskcredit quality of obligor changes leading to a fall in the value of the

loan

Concentration riskover exposure to a an individual obligor, group or industry

Correlation risk- concentration based on common risk factors between different

borrowers, industries or sectors which may lead to simultaneous default.

Factors affecting credit risk (expected and unexpected losses arising out of adverse credit events)

Exposure at Default (EAD): In the event of default, how large will be the

outstanding obligations if the default takes place.

Probability of Default (PD): The probability that the obligator or counterparty

will default on its contractual obligations to repay its debt.


8/46

8

Loss Given Default (LGD): The percentage of exposure the bank might lose in

case the borrower defaults. Usually it is taken as: 1-recovery rate.

Default Correlations: Default dependence due to common un-diversifiablfactors.

The Credit Management Process

Pre-Assessment

Pricing

Reject

Credit Grading:

CorrelationsLGDPD

CR Measurement

CR Management

Performance

EvaluationProvisioning Capital Allocn.

Accept

EAD

Model

Credit Risk Approaches in Basel II

The Accord encourages advanced risk management capabilities by stipulating three levels of

increasing sophistication with a reduction in capital charge. The approaches use differentmethods to calculate Probability of Default (PD), Loss Given Default (LGD) and Exposure at

Default (EAD).

Key Drivers of Credit Risk

Default Probability (PD) and Transition Probability

Credit Exposure (EAD)

Loss Given Default (LGD)

Default Correlation

Maturity


9/46

9

Loss Given Default (LGD)

LGD is the fraction of EAD that will not be recovered following default. It is the credit

loss incurred if an obligor of the bank defaults Loss Given Default is facility-specific because

such losses are generally understood to be influenced by key transaction characteristics such as

the presence of collateral and the degree of subordination.

Loss Given Default (LGD) = 1 - Recovery Rate

Loss Given Default is a common parameter in Risk Models and also a parameter used in the

calculation of Economic Capital or Regulatory Capital under Basel II for a banking institution.

This is an attribute of any exposure on bank's client.

If the bank uses the advanced IRB approach, then the Basel II accord allows it to use internal

models to estimate LGD. While initially a standard LGD allocation may be used (the foundationApproach), institutions that have adopted the IRB approach for the probability of default are

being encouraged to use the IRB approach for the LGD as well since it gives a more accurate

assessment of loss. In many cases, this added precision changes capital requirements.

Theoretically, LGD is calculated in different ways, but the most popular is 'Gross' LGD, where

total losses are divided by EAD. Another method is to divide Losses by the unsecured portion ofa credit line (where security covers a portion of EAD - Exposure at Default). This is known as

'historical' LGD. The historical RR is the sum of the cash flow received from defaulted loans

divided by total loan amount due at the time of default (EAD). Economic LGD is the economic

loss in the case of default, which can be very different from the accounting one. "Economic"means all costs (direct as well as indirect) incurred with recoveries have to be included in the

loss estimate, and that the discounting effects have to be integrated Loss-given-default (LGD) is

an important determinant of credit risk, and is the degree of uncertainty about how much thebank will not be able to collect if a borrower defaults. Because loan recovery periods may extend

over several years, it is necessary to discount post-default net cash flows to a common point in

time (the most suitable being the event of default). The LGD on defaulted loan facilities is thus

measured by the present value ofcash losses with respect to the exposure amount (EAD) at thedefaulted year. This can be estimated by calculating the present value of cash received post-

default over the year'snet discounted cost of recovery.

Ex-post: The ratio of losses post default to the book value of a defaulted obligation after default.

Ex-ante: The prediction of losses post default to the predicted Exposure at Default of a non-

defaulted obligation.


10/46

10

DOWNTURN LGD

Under Basel II, banks and other financial institutions are recommended to calculate 'DownturnLGD' (Downturn Loss Given Default), which reflects the losses occurring during a 'Downturn' in

a business cycle for regulatory purposes. Downturn LGD is interpreted in many ways, and most

financial institutions that are applying for IRB approval under BIS II often have differingdefinitions of what Downturn conditions are.

One definition is at least two consecutive quarters of negative growth in real GDP. Often,

negative growth is also accompanied by a negative output gap in an economy (where potentialproduction exceeds actual demand).

The calculation of LGD (or Downturn LGD) poses significant challenges to modelers and

practitioners. Final resolutions of defaults can take many years and final losses, and hence final

LGD, cannot be calculated until all of this information is ripe. Furthermore, practitioners are ofwant of data since BIS II implementation is rather new and financial institutions may have only

just started collecting the information necessary for calculating the individual elements that LGD

is composed of: EAD, direct and indirect Losses, security values and potential, expected future

recoveries. Another challenge, and maybe the most significant, is the fact that the defaultdefinitions between institutions vary. This often results in a so-called differing cure-rates or

percentage of defaults without losses. Calculation of LGD (average) is often composed of

defaults with losses and defaults without. Naturally, when more defaults without losses are addeda sample pool of observations LGD becomes lower. This is often the case when default

definitions become more 'sensitive' to credit deterioration or 'early' signs of defaults. When

institutions use different definitions, LGD parameters therefore become non-comparable.Many institutions are scrambling to produce estimates of Downturn LGD, but often resort

to 'mapping' since Downturn data is often lacking. Mapping is the process of estimating losses

under a downturn by taking existing LGD and adding a supplement or buffer, which is supposed

to represent a potential increase in LGD when Downturn occurs. LGD often decreases for some

segments during Downturn since there is a relatively larger increase of defaults that result inhigher cure-rates, often the result of temporary credit deterioration that disappears after the

Downturn Period is over. Furthermore, LGD values decrease for defaulting financial institutionsunder economic Downturns because governments and central banks often rescue these

institutions in order to maintain financial stability.

DETERMINANTS OF RECOVERY/LGD:

Empirically it has been observed that recovery rate (and hence LGD) is dependent on

The banks behavior in terms of debt renegotiation with debtors, compromise and settlements which are country specific.

The quality of collateral attached to loans Firm specific capital structure: Seniority standing of debt in the firm's overall capital

structure, leverage etc.

Industry tangibility: The value of liquidated assets dependent on the industry of theborrower.

Macro economic factors: industrial production, GDP growth, unemployment rate,

Interest rate and other macro economic factors have strong influence on LGD


11/46

11

IMPORTANCE OF LGD: LGD is not an issue for the standardized approach. The IRBF approach relies on values

furnished by the regulators. Institutions planning for the AIRB need to develop methods to estimate LGD, the credit

loss incurred if an obligor of the bank defaults, which is a key component to the credit

risk capital or risk weight. LGD is an important input for calculation of Expected and Unexpected Credit Loss and

Portfolio Economic Capital. According to BIS (June 2006) institutions implementing Advanced-IRB instead of

Foundation-IRB will experience larger decreases in Tier 1 capital, and the internal

calculation of LGD is a factor separating the two Methods.

Estimates of LGD are key parameters in a bank's risk-rating system that impact facility

ratings, approval levels, and the setting of loss reserves, as well as developing creditcapital underlying risk and profitability calculations.

For Regulatory Purpose:

CALCULATING LGD UNDER THE FOUNDATION APPROACHUnder Basel II to calculate the risk-weighted asset, which goes into the determination of therequired capital for a bank or financial institution, the institution has to use an estimate of the

LGD for each corporate, sovereign and bank exposure. There are two approaches for deriving

this estimate: a foundation approach and an advanced approach.

EXPOSURE WITHOUT COLLATERAL

Under the foundation approach, BIS prescribes fixed LGD ratios for certain classes of unsecured

exposures:

Senior claims on corporates, sovereigns and banks not secured by recognized

collateral attract a 45% LGD. All subordinated claims on corporates, sovereigns and banks attract a 75% LGD.

EXPOSURE WITH COLLATERALThe effective loss given default (LGD*) applicable to a collateralized transaction can be

expressed as

LGD* = LGD x (E* / E)

Where:LGD is that of the senior unsecured exposure before recognition of collateral (45%).

E is the current value of the exposure (i.e. cash lent or securities lent or posted)

E* should be calculated based on the following formula:E* = max {0, [E x (1 + He)C x (1HcHfx)]}

Where:

E* = the exposure value after risk mitigation

E = current value of the exposureHe = haircut appropriate to the exposure

C = the current value of the collateral received


12/46

12

Hc = haircut appropriate to the collateral

Hfx = haircut appropriate for currency mismatch between the collateral and exposure (The

standard supervisory haircut for currency risk where exposure and collateral are denominated indifferent currencies is 8%).

The *He and *Hc has to be derived from the following table of standard supervisory haircuts:

However, under certain special circumstances the supervisors, i.e. the local central banks maychoose not to apply the haircuts specified under the comprehensive approach, but instead toapply a zero H.

However, under certain special circumstances the supervisors, i.e. the local central banksmay choose not to apply the haircuts specified under the comprehensive approach, butinstead to apply a zero H.

Minimum LGD for Secured Portion of Senior Exposures:


13/46

13

LGD for unsecured & collateralized exposures as per RBIs IRB guidelines

(December 22, 2011) as per RBI:

CALCULATING LGD UNDER THE ADVANCED APPROACH

Under the A-IRB approach, the bank itself determines the appropriate Loss given default to be

applied to each exposure, on the basis of robust data and analysis. The analysis must be capable

of being validated both internally and by supervisors. Thus, a bank using internal Loss GivenDefault estimates for capital purposes might be able to differentiate

Loss Given Default values on the basis of a wider set of transaction characteristics (e.g. product

type, wider range of collateral types) as well as borrower characteristics. These values would be

expected to represent a conservative view of long-run averages. A bank wishing to use its ownestimates of LGD will need to demonstrate to its supervisor that it can meet additional minimum

requirements pertinent to the integrity and reliability of these estimates.


14/46

14

APPROACHES FOR LGD ESTIMATES UNDER AIRB APPROCH:Supervisors may permit banks to use their own internal estimates of LGD for corporate,

sovereign and bank exposures (and for retail loans also). LGD must be measured as the lossgiven default as a percentage of the EAD.

There might be significant differences in LGD estimation methods across institutions and across

portfolios some of which are listed below:

Market LGD: observed from market prices of defaulted bonds or marketable loans

soon after the actual default event. For marketable bonds and loans, the rating agenciesattempt to report the trading price of a defaulted obligation one month after default.

Implied Market LGD: LGDs derived from risky (but not defaulted) bond prices

(credit spreads) using a theoretical asset pricing model.

Workout LGD: This is the observed loss at the end of a workout process. This isbased on the discount of future cash flows resulting from the workout process from the

date of default to the end of the recovery process.

Drawback in Market LGD:

First it is limited to listed bonds, that are unsecured most of the time and thus bank cannotdraw such figures to compare their loss because bank loans are often backed by various

forms of collateral.

Secondly, secondary market prices of loans are not available in India.

With respect to the implied market LGD approach may be applicable for estimating up to BBB

category of corporate assets as bond spreads are not available for non-investment grades (below

BBB). For most bank loans, such market information will not be available, and a bank will haveto calculate the economic LGD from its own internal records. This requires discounting of all net

cash flows received at an appropriate discount rate. To determine LGD, a bank must be able toidentify accurately the borrowers that actually defaulted, the exposures outstanding at the time of

default and the amount and timing of repayments ultimately received along with the carryingcosts of non-performing loans/facilities (e.g. interest income foregone and workout expenses like

collection charges, legal charges, etc.). In addition, demographic information pertaining to the

borrower, including industry assignment, public or private designation (constitution) andgeographic domicile, are important for developing LGD estimates that are segmented according

these characteristics. Finally the structural elements of the defaulted facilities, such as whether

the banks interest is senior (according to claim: viz. first charge or second charge) orsubordinated and whether it has received any collateral, collateral segmentation etc. should be

noted.

Therefore I will be doing my project primarily on workout LGD. The workout LGD, values lossusing information from the workout. The loss associated with a defaulted facility is calculated bydiscounting the cash flows, including costs, resulting from the workout from the date of default

to the end of the recovery process. The loss is then measured as a percentage of the exposure at

default. The timing of cash flows and both the method and rate of discount are crucial in this

approach.There are four main issues that arise when using the workout approach to compute the loss of a

defaulted facility.


15/46

15

First, it is important to use the appropriate discount rate.

Second, there are different possibilities about how to treat zero or negative LGDobservations in the reference data.

Third, the measurement and allocation of costs associated with workout can be

complicated. Fourth, it is not clear how to define the completion of a workout.

Workout LGD

This section looks at the process of computing the workout loss of a defaulted facility, anddiscusses issues related to the measurement of the various components of the workout LGD

including recoveries, costs and the discount rate.

Components of workout LGDThere are three main components for computing a workout loss: the recoveries (cash or

noncash), the costs (direct and indirect) and the discount factor that will be fundamental to

express all cash-flows in terms of monetary units at the date of default. If all the cash flowsassociated with a defaulted facility from the date of default to the end of the recovery process are

known (i.e. we have complete information) then the realised LGD, measured as percentage of the

EAD at the time of default, is given by:

Realised LGD=1- [Ri(r)-Pj(r)/EAD]

where Ri is each of the i discounted recoveries of the defaulted facility, Pj is each of the jdiscounted payments or costs during the recovery period and r represents a discount rate.

Objective:

To get an insight of whole Risk Management currently being practiced in Indian Banks.

To study the current method use for estimating LGD and recovery factors by Indian

Banks..

To see the effect of various extreme scenarios through creating templates and how

historical and economic LGD is changing accordingly.

To study the dynamic method of calculating LGD and scope for its implementation inIndian Banks.

To do analysis of LGD estimates with help of a sample data of a bank and comparing the

results.


16/46

16

Scope of Project:

Loss Given Default is an essential input into the process of lending, investing, trading, or pricing

of loans, bonds, preferred stock, lines of credit, and letters of credit. Accurate LGD estimates are

important for provisioning reserves for credit losses, calculating risk capital, and determining fairpricing for credit risky obligations. Accurate estimates of LGD are fundamental to calculating

potential credit losses. So it is extremely important for Indian Banks to have effective models in

place to measure LGD and this project helps to determine various methods by which LGD can bemeasured for loans, bonds and stocks and various validation methods for these models.


17/46

17

Chapter II

Review of literature


18/46

18

A lot of literature has been written about this particular subject vis a vis research papers, books,

presentations, articles by authors, financial institutions ,consultancy firms, rating agencies etc.

Risk Management has been talked about a lot in recent past and after Sub Prime crisis there has

been an increased inclination towards this area. Books that I referred to are: Financial Risk

Management by Dun & Bradstreet which talks about Understanding Risk, Risk classification,

Measuring Risk, Tools for Risk management, Basel II etc and gives an insight of some useful

concepts.

My main focus was on studying papers issued by different authors for this I have read :

1. Losscalc: model for predicting loss given default (LGD) by Greg M. Gupton and Roger

M. Stein

2. What do we know about Loss given default by Til Schuermann3. Estimating Expected Loss Given Default by Petr Jakubik and Jakub Seidler

4. Estimating Recovery Rates on Bank's Historical Loan Loss Data by Arindam

Bandyopadhyay and Pratima Singh.

5. Losscalc v2: dynamic prediction of LGD modeling methodology by Greg M. Gupton andRoger M. Stein

6. Measuring LGD on Commercial Loans by Michel Araten, Michael Jacobs Jr., and

Peeyush Varshney JPMC ,RMA paper7. Pitfalls in Modeling Loss Given Default of Bank Loans by Marc Grtler and Martin

Hibbeln

8. Poor Default History Of Indian banks by Arindam bandyopadhyay

9. Stressed LGD in Capital analysis, Gary Wilhite.

Hand Book of Risk management II by Dr. Arindam Bandyopadhyay mainly for understanding

LGD.

Presentation and reading material on credit risk by Dr. Arindam Bandyopadhyay showing

prediction of LGD.


19/46

19

Summary of empirical studies of Loss Given Default:

Bibliographic

Reference

Period Facilities included Methodology and findings

Altman, Edward I. and

Vellore M. Kishore

(1996), Almost

Everything You Wanted

to Know About

Recoveries on

Defaulted Bonds.

19781995 696 defaulted bond issues.

By seniority.

By industrial class.

Market LGD.

Overall average LGD is 58.3%.

Average LGD for senior

unsecured debt 42%; senior

unsecured

52%; senior subordinate 66% and

junior subordinate 69%.

Statistically different LGD by

industry class even when

adjusted for seniority.

Initial rating of investment

grade versus junk-bond category

has no effect on recovery when

adjusted for seniority.

Time between origination and

default has no effect on recovery.

No statistical relationship

between size and recovery.

Araten, Michel, Michael

Jacobs Jr., and Peeyush

Varshney (2004),Measuring LGD on

Commercial Loans: An

18-Year Internal Study.

RMA Journal 86 (8), p.

96103.

19821999 3,761 large corporate loans

originated by JP Morgan

Workout LGD.

Mean LGD 39.8%, standard

deviation 35.4%. Range from -10% to 173%.

Model LGD approximately 5%.

Broke down LGD by type of

collateral and found LGD lowest

for loans collateralized by

accounts receivable.

Found a positive correlation

between LGD and the default

rate

using annual data from 1986

1999.

Gupton, Greg M. and

Roger M. Stein

(2002), LossCalc:

Moodys Model for

Predicting Loss Given

Default (LGD).

19812002 1,800 defaulted loans, bonds

and preferred stock.

U.S. debt obligations only.

Both senior secured and senior

unsecured loans.

Also includes Corporate

Market LGD.

Default price one-month after

default.

Beta distribution fits the

recovery data better than a

normal distribution.


20/46

20

Special Comment.

Moodys Investors

Services, February.

mortgages and industrial

revenue bonds.

Over 900 defaulted public and

private firms.

Issue size is US$ 680,000 to

US$ 2 billion; median size US$100 million.

There are a small number of

LGD less than zero (gains).

LossCalc predicts immediate

LGD and one-year horizon LGD.

Methodology is to map the

beta distribution of the LGDs to anormal distribution and then

perform OLS regression.

Historical averages of LGD by

debt type (loan, bond, preferred

stock) are an explanatory factor

for facility level LGD.

Historical averages of LGD by

seniority (secured, senior

unsecured, subordinate etc) are

an explanatory factor for LGD. Except for financial firms or

secured debt, firm leverage is an

explanatory factor for LGD.

Moving average recoveries for

12 broad industries are an

explanatory factor for LGD.

One-year PDs from Risk Calc

are an explanatory factor for

LGD.

Moodys Bankrupt Bond Index

is an explanatory factor for LGD. Average default rates for

speculative grade bonds from 12-

months prior to facility default

are an explanatory factor only for

immediate LGD.

Changes in the index of leading

economic indicators are an

explanatory factor for LGD

Hamilton, David T.,

Praveen Varma, SharonOu and Richard Cantor

(2003),

Default and Recovery

Rates of Corporate

Bond Issuers: A

Statistical Review of

Moodys Ratings

19822002 2,678 bond and loan defaults.

Includes 310 senior securedbank loan defaults

Market LGD.

Default price measures onemonth after default.

Distribution of recovery rates is

a beta distribution skewed

towards high recoveries (low

LGD).

Average LGD for all bonds is

62.8%; median LGD for all bonds


21/46

21

Performance 1920

2002. Special

Comment, Moodys

Investors Service.

is 70%.

Average LGD for senior secured

bank loans is 38.4%; median LGD

for senior secured bank loans

33%.

LGD in all debt instrumentsincreased in 2001 and 2002.

Average LGDs vary by industry.

LGD and default rates are

positively correlated.

Greg M. Gupton ,Roger

M. Stein(January 2005),

LossCalc: Moodys

Model for Predicting

Loss Given Default

(LGD) version 2.0

1981-2004 LossCalc is built on a global

dataset of 3,026 recovery

observations for loans, bonds,

and For loans, bonds, and

preferred stocks It projects LGD

for defaults occurringimmediately and for defaults

that may occur in one year.

Market LGD.

Default price one-month after

default.

Beta distribution fits the

recovery data better than a

normal distribution. Historical averages of LGD by

debt type (loan, bond, preferred

stock) are an explanatory factor

for facility level LGD.

Dynamic process for LGD

estimation.


22/46

22

Chapter III

Description of the Data, Variables and

Summary Statistics


23/46

23

The Data sample used here is for a small size south based bank of 64 borrowers.

The DETERMINANTS OF RECOVERY/LGD:

Empirically it has been observed that recovery rate (and hence LGD) is dependent on

The banks behavior in terms of debt renegotiation with debtors, compromise and settlementswhich are country specific.

The quality of collateral attached to loans.

Firm specific capital structure: Seniority standing of debt in the firm's overall capital structure,leverage etc.

Industry tangibility: The value of liquidated assets dependent on the industry of the borrower.

Macro economic factors: industrial production, GDP growth, unemployment rate, interest rateand other macro economic factors have strong influence on LGD.

CLASSIFICATION OF FACILITY TYPE:

Various credit facilities extended by bank can be classified into two categories viz. fund based

and non-fund based. When bank places certain funds at the disposal of borrowers and

borrowers avail these funds, such types of credit facilities are known as fund based.However,

there are certain types of advances which do not involve deployment of funds at least at theinitial stage though in contingencies funds are also involved. These arecalled non-fund based

advances.

Fund based credit facilities included in the database, used for the study are as following :-

1. Overdraft2. Cash-credit

3. Demand loan

4. Term loan5. Purchasing/Discounting of bills

6. Advanced Bills

7. Packing Credit8. Foreign Bills Purchased

9. Working Capital loan

1. Non-Fund based facility are as follows:

1. Letter of Credit2. Bank Guarantee

Above mentioned all the 10 categories of facility type were considered in the study. The sample Ihave taken for the bank which is south based small sized bank the facility with the borrowerswas only Term loan and Working capital loan wise.


24/46

24

COLLATERAL CLASSIFICATION:The borrowers were segregated on the basis of:

1. Accounts Receivable2. Cash and Marketable Securities#3. Cash and Marketable Securities#Inventory#Fixed Assets and/or

Equipment#Commercial Real Estate#

4. Commercial Real Estate#5. Fixed Assets and/or Equipment#6. Fixed Assets and/or Equipment#Commercial Real Estate#7. Inventory#Commercial Real Estate#8. Residential Real Estate#

11

54

No.of borrowers

Term Loan

Working Capital

33

2

13

533

17

No.of borrowers

Accounts Receivable

Cash and Marketable Securities#

Cash and Marketable

Securities#Inventory#Fixed Assets

and/or Equipment#Commercial Real

Estate#


25/46

25

TYPES OF BORROWERS:There are several types of borrowers. The database compiled for this study includes the

following types of borrower:

1. Sole proprietorship firms2. Partnership firms3. Private Limited companies4. Public Limited Company5. Others

From the pie chart above it is clear that the number of borrowers are maximum in case of public

limited liability companyunlisted.

STATE WISE CLASSIFICATION:All the borrowers are assigned a zone on the basis of its location. There are in all 6 zones in thestudy.

1. Andhra Pradesh2. Delhi3. Gujarat4. Karnataka5. Maharastra6. Kerala

6

3

8

2

No.of borrowers

Partnership

Private limited liabilitycompany

Public limited liability

company - unlisted

Others


26/46

26

SECTOR CLASSIFICATION:The borrowers belong to 6 different sectors. The definition used for classification as follows:

1. Financial Services

2. Infrastructure

3. Manufacturing4. Trading and Merchant Exports

5. Others

6

11

117

18

3

No.of borrowers

Andhra Pradesh

Delhi

Gujarat

Karnataka

Maharastra

Kerala

8

7

26

6

19

No.of borrowers

Financial Services

Infrastructure

Manufacturing

Others

Trading and Merchant

Exports


27/46

27

INDUSTRY CLASSIFICATION:The borrowers are broadly segregated into 12 different industries category. There is a separatecategory as other industries and trading.

1. Trading: It consists of all the traders belonging to different industries.

2. Transport equipment industry: It includes auto ancillary and automobiles.3. Chemicals Industry: it include pharmaceuticals and other chemicals

4. Construction Industry: It is comprises of infrastructure and real estate

5. Machinery Industry: It is composed of engineering and electronics6. Food and Beverages: It includes vegetable oil and beverages industries

7. Textiles Industry: it includes all the sub categories of yarn and textiles within it.

8. Metal and Metal Products: It consists of all the industries falling in the category of

metals both ferrous and non-ferrous and iron & steel.9. Leather Products: it includes the leather industry, shoes industry and other leather

products

10.Non-Banking Finance Company (NBFC)11.Paper Industry: all paper products fall in this category.12.Other Industries: It include industries like jems & jewellery, cement, ship breaking,

sugar, tea, agriculture, rubber, IT etc.

The pie chart shown here ,industries are being merged together so as to get better results.

6

9

10

9

10

8

12

No.of borrowers

Automobile

Chemicals,Dyes,Paints

Construction

Financial services

Food Processing

Textiles

Other industries


28/46

28

Interest rate Classification:

1. Fixed rate2. PLR linked

Bank Arrangement Classification:The borrowers are classified as:

1. Consortium2. Multiple banking3. Sole banking

The maximum number of borrowers are in sole banking arrangement.

38

27

No.of borrowers

Fixed rate

PLR linked

18

741

No.of borrowers

Consortium

Multiple banking

Sole Banking


29/46

29

BALANCE OUTSTANDING AT THE TIME OF DEFAULT:It is the outstanding balance at the time of default and it includes the interest suspenseamount. However, in many cases this amount was not available with the bank as the

amount outstanding was adjusted with the amount recovered during the years of default

(period between the date of default and date of compromise) thus reducing it by therecovered amount.


30/46

30

Chapter IV

Methodology


31/46

31

The Data sample used here is for a small size south based bank of 64 borrowers. Data

used here is as per a format having information related to the above mentioned variables

categories and other factors like IIP, contractual lending rates etc. Based on the data the

LGD was calculated in two ways:

1) HISTORICAL LGD:

This is a simply 1- recovery rate. The recovery till date had been reduced by the recovery

cost. This amount was used in the numerator.

This was divided by the EAD (denominator). This gave the recovery rates.

2) ECONOMIC LGD (DISCOUNTED CASHFLOW LGD):

The amount recovered till date was reduced by the recovery cost. This amount was

discounted @10.00% for the number of years in default. This gave the discounted

recovery. It was then divided by EAD (denominator). This rate is the recovery rate. The

recovery was flatly discounted by a single rate of discount.


32/46

32

Chapter V

Analysis/Results


33/46

33

A Data Sample of 68 borrowers is used here and with the help of LGD model following is the

results analysis and interpretation.

The results are as follows:

No.ofborrowers Historical LGD Economic LGD

68

Mean 68.52% 67.53%

Stdev 39.41% 36.29%

This is an average result for all the borrowers.

SOME OTHER INTERESTING RESULTS ARE AS FOLLOWS:

Sector Wise LGD:

The sector wise LGD is highest in case of Trading and Merchant Exports and lowest in case of

other industries. The highest contribution in the database is by the Manufacturing sector, withmore than half of the borrowers belonging to this sector. For sectors like financial services,

Infrastructure and manufacturing the mean LGD figures are almost in the same range. The sector

wise LGD is as follows:

Sector

No.of

borrowers

Historical

LGD Eco. LGD

Financial Services 8

Mean 65.58% 75.71%

Stdev 44.58% 31.74%

Infrastructure 7

Mean 61.69% 73.20%

Stdev 46.63% 33.15%

Manufacturing 26

Mean 64.91% 75.34%

Stdev 44.24% 31.40%

Others 6

Mean 64.28% 75.30%

Stdev 45.48% 31.66%

Trading and Merchant

Exports 19

Mean 66.19% 76.31%

Stdev 44.09% 31.33%


34/46

34

Constitution Wise LGD:

The LGD figure is highest for private limited liability company which means that the loss at the

time of default will be maximum for private companies as compared to public limited and

recovery will be lowest, as the LGD figure is lowest for public Limited Company. While for

Partnership Firms also LGD figure is quite high which is 61.62%. The constitution wise LGD isas follows:

Constitution

No.of

borrowers

Historical

LGD Eco. LGD

Partnership 6

Mean 61.62% 73.77%

Stdev 43.57% 30.36%

Private limited liability

company 3

Mean 69.17% 78.85%

Stdev 43.00% 29.90%

Public limited liability company

unlisted 8

Mean 67.24% 75.91%

Stdev 44.23% 31.41%

Others 2

Mean 67.20% 74.98%

Stdev 40.04% 31.63%

State wise LGD:

The recovery in case of southern state i.e. Kerala is highest and lowest in case of Andhra Pradesh

and Gujarat(Although it is a south based bank). The LGD figure for Kerala is 18.01% but the

number of borrowers here are three only, while in case of Gujarat LGD is 64.86% but the

number of borrowers here are 11 which may be the case for higher value of LGD in this state.The state wise LGD is as follows:

States No.of borrowers Historical LGD Eco. LGD

Andhra Pradesh 6

Mean 62.47% 69.79%

Stdev 40.27% 34.90%

Delhi 11

Mean 56.79% 70.91%

Stdev 46.40% 32.07%

Gujarat 11

Mean 64.86% 75.50%

Stdev 49.02% 34.09%

Karnataka 7

Mean 61.71% 73.76%

Stdev 46.31% 32.23%

Maharastra 18

Mean 46.74% 63.81%

Stdev 51.75% 35.92%

Kerala 3

Mean 18.01% 41.69%

Stdev 44.80% 29.91%


35/46

35

Facility Wise LGD:

The facility to the borrowers in the sample was as Term loan and Working Capital Loan. The

recovery in case of the Term Loan is less than in case of Working Capital loan. This is as per

expected.

Facility type No.of borrowers Historical LGD Eco. LGD

Term Loan 11

Mean 62.18% 73.50%

Stdev 45.49% 32.31%

Working

Capital 54

Mean 66.66% 76.50%

Stdev 44.28% 31.54%

Collateral type wise LGD:

The LGD is maximum for Inventory (commercial real estate) where average mean of LGD is

95.29% with standard deviation of 10.42% where there are 3 borrowers and minimum for Fixed

Assets and/or Equipment# Commercial Real Estate# where there are 13 borrowers and it can be

implied that recovery in this type of collateral will be higher. The collateral wise LGD is as

follows:

Collateral type

No.of

borrowers

Historical

LGD

Eco.

LGD

Accounts Receivable 3

Mean 63.93% 74.36%

Stdev 43.16% 31.33%

Cash and Marketable Securities# 3

Mean 54.33% 69.14%

Stdev 49.01% 32.75%

Cash and Marketable Securities#Inventory#Fixed

Assets and/or Equipment#Commercial Real

Estate# 2

Mean 95.00% 96.32%

Stdev 10.90% 8.89%

Commercial Real Estate# 13

Mean 55.40% 69.36%

Stdev 47.60% 33.24%

Fixed Assets and/or Equipment# 5

Mean 60.98% 72.19%

Stdev 43.33% 32.01%

Fixed Assets and/or Equipment#Commercial Real

Estate# 3

Mean 42.76% 60.40%

Stdev 54.33% 37.67%

Inventory#Commercial Real Estate# 3

Mean 95.29% 95.94%

Stdev 10.42% 8.90%

Residential Real Estate# 17

Mean 63.57% 74.26%

Stdev 45.35% 32.29%


36/46

36

LGD according to collateral frequency:

The collateral which was revalued semiannually has higher recovery but this was not obvious as

for Quarterly or more frequently the recovery should be more. The reason might be in

semiannual collateral value the number of borrowers are 3 only and there can be the case that the

exposure here is small and recovery is also good but for quarterly or more there are 53 borrowerswith LGD of 63.57% and recovery here will be less. The LGD collateral frequency wise is as

follows:

Collateral frequency

No.of

borrowers

Historical

LGD Eco. LGD

Less Frequently than

annual 10

Mean 66.50% 76.50%

Stdev 43.83% 31.13%

Quarterly or more

frequently 53

Mean 63.57% 74.26%

Stdev 45.35% 32.29%

Semiannually 2Mean 43.74% 61.29%Stdev 55.30% 38.51%

Type Guarantee wise LGD:

Almost all the collaterals by borrowers has guarantee and it is owner /shareholder guarantee with

58 borrowers and the LGD for this is 66.50%,while other third party guarantee number of

borrowers is 6 with lower LGD of 62.20%.The guarantee wise LGD is as follows:

Type of Guarantee

No.of

borrowers

Historical

LGD Eco. LGD

Other third Party 6

Mean 62.20% 73.39%

Stdev 46.07% 32.81%

Owner/shareholder 58

Mean 66.50% 76.50%

Stdev 43.83% 31.13%

Interest rate type LGD:

The LGD for both fixed rate and PLR linked loan is almost same with number of borrowers 38

and 27 respectively. The interest rate type LGD is as follows:

Interest Rate

type

No.of

borrowers

Historical

LGD Eco. LGD

Fixed rate 38

Mean 66.50% 76.50%

Stdev 43.83% 31.13%

PLR linked 27

Mean 65.66% 75.91%

Stdev 44.23% 31.41%


37/46

37

Industry Wise LGD:

The industry-wise historical and economic LGD is the highest in case of Automobile industry

and it is lowest for construction industry. The industry wise LGD is as follows:

Industry wise LGD No.ofborrowers Historical LGD Eco. LGD

Automobile 6

Mean 72.35% 77.80%

Stdev 37.82% 32.55%

Chemicals,Dyes,Paints 9

Mean 66.50% 68.55%

Stdev 40.27% 35.25%

Construction 10

Mean 65.51% 67.16%

Stdev 41.07% 35.75%

Financial services 9

Mean 67.73% 67.80%

Stdev 40.00% 36.11%

Food Processing 10Mean 68.06% 69.62%Stdev 39.88% 35.07%

Textiles 8

Mean 68.49% 74.80%

Stdev 41.61% 33.02%

Other industries 12

Mean 69.13% 67.30%

Stdev 39.62% 36.71%

Bank Arrangement Wise LGD:

The LGD is maximum for borrowers with sole banking arrangement and lowest with multiplebanking as shown in figures below:

Bank

Arrangement

No.of

borrowers Historical LGD Eco. LGD

Consortium 18

Mean 65.14% 67.30%

Stdev 40.31% 35.22%

Multiple

banking 7

Mean 63.19% 73.11%

Stdev 43.88% 32.89%

Sole Banking 41

Mean 68.24% 66.72%

Stdev 39.88% 36.82%


38/46

38

Debt Bank facility:

The LGD for borrowers with senior debt facility is lower and thus recovery is higher as

compared to borrowers with subordinated debt facility which is very obvious. The Debt bank

facility wise LGD is as follows:

Debt bank

Facility

No.of


Senior 64

Mean 68.52% 67.53%

Stdev 39.41% 36.29%

Subordinated 2

Mean 78.53% 43.11%

Stdev 36.47% 44.39%

Compromise loan wise LGD:

There were some accounts which were compromised but not settled till date. However, regularrecoveries were observed in these cases. There were 15 such partially settled accounts in thedatabase. The historical or actual LGD for these accounts is higher than the settled accounts.

This is as expected. The model predicts the LGD for not settled accounts as 69.82% and for the

settled accounts as 62.03%.

Compromised

No.of


Yes 27

Mean 62.03% 69.15%

Stdev 39.73% 34.84%

No 15Mean 69.82% 68.09%Stdev 39.76% 35.61%

Blank 24

Mean 68.24% 66.72%

Stdev 39.88% 36.82%

Margin wise LGD:

I have calculated Loan to value ratio and inverse of it as margin.Now we can see from the figures

below the LGD for borrowers with margin less than 50% is highest means recovery will be

lowest for these borrowers which has to be the case.The mean economic LGD with 10% discount

rate is 84.14%.While it is minimum for borrowers with margin of greater than 100% and it is

66.49% which was very obvious.

Margin

No.of

borrowers Economic LGD

Less than 50% 2

Mean 84.14%

Stdev 5.19%


39/46

39

Margin

No.of


More than 50%

and less than100% 17

Mean 75.22%

Stdev 25.42%

Margin

No.of


More than 100% 41

Mean 66.49%

Stdev 39.45%

DIFFICULTIES & SUGGESTIONS:

First, data limitations pose an important challenge to the estimation of LGD parameters ingeneral, and of LGD parameters consistent with economic downturn conditions in particular.

Hence we suggest, that bank should have a systematic method of recording the data.

Second, Due to the non-availability of some of the important data with the bank like year-wise

recovery, amount outstanding at the time of default etc., some assumptions were used (explainedin the detailed description of the data). This has led to the lower accuracy in the estimation of the

LGD.

Third, the bank needs to have some people who are especially involved in the process ofcollecting the required data for the future studies. The appropriate data is the backbone of anystatistical study.

Fourth, had there been enough data, the model would have given better results. Therefore, we

suggest bank to recalibrate the model with bigger sample data and validate it with a handout

sample before putting it to use.

Fifth, Since LGD estimation is the dynamic process, we suggest the bank to revalue collateralin a more frequent interval.


40/46

40

Chapter V

Suggestions for Future research


41/46

41

RECOMMENDATIONS FORFUTURE STUDY:

This study can be further extended to rate the borrower through a LGD rating model.

Bank can rate the borrower on the basis of the predicted recovery rates, which the model

will estimate. Based on this estimation benchmarking can be done for different category

of borrowers. The borrower with the highest recovery rates will be given AAA grade.Similarly grades can be benchmarked for the lower grades. Thus the facility rating and

the borrower rating can be clubbed together, if the bank uses the LGD rating Model.

However before carrying forward the exercise the model should be validated with

handout sample. Moreover bank need to have the rating information of the borrower.

Also here the static results of collateral are used but in future bank must focus on

dynamic value of collateral for which bank must do revaluation of collateral regularly.


42/46

42

Chapter VI

Executive Summary


43/46

43

Building a Recovery Model from loss perspective is hard. This is because it is difficult to

get enough predictive data since there are few commercial defaults. Therefore, the time

needed to collect default data was substantial. Total default data used in estimating therecovery model was of 68 defaulted borrowers.

There are different approaches to calculate LGD. In this project, on the basis of data

available for study, Ive calculated it in two ways, namely the historical or accounting

LGD, Economic LGD. The historical LGD is the calculated by directly dividing the net

recovery till date with the EAD, the economic LGD was calculated by discounting the net

recovery till date, at the rate 10.00%.

Some interesting results were as follows: The weighted average actual LGD for the bank

studied in this paper is 68.52%.

On further analysis it was found that the Automobile industry had higher LGD compared

to the manufacturing or trading industry. Facility-wise analysis shows that the bank hashigher recovery rates in case of working capital loan facility than in case of term loan

facility.

The defaulted accounts that were compromised and settled had lower LGD than the

accounts that were compromised and partly settled (recovery was in progress).

The sector-wise analysis indicated that the highest rate of recovery was in the other sector

followed by the manufacturing and infrastructure sector, while the lowest recovery was

from the Trading and merchant export sector.

The margin less than 50% is highest means recovery will be lowest for these borrowers

which has to be the case. The mean economic LGD with 10% discount rate is

84.14%.While it is minimum for borrowers with margin of greater than 100% and it is66.49% which was very obvious.


44/46

44

Chapter VII

Bibliography


45/46

45

Altman, E. I., and V. Kishore, 1996 (November-December), Almost Everything

You wanted to Know about Recoveries on Defaulted Bonds,Financial Analysts

Journal, pp. 57-64.

Altman, E. I., D. Crooke, and V. Kishore, 1999, Defaults and Returns on High-

Yield Bonds: Analysis through 1998 and Default Outlook for 1998-2000.

Araten, M., M. Jacobs Jr, and P. Varshney, 2004, Measuring LGD on

Commercial Loans: An 18-Year Internal Study, RMA Journal, May.

Greg M. Gupton ,Roger M. Stein(January 2005), LossCalc: Moodys Model for

Predicting Loss Given Default (LGD) version 2.0 .

Hamilton, David T., Praveen Varma, Sharon Ou and Richard Cantor (2003),Defaultand Recovery Rates of Corporate Bond Issuers: A Statistical Review of Moodys

Ratings Performance 1920-2002. Special Comment, Moodys Investors Service.

Gupton, Greg M. and Roger M. Stein(2002), LossCalc: Moodys Model for

Predicting Loss Given Default (LGD). Special Comment. Moodys Investors

Services, February 2002.

48 & 49 & Appendix 3 of RBI circular (Dec 2011 document).

Schuermann, T., 2004, What Do We Know About Loss Given Default?,

Working Paper, Wharton, Federal Reserve Bank of New York.

Asarnow, Elliot, and David Edwards. "Measuring Loss on Defaulted Bank Loans:

A 24-Year Study", Journal of Commercial Lending, (Mar-1995), Vol. 77, No.7,

pp. 11-23.

Estimating Recovery Rates on Bank's Historical Loan Loss Data by Arindam

Bandyopadhyay and Pratima Singh,February 2007.


46/46

finally final report

Documents