Credit Credit- -Rating Shopping, Selection and Rating Shopping, - - PowerPoint PPT Presentation

1 1 1 1

“ “Credit Credit-

• Rating Shopping, Selection and

Rating Shopping, Selection and the Equilibrium Structure of Ratings the Equilibrium Structure of Ratings” ”

Francesco Francesco Sangiorgi Sangiorgi Stockholm School of Economics Stockholm School of Economics Jonathan Jonathan Sokobin Sokobin Securities and Exchange Commission Securities and Exchange Commission Chester S. Chester S. Spatt Spatt Tepper Tepper School of Business School of Business Carnegie Mellon University and Carnegie Mellon University and National Bureau of Economic Research National Bureau of Economic Research 2 0 0 9 UBC Sum m er Finance Conference 2 0 0 9 UBC Sum m er Finance Conference W estbank W estbank , British Colum bia , British Colum bia July 28, 2009 July 28, 2009

2 2 2 2

Selling Information Selling Information

• “

“ Investor pays Investor pays” ” model model

• Conflicts of interest (limited)

Conflicts of interest (limited)

• Difficulty of exclusion

Difficulty of exclusion

• Availability of ratings and regulation

Availability of ratings and regulation

• “

“Issuer/ Issuer/Securitizer Securitizer pays pays” ” model model

• Conflicts of interest (extensive)

Conflicts of interest (extensive)— —Ratings shopping for individual Ratings shopping for individual transactions, but also positive views about securitization and transactions, but also positive views about securitization and tranching tranching

• Limited incentive for re

Limited incentive for re-

• rating

rating

• Issuer chooses which ratings to purchase, publish

Issuer chooses which ratings to purchase, publish

3 3 3 3

Demand for High Ratings Demand for High Ratings

Information and asset demand Information and asset demand Net capital rules Net capital rules Money market funds; permissible holdings Money market funds; permissible holdings “ “Investment grade Investment grade” ” and suitability and suitability Purchase from agency offering the highest rating Purchase from agency offering the highest rating If multiple ratings at a level required, then nth If multiple ratings at a level required, then nth-

• order statistic key
• rder statistic key

Manufacturing rating, structuring ( Manufacturing rating, structuring (“ “regulatory arbitrage regulatory arbitrage” ”) )— —AAA AAA meaning very distorted (corporate bonds vs. municipals too) meaning very distorted (corporate bonds vs. municipals too)

4 4

Spotlight on Credit Rating Agencies

Scope to mis-value an entire asset class rather than specific loans Potential for systemic risk as investors relied upon CRAs While reputation weakened, investors and regulators look to CRAs Conflict of interest and the payment model – “Selection” and rating shopping Outsourcing due diligence (especially to few players) is an odd basis for asset management, despite scale economies – creating diverse signals Move towards reduced regulatory reliance on ratings

5 5 5 5

Related Regulatory Proposals Related Regulatory Proposals

What factors were considered by the rating agency? What factors were considered by the rating agency? What was the basis of the ratings? What was the basis of the ratings? – – Especially important for structured financing (encourage Especially important for structured financing (encourage unsolicited ratings) unsolicited ratings) Track record disclosure Track record disclosure (Reputation is apparently not powerful enough to force above (Reputation is apparently not powerful enough to force above disclosures) disclosures)

6 6

Reduced Regulatory Reliance on Ratings ?? Reduced Regulatory Reliance on Ratings ??

– – Mitigate systemic risk ( Mitigate systemic risk (mis mis-

• value an asset class)

value an asset class) – – Avoids allowing agencies to sell regulation and amplifies Avoids allowing agencies to sell regulation and amplifies conflict of interest conflict of interest – – Ratings for different products have different meanings Ratings for different products have different meanings--

• -reduce

reduce effort to engage in effort to engage in “ “regulatory arbitrage regulatory arbitrage” ” – – Encourages decentralized and competing due diligence Encourages decentralized and competing due diligence – – ” ”Dead on Arrival Dead on Arrival” ”: Asset managers like current legal safe : Asset managers like current legal safe harbors harbors

7 7

Economics of Selection and Regulation

“Selection” theme Solicited vs. unsolicited ratings “Notching” Comparative statics of cost, correlation and transparency Extension to “winner’s curse”

8 8

Solicited vs. Unsolicited Ratings Solicited vs. Unsolicited Ratings

Solicited Solicited— —Issuer purchases Issuer purchases Unsolicited is voluntary choice of agency Unsolicited is voluntary choice of agency – – Difficult for securitization Difficult for securitization – – Regulators discouraged this for several years (were Regulators discouraged this for several years (were unsolicited ratings punitive?), now trying to promote to unsolicited ratings punitive?), now trying to promote to limit conflicts of interest from solicited ratings limit conflicts of interest from solicited ratings— —example example

• f
• f “

“ unintended consequences unintended consequences” ” and regulation and regulation – – “ “ Conflicts of interest Conflicts of interest ” ” and the analogy to and the analogy to “ “ second second-

• best

best ” ” frictions (should we shut down individual frictions?) frictions (should we shut down individual frictions?) Due to unsolicited ratings hard to charge (incentive Due to unsolicited ratings hard to charge (incentive constraint) for solicited ones constraint) for solicited ones— —unless unsolicited ratings unless unsolicited ratings tend to be lower tend to be lower

– – Otherwise, no incentive to purchase rating Otherwise, no incentive to purchase rating

9 9 9 9

Otherwise, no incentive to purchase rating Otherwise, no incentive to purchase rating Motive need not be punitive Motive need not be punitive Selection story Selection story— —Solicited ratings have access to fine details Solicited ratings have access to fine details

– – Firms for which that would be beneficial net of cost will pay fo Firms for which that would be beneficial net of cost will pay for a rating r a rating

Are unsolicited ratings artificially low or solicited ratings ar Are unsolicited ratings artificially low or solicited ratings artificially tificially high due to ratings shopping? high due to ratings shopping? Which is the important conflict of interest? Which is the important conflict of interest?

Why are Unsolicited Ratings Lower than Why are Unsolicited Ratings Lower than Solicited Ones? Solicited Ones?

10 10 10 10

Why are Unsolicited Ratings Lower than Why are Unsolicited Ratings Lower than Solicited Ones? (cont.) Solicited Ones? (cont.)

Unsolicited rating if firm Unsolicited rating if firm’ ’s estimate is below s estimate is below x x* and solicit if estimate * and solicit if estimate exceeds exceeds x x*. Optimal *. Optimal x x* is determined by ratings cost. * is determined by ratings cost. (If cost = 0, then (If cost = 0, then x x* = 0 as in * = 0 as in Akerlof Akerlof [1970]) [1970]) Under ratings shopping firms have choice of which agency or Under ratings shopping firms have choice of which agency or agencies to solicit agencies to solicit— —reinforces the effect reinforces the effect Multiple agencies extend the Multiple agencies extend the Verrecchia Verrecchia [1983] disclosure intuition [1983] disclosure intuition

a

X*

b

11 11 11 11

Selectivity Framework Selectivity Framework

Agency i possesses own signal about distribution of one Agency i possesses own signal about distribution of one-

• period

period payoff, payoff, f fi

i

• - simplified view of a rating

simplified view of a rating

Role of signal is to classify asset for regulatory objectives Role of signal is to classify asset for regulatory objectives

– – each agency treated equivalently each agency treated equivalently – – high ratings desired to minimize need for regulatory capital high ratings desired to minimize need for regulatory capital

Diverse models, common knowledge Diverse models, common knowledge Goal is to maximize NPV of the issuer Goal is to maximize NPV of the issuer

12 12 12 12

Ratings Shopping Ratings Shopping

Implicit shopping uses prior knowledge Implicit shopping uses prior knowledge— —can be transparent or can be transparent or somewhat somewhat noisy noisy Explicit shopping also can be perfect, but depends upon extent o Explicit shopping also can be perfect, but depends upon extent of f search search costs costs Either can produce a substantial selection effect Either can produce a substantial selection effect— —purchase highest purchase highest rating identified rating identified Selectivity even if legal standard based upon at least two ratin Selectivity even if legal standard based upon at least two ratings gs (2 (2nd

nd-

• order statistic in auctions) instead of one
• rder statistic in auctions) instead of one

Charging for indicative rating (NY AG settlement) reduces explic Charging for indicative rating (NY AG settlement) reduces explicit it shopping shopping

• -Need not reduce

Need not reduce “ “selection selection” ” because extent of transparency is because extent of transparency is endogenous (i.e., less noise could result) endogenous (i.e., less noise could result)

• -How do

How do “ “costs costs” ” influence equilibrium regulation? influence equilibrium regulation?

13 13 13 13

Selection and Issuer Pays Selection and Issuer Pays

“ “Shadow Shadow” ” or

• r “

“v virtual irtual” ” ratings are below published ones ratings are below published ones

– – Does the issuer purchase the Does the issuer purchase the “ “high high” ” or the

• r the “

“low low” ” rating? rating?

Import of not being rated Import of not being rated

– – in general in general – – by particular agencies by particular agencies

Single vs. multiple ratings at a level Single vs. multiple ratings at a level Split ratings (empirical literature Split ratings (empirical literature— —different inferences) different inferences)

14 14 14 14

Notching and Competition Notching and Competition

Notching Notching— —formulaic haircutting of ratings from other agencies in re formulaic haircutting of ratings from other agencies in re-

• rating

rating components of securitization structures components of securitization structures Selectivity implies notching; Virtual ratings if not selected ty Selectivity implies notching; Virtual ratings if not selected typically lower pically lower 2007 SEC framework allows notching by raters 2007 SEC framework allows notching by raters – – Promotes competition in standards and development of distinct Promotes competition in standards and development of distinct reputations (rationale for 2007 framework) reputations (rationale for 2007 framework) Mutual notching due to heterogeneity in models Mutual notching due to heterogeneity in models – – heterogeneity is greatest when correlation in agency signals is heterogeneity is greatest when correlation in agency signals is least least – – particularly relevant for securitizations particularly relevant for securitizations Example of anti Example of anti-

• competitive and punitive effect (via pricing)

competitive and punitive effect (via pricing)

– – notching when the different agencies use the same model notching when the different agencies use the same model

15 15 15 15

Sequential Structure Sequential Structure

Our formal model assumes the decision to solicit additional Our formal model assumes the decision to solicit additional agencies is sequential agencies is sequential— —analogy to search analogy to search If two agencies and costly to solicit, then the second is If two agencies and costly to solicit, then the second is solicited only if the indicative rating from the first is in an solicited only if the indicative rating from the first is in an “ “ interior interior” ” interval (assuming correlated underlying signals interval (assuming correlated underlying signals and ratings) and ratings) — —if first indicative rating is very low, then if first indicative rating is very low, then utilize unconditional rating and if initial indicative utilize unconditional rating and if initial indicative assessment is very high, then it assessment is very high, then it ’ ’s not worth the cost to s not worth the cost to potentially get a slightly higher one (e.g., suppose already potentially get a slightly higher one (e.g., suppose already AAA!) AAA!)

16 16 16 16

17 17

18 18 18 18

19 19 19 19

20 20 20 20

21 21

22 22

“Winner’s Curse” and Credit Ratings

Auction analogy Auction analogy— —Should the information content of a rating being Should the information content of a rating being “ “published published” ” be reflected in its rating? be reflected in its rating? Should agencies adjust for Should agencies adjust for “ “winner winner’ ’s curse s curse” ” as only purchased when as only purchased when an outlier? an outlier?

• -What are the ratings supposed to capture?

What are the ratings supposed to capture?

If not, should regulators adjust standards to reflect the streng If not, should regulators adjust standards to reflect the strength of th of the the “ “winner winner’ ’s curse s curse” ”? ?— —as in auction theory key is cross as in auction theory key is cross-

• sectional

sectional dispersion in signals dispersion in signals Number of signals (agencies), techniques Number of signals (agencies), techniques

• -Interpretation of maximum signals changes

Interpretation of maximum signals changes

• -Selection over likely ratings net of cost

Selection over likely ratings net of cost

23 23

“ “Winner Winner’ ’s Curse s Curse” ” and Credit Ratings (cont.) and Credit Ratings (cont.)

For example, winner For example, winner’ ’s curse correction implies ratings decline with s curse correction implies ratings decline with more agencies more agencies Yet, Becker and Yet, Becker and Milbourn Milbourn [2008] document that Fitch entry led to [2008] document that Fitch entry led to higher higher ( (not lower not lower) ratings ) ratings— —competitive (bias) effect competitive (bias) effect— —consistent with consistent with ratings shopping, but not winner ratings shopping, but not winner’ ’s curse s curse

24 24

Biases in Ratings and the Biases in Ratings and the “ “Winner Winner’ ’s Curse s Curse” ”

Agency producing high ratings (first Agency producing high ratings (first-

• order stochastic dominance) for
• rder stochastic dominance) for

a fixed ease of valuation or market share requires more notching a fixed ease of valuation or market share requires more notching downward than reverse (under winner downward than reverse (under winner’ ’s curse) s curse) Agency producing less precise ratings in a space requires more Agency producing less precise ratings in a space requires more notching downward than agency with more precise ratings (greater notching downward than agency with more precise ratings (greater “ “winner winner’ ’s curse s curse” ”) ) Difficult to value instruments such as securitizations require m Difficult to value instruments such as securitizations require more

• re

notching notching— —regulatory issue emerged only for securitizations as much regulatory issue emerged only for securitizations as much heterogeneity in valuation needed for large effects heterogeneity in valuation needed for large effects Greater selectivity and equilibrium notching for longer Greater selectivity and equilibrium notching for longer-

• term bonds,

term bonds, lower lower-

• rated bonds, and tranches

rated bonds, and tranches

25 25

Tranching Tranching and Securitization and Securitization

Basic selection argument does not require Basic selection argument does not require tranching tranching or even

• r even

securitization securitization However, magnitudes are much larger due to the extent of However, magnitudes are much larger due to the extent of heterogeneity in the rating agency heterogeneity in the rating agency’ ’s assessment. Heterogeneity, s assessment. Heterogeneity, which arises from instruments being hard to value (as in which arises from instruments being hard to value (as in tranching tranching and securitization), is crucial to selectivity and securitization), is crucial to selectivity Because these are costly to value, rating agencies have made Because these are costly to value, rating agencies have made available a range of techniques available a range of techniques— —by offering more choices to the by offering more choices to the borrower these greatly enhance selection borrower these greatly enhance selection When substantial selectivity biases arise, concerns about the When substantial selectivity biases arise, concerns about the reliability of ratings are stronger reliability of ratings are stronger--

• -What do the ratings mean?

What do the ratings mean?

26 26

Additional Work on Rating Shopping Additional Work on Rating Shopping

Skreta Skreta, V. and L. , V. and L. Veldkamp Veldkamp, , “ “ Rating Shopping and Asset Rating Shopping and Asset Complexity: A Theory of Ratings Inflation, Complexity: A Theory of Ratings Inflation,” ” Stern School, Stern School, New York University, presented at Carnegie New York University, presented at Carnegie-

• Rochester

Rochester conference, fall 2008 (most closely related) conference, fall 2008 (most closely related)