Are Credit Rating Agencies Discredited? Price Effects from Agency - - PowerPoint PPT Presentation

Are Credit Rating Agencies Discredited? Price Effects from Agency Sovereign Debt Announcements

Mahir Binici (Central Bank of Turkey), Michael Hutchison (UC Santa Cruz, California, U.S.A.), Evan Weicheng Miao (UC Santa Cruz, California, U.S.A.) Presented by Evan Weicheng Miao August 10th, 2017

Presented by Evan Weicheng Miao Binici, Hutchison and Miao 2017 August 10th, 2017 1 / 42

The view does not represent the Central Bank of ROC, Taiwan or Central Bank of Turkey.

Presented by Evan Weicheng Miao Binici, Hutchison and Miao 2017 August 10th, 2017 2 / 42

Outline

1

Introduction

2

Literature Review

3

Data and Methodology

4

Empirical Results

5

Conclusion

Presented by Evan Weicheng Miao Binici, Hutchison and Miao 2017 August 10th, 2017 3 / 42

Credit Rating Agencies and Announcements

1 Credit rating agencies (CRA), assigns credit ratings (e.g., AAA, BB),

which rate a debtor’s (firm or country) ability to pay back debt and the likelihood of default.

2 "Three Big" credit rating agencies: S&P, Moody’s, Fitch. 3 They also assign "watch" (negative or positive), "outlook" (negative

• r positive), "stable" and "developing" designation, not credit rating

changes but signals that there may be (or not) credit rating changes.

4 Developing: a review is ongoing; outlook: longer term; watch:

short-term.

5 For example, if a country is on negative watch, the country is very

likely, but not certain to be downgraded within the next 90 days.

Presented by Evan Weicheng Miao Binici, Hutchison and Miao 2017 August 10th, 2017 4 / 42

Credit Ratings (S&P)

1 25 rating categories (From highest AAA to lowest D)

Judgment about ability and willingness of debtor to repay the debt; the likelihood of default Credit ratings in S&P, Moody’s, and Fitch are generally comparable.

2 5 main factors in sovereign ratings

institutional effectiveness and political risks; economic structure and growth prospects; external liquidity and international investment position; fiscal performance and flexibility, as well as debt burden; monetary flexibility

Presented by Evan Weicheng Miao Binici, Hutchison and Miao 2017 August 10th, 2017 5 / 42

Literature on CRA Announcements on Financial Asset Price

1 CRA (mostly) on corporate and sovereign bonds and CDS spreads; (a

few) on other assets, such as equity prices, foreign exchange rates

2 Asymmetries in literature:

Outlook and Watch more important than credit rating changes Negative events more important than positive events; some literature finds the opposite Diverse studies

Presented by Evan Weicheng Miao Binici, Hutchison and Miao 2017 August 10th, 2017 6 / 42

Cantor and Packer (1996)

1 They consider announcements by Moody’s and S&P’s, 1987-1994 2 Daily sovereign bond prices for 18 countries, 79 announcements, 39 of

which are actual rating changes

3 Results:

Relative spread

• rose by 0.9%

for negative announcements. fell by 1.3%, for positive announcements; statistically significant.

Presented by Evan Weicheng Miao Binici, Hutchison and Miao 2017 August 10th, 2017 7 / 42

Ismailescu and Kazemi (2010)

1 The effect of sovereign credit rating change announcements by S&P

• n the CDS spreads for 22 emerging markets.

2 Method/Data: event study methodology / daily data over 2001-2008. 3 Results 1 (Mean Changes; NOT statistically significant):

Relative spread

• upgrade rating

lower the spreads by 11bps. downgrade rating, raise spreads by 67 bps but NOT statistically significant; means are affected by outliers.

4 Look at (1)median changes (2) proportion of negative/positive CDS

spread changes over the event window.

5 Results 2 (Median Changes and Proportion of Neg/Pos

Changes; partly Statistically Significant):

Median changes are significant for both negative/positive events. Positive rating events reduce the CDS spreads, statistically significant; not so for Negative events

Presented by Evan Weicheng Miao Binici, Hutchison and Miao 2017 August 10th, 2017 8 / 42

What’s missing? Evaluating Conditional Announcements

1 Consider the following two downgrade scenarios:

(a) Negative Watch → Downgrade: The downgrade is to some extent anticipated; less of a surprise (b) Stable → Downgrade: The downgrade is more of a surprise

The effects should be different: b>a

2

Neg Watch Neg Outlook Stable/Developing Pos Outlook Pos Watch Likely to upgrade Likely to downgrade

Presented by Evan Weicheng Miao Binici, Hutchison and Miao 2017 August 10th, 2017 9 / 42

Research Questions: CRA Effects on Sovereign CDS Markets

Do Credit Rating Agency Announcements on Sovereign Bonds Affect Market Expectations of Default Probability?

1 How do different types of announcements affect expectations?

(a) Credit rating changes (b) Watch changes (c) Outlook changes

2 Are the effects of rating changes conditional upon prior outlook/watch

state and prior grade level of bond?

3 Are there asymmetric effects between downgrades and upgrades? 4 Are CRA announcements anticipated? 5 Has the market effect of CRA announcements changed since the

global financial crisis (when CRA credibility damaged)?

Presented by Evan Weicheng Miao Binici, Hutchison and Miao 2017 August 10th, 2017 10 / 42

Signal Interpretation

1

Upgrade (Downgrade): always positive (negative) event

2

Transition between Outlook/Watch/Stable Neg Watch Neg Outlook Stable/Developing Pos Outlook Pos Watch Positive events Negative events

Presented by Evan Weicheng Miao Binici, Hutchison and Miao 2017 August 10th, 2017 11 / 42

Hypotheses 1: Effects of Positive/Negative Announcements

1 H1: CRA announcements of positive (negative) credit rating changes,

watch/review designations or outlook designations conditionally lower (raise) CDS spreads. Positive events Negative events Spread ↓ Spread ↑ Upgrade Downgrade Outlook/Watch transition Outlook/Watch transition

Presented by Evan Weicheng Miao Binici, Hutchison and Miao 2017 August 10th, 2017 12 / 42

Hypotheses 2-3: Effects of Credit Rating Changes on Different Outlook/Watch Status; Surprise or not

1 H2: CRA announcements of credit rating downgrades (upgrades) have

larger impact, in absolute value, when on prior stable/developing assignment status than when on prior negative (positive) outlook or watch status.

2 H3: CRA announcements of credit rating downgrades (upgrades) have

larger impact, in absolute value, when on prior negative (positive)

• utlook status than when on prior negative (positive) watch status.

Neg Watch Neg Outlook Stable/Dev Pos Outlook Pos Watch Upgrade effect:High Upgrade effect:Low Dowgrade effect:High Downgrade effect:Low

Presented by Evan Weicheng Miao Binici, Hutchison and Miao 2017 August 10th, 2017 13 / 42

Hypothesis 4-5: Effects (Direction) of Transition between Outlook/Watch/Stable Status

1 H4: CRA announcements of stable/developing status lower (raise)

CDS spreads from negative watch or outlook status, and raise (lower) CDS spreads from positive watch or outlook status.

2 H5: CRA announcements of negative (positive) outlook status raise

(lower) CDS spreads when bonds are on prior stable/developing status, and lower (raise) CDS spreads when bonds are on prior negative (positive) watch status. Neg Watch Neg Outlook Stable/Dev Pos Outlook Pos Watch Spread ↓ Spread ↑

Presented by Evan Weicheng Miao Binici, Hutchison and Miao 2017 August 10th, 2017 14 / 42

Hypotheses 6: Effects (Magnitude) of Transition between Outlook/Watch/Stable designation

1 H6: CRA announcements of negative (positive) watch status have a

larger impact for bonds on prior stable/developing status than those

• n prior negative (positive) outlook status.

Neg Watch Neg Outlook Stable/Dev Pos Outlook Pos Watch Larger effect Larger effect Smaller effect Smaller effect

Presented by Evan Weicheng Miao Binici, Hutchison and Miao 2017 August 10th, 2017 15 / 42

Hypotheses 7: Effects before and after Global Financial Crisis

1 H7: CDS spreads responds less in absolute value to CRA

announcements in the post-GFC period compared to the pre-GFC period. This hypothesis suggests that that CRA have been discredited since the GFC, with markets discounting their information value as to the default risk of a sovereign bond.

Presented by Evan Weicheng Miao Binici, Hutchison and Miao 2017 August 10th, 2017 16 / 42

Credit Default Swap (CDS) – our market-based measure of default risk

1 A credit default swap (CDS) is a financial contract that the seller of

the CDS will compensate the buyer in the event of a loan default (by the debtor) or other credit event.

2 The seller of the CDS insures the buyer against some reference loan

• defaulting. The buyer of the CDS makes a series of payments (the

CDS "fee" or "spread") to the seller and, in exchange, receives a payoff if the loan defaults.

3 CDS spread = insurance premium rate, which depends on how "risky"

the debtor is. High likelihood of default = Higher CDS spreads.

Presented by Evan Weicheng Miao Binici, Hutchison and Miao 2017 August 10th, 2017 17 / 42

What happened in case of a credit event, e.g., a downgrade?

1 Example: Brazil bonds, CDS widen after Fitch downgrade 2 New York, Dec 16,2015 (Reuters)- Brazil’s sovereign bonds and credit

default swaps widened by around 30bp Wednesday after Fitch stripped the sovereign of its investment-grade rating and lowered the credit to BB+.

3 On a priori ground, downgrade (or a negative credit rating

announcement) ⇒ CDS spread ↑.

4 Empirical question: Is the impact of CRA announcement on the

market statistically significant?

Presented by Evan Weicheng Miao Binici, Hutchison and Miao 2017 August 10th, 2017 18 / 42

Greek CDS spread ↑ when downgraded

Presented by Evan Weicheng Miao Binici, Hutchison and Miao 2017 August 10th, 2017 19 / 42

Data

1 55 countries, January 1, 2005 - December 31st, 2012; daily data 2 CDS 5-year tenor on sovereigns 3 CRAs: S&P, Moody’s, Fitch (Big 3) Presented by Evan Weicheng Miao Binici, Hutchison and Miao 2017 August 10th, 2017 20 / 42

Research Question

1 How much are the effects of CR announcements on market price of

CDS spreads? Are they statistically significant?

Presented by Evan Weicheng Miao Binici, Hutchison and Miao 2017 August 10th, 2017 21 / 42

"Abnormal" returns: disentangle the country-specific effect from the market-wide movement.

1 Apple’s Stock Dips After Death of Steve Jobs (CBS News, Oct 6,

2011)

2 How much of the stock price drop is attributable to the death of Steve

Jobs?

3 Stock price change = overall market effect + stock-specific effect 4 Need to look at S&P 500 or NASDAQ (market portfolio) and

disentangle the (Apple) stock-specific effect (abnormal return) from the market-wide movement.

5 Our application:

Need to look at CDS for other countries, and disentangle the country-specific effect (abnormal return) from the market-wide movement.

Presented by Evan Weicheng Miao Binici, Hutchison and Miao 2017 August 10th, 2017 22 / 42

Event Study Methodology I

1 Event: CRA announcements 2 Response: CDS “abnormal” returns 3 Define Day 0: event day; +1: next weekday day; −5: one week prior

to the event day

4 Event windows:

Two-day window: [0, +1] Pre-event windows: [-40, -15] (8 to 3 week prior to the event) and [-14, -1] (3 week to 1 day prior to the event) Post-event windows [+2, +14]

Presented by Evan Weicheng Miao Binici, Hutchison and Miao 2017 August 10th, 2017 23 / 42

Event Study Methodology II

1 Excluding overlapping events: Exclude any event that is preceded by

another event within 3 weeks (If one event occurs shortly after an earlier event, then the latter event may be less informative)

2 If 2 events of opposite nature on same day, exclude both events 3 If 2 events of same type (e.g. downgrades by 2 agencies for same

country), then counted as one event

Presented by Evan Weicheng Miao Binici, Hutchison and Miao 2017 August 10th, 2017 24 / 42

Event Study Methodology III

1 Following CAPM:

CDS return for country i, date t: RC

i,t = Si,t Si,t−1 − 1

CDS Market Spread Sk,t = 1

N

N

i=1 Si,t

Return on CDS Market Portfolio RC

k,t = Sk,t Sk,t−1 − 1

Abnormal return for country i, date t ARC

i,t = RC i,t − αC i − βC i RC k,t

Cumulative abnormal return CARC

i,[w1,w2] = w2 t=w1 ARC i,t

2 Find the average effect of a class of credit rating events (e.g., all

downgrades on negative watch) by averaging the SCAR (CAR standardized by its standard deviation) in the same class.

Presented by Evan Weicheng Miao Binici, Hutchison and Miao 2017 August 10th, 2017 25 / 42

Event Study Methodology IV

1 Alternative: signs of spread changes in event window 2 A priori expected change (pos or neg) 3 Test whether proportion of Pos/Neg spread change significantly

greater than 1/2.

4 Chi-square test Presented by Evan Weicheng Miao Binici, Hutchison and Miao 2017 August 10th, 2017 26 / 42

Results 1

1 What are the effects of conditional CRA

announcements?

2 Do they conform to our priors?

Ans: Yes for both

Presented by Evan Weicheng Miao Binici, Hutchison and Miao 2017 August 10th, 2017 27 / 42

Full sample Up/Downgrade effects

1 Most highly significant: t-stats (7/8) and spread changes (6/8) 2 All significant, expected signs; all spread changes direction expected 3 Downgrades larger than upgrades 4 Biggest downgrade effects: from neg. outlook and stable 5 Biggest upgrade effects: from stable 6 Smallest effects: from positive watch (insignificant) and neg watch 7 Unconditional = “all” . . . more information from conditional events Presented by Evan Weicheng Miao Binici, Hutchison and Miao 2017 August 10th, 2017 28 / 42

Full sample Outlook/Watch Announcements

1 All neg. watch significant and generally expected signs 2 Big rise in CDS from neg. outlook to watch 3 Neg. outlook sig. and large effects from stable 4 Most pos. watch / outlook NOT significant, exception is positive

• utlook from stable (as expected)

Presented by Evan Weicheng Miao Binici, Hutchison and Miao 2017 August 10th, 2017 29 / 42

Results 2 Are events anticipated? Ans: Only negative watches are partly anticipated.

Presented by Evan Weicheng Miao Binici, Hutchison and Miao 2017 August 10th, 2017 30 / 42

Full sample Credit Rating Changes, pre-event window [−40, −15] and [−14, −1]

1 No evidence of "anticipation" of CR changes in pre-event windows (no

significant SCARs)

Presented by Evan Weicheng Miao Binici, Hutchison and Miao 2017 August 10th, 2017 31 / 42

Full sample Outlook/Watch events, pre-event window [−40, −15]

1 Negative watch announcements anticipated: Significant positive

SCARs, 3-8 weeks prior to negative watch announcements (from neg

• utlook and stable)

2 Negative news already incorporated into CDS (rising CDS) 3 Contrast: In corporate CDS, some evidence that positive news

"leaked" and therefore CRA credit rating changes not significant news

Presented by Evan Weicheng Miao Binici, Hutchison and Miao 2017 August 10th, 2017 32 / 42

Full sample Outlook/Watch events, pre-event window [−14, −1]

1 No evidence of anticipation of outlook/watch announcement in the

3-week pre-event window.

Presented by Evan Weicheng Miao Binici, Hutchison and Miao 2017 August 10th, 2017 33 / 42

Results 3 How has the relative effects of CRA announcements changed since the Global Financial Crisis? Ans: CRAs are not discredited, at least in terms

• f credit rating changes.

Presented by Evan Weicheng Miao Binici, Hutchison and Miao 2017 August 10th, 2017 34 / 42

Credit Rating Changes, Pre-crisis

1 Some significant effects (3/8) during "normal" period 2 Little response on upgrades, except from stable (similar to full sample,

as expected)

3 No impact from neg watch to downgrade, biggest effect from neg

• utlook to downgrade (expected & same as full sample)

Presented by Evan Weicheng Miao Binici, Hutchison and Miao 2017 August 10th, 2017 35 / 42

Credit Rating Changes, Post-crisis

1

CRA NOT discredited in terms of credit rating changes; 5/8 significant after the crisis, compared with 3/8 during the normal crisis.

2

All downgrade categories significant, SCAR & spread directional change, and expected signs– more significant results than pre-crisis

3

Biggest "negative" event is from Stable (by CAR mean/median), as expected; overall magnitudes often smaller than pre-crisis

4

Only pos outlook to upgrade significant. Low power of tests? Comparison: Pre-crisis stable to upgrade significant (only)

Presented by Evan Weicheng Miao Binici, Hutchison and Miao 2017 August 10th, 2017 36 / 42

Outlook/Watch Changes, Pre-crisis

1 All Negative events significant w/ correct signs, large effects 2 Note: All to Neg Watch significant despite few events 3 Larger effects than full sample (CDS rise) Only one positive signal

significant: stable to pos outlook; same as full sample

Presented by Evan Weicheng Miao Binici, Hutchison and Miao 2017 August 10th, 2017 37 / 42

Outlook/Watch Changes, Post-crisis

1 Only 1/7 outlook/watch announcements significant, compared to 4/6

in pre-crisis period

2 Only neg outlook significant– from stable to negative outlook Presented by Evan Weicheng Miao Binici, Hutchison and Miao 2017 August 10th, 2017 38 / 42

Results 4 Do CRA announcement effects depend on prior level of rating grade? Ans: Yes, but one caveat applies.

Presented by Evan Weicheng Miao Binici, Hutchison and Miao 2017 August 10th, 2017 39 / 42

CRA announcement effects and the prior level of rating grade

1 High ratings very sensitive to downgrades 2 Low rating sensitive to upgrades and downgrades (from outlook) 3 Caveat: Sample size for classes other than high rating are low, which

may result in a lower testing power

Presented by Evan Weicheng Miao Binici, Hutchison and Miao 2017 August 10th, 2017 40 / 42

Comparison with other studies

1 Diverse studies in the literature

(1) Norden and Weber (2004), Hull et al. (2004) and Norden (2008) find that only negative CR announcements significant in corporate CDS (2) Ismailescu and Kazemi (2010) find only positive events matter (3) IMF (2010) and Alsakka and ap Gwilym (2013) claim watch/outlook most important

2 We find upgrades and downgrades important and negative watch lead

to significant effects

3 Our results, which condition upon the prior outlook/watch status, may

reconcile part of the contradictory results in literature

Presented by Evan Weicheng Miao Binici, Hutchison and Miao 2017 August 10th, 2017 41 / 42

Conclusion

1 CRAs not “discredited” after GFC; negative announcements and esp.

downgrades have larger effects

2 Conditioning of CR announcements on prior outlook/watch state

critical to precise measurement of effects, as expected

3 Effects largest for upgrades/downgrades if NOT on watch; largest

effects if on stable or outlook status prior.

4 CR changes not anticipated 5 Negative watch partly anticipated 6 High-rated bonds sensitive to downgrades, and low-rated bonds

sensitive to downgrades and upgrades

Presented by Evan Weicheng Miao Binici, Hutchison and Miao 2017 August 10th, 2017 42 / 42