Linking Equity and Credit Risk - Northfield · y & Credit Risky & Credit Risk ... • Gray Merton &...
Transcript of Linking Equity and Credit Risk - Northfield · y & Credit Risky & Credit Risk ... • Gray Merton &...
Linking EquityLinking Equity
Northfield InformatioL d NLondon, Nove
y & Credit Risky & Credit Risk
on Services Seminarb 5 2009ember 5, 2009
OutOut
Which Models out there What can we @ Northfi
A crude test - Is there aEquity risk forecasts havCreditworthiness?
Next Steps
tlinetline
e link Equity & Credit?ield do to contribute?ny evidence of Northfield's yving any predictive power on
ExistingExisting
M t “O th P i i Merton, “On the Pricingrisk structure of interestFi V l 29 IFinance, Volume 29, Is
Leland, “Debt Value, BoOptimal Capital Structu
Gray Merton & Bodie; Gray, Merton & Bodie; Approach to MeasuringSovereign Credit Risk”Sovereign Credit Risk
ModelsModels
f C t D bt thg of Corporate Debt: the t rates” - Journal of
2 M 1974ssue 2, May 1974ond Covenants and
ure” “Contingent ClaimsContingent Claims
g and Managing
MertonMerton
C ll ti l h h ld Collectively, shareholde A) a call option on ow) p
assets (by paying off B) a put option – by d B) a put option – by d
company's creditors in exchange for paymin exchange for paym
Putting this in an optionyieldsyields...
Model Model
b id t hers can be said to have:wnership of a firms pf the firm's debt);defaulting forcing thedefaulting, forcing the to take the firms assets
mentment.n pricing framework
Merton Model (viMerton Model (vi• C(
– ica
• K =theouou
• S =fof
• SigE[E[e
• Solife
a Black Scholes)a Black Scholes)S,t) = price of the call option n this case, equity market p;= strike price – in this case e amount of debt tstanding;tstanding;= spot price – present value th t it + d btthe assets – equity + debt;gma = Asset vol – approx. by
it l] d fl t d b d btequity vol] deflated by debt; olve for (T-t) = option implied e of the debt;
Merton ModeMerton Mode
A t t• Assumes a constant rtherefore underestima
• Assumes bankruptcy nothing;g
• Assumes a uniform booption-less perpetualoption less, perpetual
• Can only be applied to• The model is Robust h
extended to address m
el Limitationsel Limitations
i kf t f i t triskfree rate of interest r –ating asset volatility;proceedings cost
ody of vanilla discount, ly rolling debt;ly rolling debt;o Corporate issuers;however as it can be most of these issues
Merton ModeMerton Mode
M t t d d hi• Merton extended his ocoupons & callable bo
• Leland (94) extends thcomplex corporate strsubordinated debt, inccost, taxes – e.g. debt
• Gray Merton & Bodie apply it to sovereign dapply it to sovereign d
el Extensionsel Extensions
d l t h dlown model to handle onds;he model to handle more ructures including gcorporating bankruptcy t restructuringextend the model to
debtdebt
Where NorthWhere North
O f th i i t One of the main inputs estimation) to all the moV l tilit A thi iVolatility – As this, in prby E[equity vol] deleverdebt an issuer has thdebt an issuer has – thcore business...
The question that needwould the equity vol be
fdebt” - Northfield's in a address this question
hfield fits inhfield fits in...
(th l th t d(the only one that needs odels discussed is Asset
ti i i t dractice is approximated raged by the amount of is variable IS Northfield'sis variable IS Northfield's
ds answering is “what if the issuer has no unique position to
Proove it AProove it - A
I th i i l Is there any empirical eEquity models being ab
dit thi ?creditworthiness? Or, what happens when
– 20 years of Equity rA graduate student;– A graduate student;
– A library...
A Crude TestA Crude Test
id f N thfi ld'evidence of Northfield's ble to forecast
n you've got:y grisk model exposures;; and; and
TheThe
17 f kl i• 17 years of weekly isschanges (moodys, S&
t d b Breported by Barrons, aintern, Steve Dyer (wipresentation would bepresentation would be3 months; a total of 89
• 20 years of the NIS Fucontinuous factor exp
&risk & total risk.
DataData
l l tisuer level ratings &P, Duff & Phelps, Fitch)
ll t d bas collected by our thout whom this
e a lot more boring) overe a lot more boring) over 956 entries.undamental model osures, stock specific
FundamenFundamenntal Modelntal Model • The 11 fundamentals:
– Earnings/Price– Book/Price– Div Yield– Trading Activity– 12 month Rel Strength– Log Mrkt Cap– Earnings Var– EPS Growth Rate– Revenue Price– Debt / Equity– Price Vol
The MethodThe Method
T f ll d / dTransform all upgrades / dscale
– Moody's• +1 for a full upgra• +.33 for every su
– S&P• +1 for a full upgra• +.5 for a subgrad
Normalize Beta, stock spethe fundamental model
d (data prep)d (data prep)
d d t idowngrades to a numeric
ade: B1 to Ba1ubgrade: B2 to B1
ade: A to AAde: AAA to AAA+
ecific & total risk fields in
The Method (straThe Method (stra
Li it l t 1265• Limit sample to 1265 ratings changes;
• Perform cross sectionfactor exposure, with tchanges (Y) as the decorresponding, laggeindependent variable,
– Y(t) = a + b*Price to( )• Increase the lag & see
numbersnumbers
aight regressions)aight regressions)
i d d t M d 'independent Moody's
nal regressions, one per the codified ratings gependent variable & the ed factor exposure as the for example...
o book(t-lag) +e( g)e what happens to the
Straight RegressioStraight Regressioa b
beta -0.1inDepVarbeta 0.1
-0.1-0.02
0 1
earningsToPricebookToPricedividendYield -0.1
-0.09-0.03
dividendYieldtradingActivityrelativeStrength
-0.08-0.05-0 09
glogOfMarketCapearningsVariablilityepsGrowthRate -0.09
0.02-0.07
epsGrowthRaterevenueToPricedebtToEquity
-0.05-0.1-0.1
priceVolatilityssRisktotalRisk 0.1totalRisk
on Results (0 Lag)on Results (0 Lag)t-a t-b r2
-0.12 -6.08 -7.08 0.040.12 6.08 7.08 0.040.06 -6.04 4.21 0.01
-0.13 -1.27 -12.48 0.110 12 6 27 7 33 0 04-0.12 -6.27 -7.33 0.04
-0.06 -5.42 -3.11 0.010.24 -1.61 17.76 0.20.09 -5.03 7.17 0.04
-0.17 -2.71 -10.41 0.080 07 -5 34 3 73 0 010.07 -5.34 3.73 0.01
-0.14 0.92 -13.22 0.12-0.05 -3.92 -3.64 0.01-0.12 -2.93 -9.36 0.06-0.19 -6.28 -11.79 0.1
-0.2 -6.31 -12.4 0.110.2 6.31 12.4 0.11
The “StraigThe Straig• Increasing the lag for the
DECREASES explanatorDECREASES explanator• Strongest factor is Relati
h th th t k i hwhether the stock price hthe market over 12 mont
iti b– a positive number meaupgrade is close on its
– Note the implied 1 year– Note the implied 1 yearvariable itself...
• Strong results for Total Rglogical – the riskier a comcredit downgrade
ght” Storyght Story...ese regressions ry powerry powerve strength a measure of
h i f ll f t thhas risen or fallen faster than hs
th t k i d i ll &ns the stock is doing well & an heels -
r lag in the composition of ther lag in the composition of the
Risk and Stock Specific risk –pmpany, the more likely a
“Straight” StStraight St
R l ti t th b i• Relative strength being sprice than book to price sare more important thanare more important than predicting ratings change
• Rsquared numbers are r• Rsquared numbers are rhigh... How do we interprthat later...
tory (contd )tory (contd.)
t th tstronger than revenue to suggests that market forces company fundamentals incompany fundamentals in
es (!)relatively low / T stats veryrelatively low / T stats very ret these numbers? More on
The Method (deThe Method (de
M ti l• More cross sectional rthe price to book exam
– Y(t) = a + b * [P/B (t• Increase lag & see whc ease ag & see
elta regressions)elta regressions)
i thi tiregressions... this time mple looks like this:t-1) - P/B (t)]hat happensat appe s
Delta Regressioga b
beta -0.11 -depVar
-0.1-0.08 -
0 1
earningsToPricebookToPricedividendYield -0.1 -
-0.1 --0.1
dividendYieldtradingActivityrelativeStrength
-0.06-0.1 -
-0 11
logOfMarketCapearningsVariablilitepsGrowthRate 0.11
-0.07 --0.11 -0 07
epsGrowthRaterevenueToPricedebtToEquity
i V l tilit -0.07 --0.11 --0.11 -
priceVolatilityssRisktotalRisk
ns (9 month lag)( g)t-a t-b r2
-0.14 -6.42 -8.37 0.050.06 -6.02 4.65 0.02-0.11 -4.82 -7.02 0.040 12 5 89 3 78 0 01-0.12 -5.89 -3.78 0.01
-0.07 -5.46 -3.57 0.010.08 -5.51 5.38 0.020.52 -3.67 16.15 0.18
-0.18 -5.96 -8.23 0.050 03 -6 41 1 99 00.03 6.41 1.99 0
-0.21 -3.91 -11.14 0.09-0.04 -6.27 -2.04 00 14 3 93 9 23 0 06-0.14 -3.93 -9.23 0.06
-0.16 -6.46 -9.28 0.07-0.17 -6.49 -9.99 0.08
The DeltaThe Delta
H L f M k t C• Here Log of Market C– increase in Market c
ti hratings change – samargument in previous
• Very strong T-Stats forisk – if the risk modethe likelihood of a dow
a Storya Story ...
C i th d i t f tCap is the dominant factor cap forecasts positive
l ti t the as relative strength regression
or stock specific & total ls estimate INCREASES, wngrade increases
Aside on Credit AAside on Credit A
C dit ti i d• Credit rating agencies donthan quarterly since that isstatements are issued.
• Full annual reports are onreasonable guess for frequg qaverage of every six monthbeing diligent.
• Our results show that thesepredictable, even with a timimplies that Credit Agencieimplies that Credit Agencie
• In the interim, why not useprobability that can be estiprobability that can be esticurrent?
Agency DiligenceAgency Diligence
't i ti ft't review ratings more often how often interim financial
ly issued once a year, so a uency of reviews might be an y ghs, if the rating agencies are
e changes are somewhat me lag of 9 months which es are behindes are behind a different measure of default mated at any time & bemated at any time & be
Problems withProblems with
Th d• There are no records upgrades / downgrade
hi h t t kiwhich amounts to a ki– Even so, if we were
assuming say 30,00that T stats in the 5+have many results ihave many results i
• We are using regressdependent variabledependent variable –deflates our R2 meas
h the Methodh the Method
f ll th ditfor all the credit es that didn't happen i d f i hi biind of survivorship bias
e to adjust for this bias by 00 “0” events we can still say + range are significant – we n this rangen this range...ion to estimate a discrete this almost certainlythis almost certainly ures.
Fixing the PFixing the P
I t d d• Insert dummy recordsthink a credit rating ag
d l ft th tiand left the rating unc– Count the number o
level of debt would investigation, subtrachanges & insert thchanges & insert threcords...
• Use discrete statistics• Use discrete statisticscontingency tables to in addition to our tradiin addition to our tradi
ProblemsProblems
f ll th tis for all the times we gency looked at an issuer h dchanged:
of issuers each month whose have warranted an act the number of actual at resulting number of “0”at resulting number of 0
s such as Chi^2s such as Chi^2 estimate goodness of fit
itional R^2itional R 2
How can Northresu
C diti i (b k d l• Conditioning (backward ldata with model forecast
if th hi t i b– e.g. if the historic probaour model predicts %1increase in bad events
• Provides an intellectual bModel (with Northfield asfuture forecasting;
• Start calculating & reportg pEE problems;
• Improve credit risk estimplevel in the EE Model...
field use these ults?l ki ) t iti t ilooking) transition matrix ts,
bilit f d d i 1 % dability of downgrade is 1 % and .5 this could be a big deal (50% ) for investors;)
basis for using the Merton sset volatility forecasts) for
ting issuer specific risk for g p
mation at the individual bond
One FinaOne Fina
R l ti hi i• Relationships in rever• What can Debt risk sa• Is a change in OAS lin
Si il t f l– Similar type of analycross sectional regrthe independent vathe independent vavol as the dependen
– Could work becauseCould work becauseliquid indications of issuer creditworthin
l Thoughtl Thought
rse...ay about credit?ynked to Equity volatility?
i i tysis, in reverse... set up a ression with delta OAS as riable and Equity returns orriable and Equity returns or nt variable...e OAS are perhaps the moste OAS are perhaps the most what market perception of ess...
big thanks to Sbig thanks to SSteve DyerSteve Dyer ...