1 / 15

Apoorva Javadekar - Ratings Quality Under ’Investor-Pay Model

Apoorva Javadekar - <br>

Download Presentation

Apoorva Javadekar - Ratings Quality Under ’Investor-Pay Model

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript


  1. Ratings Quality Under ’Investor-PayModel’ ApoorvaJavadekar BostonUniversity September 13,2013 A. Javadekar () September 13,2013

  2. Motivation Recent crises - also a ’ratings crises’ (Benmelech, Dlugosz) ’Inflations Ratings’ Phenomenon - Risky securities rated AAA Fallen Angels: 60% of the initially AAA rated productsduring 2005-07, rated below investment grade by 2009 for S&P (IMF, GFSR, 2009) Reasons - conflict of interest, ratings shopping, lack of effective monitoring, faulty risk models etc. Conflict of interest - Result of ’Issuer paysystem’ fees contingent upon ratings (Ratings shoppings) CRA’s trade off - current income vs future reputation A. Javadekar () September 13,2013

  3. Reforms Reforms suggested - Cuomo Commission, Dodd Frank Act, European UnionCommission Examples - Non - contingent fee under issuer pay, Investor Pay system, Platform system, rotation scheme, discloser norms Investor Paysystem Investors pays a non - contingent fee to CRA for ratings Eliminates ’conflict ofinterest’ Possible problems - free riding, regulatory arbitrage (Acharya, Calomiris) Objective: Understand the incentives of CRA under Investor pay system and shed light on possibleissues/problems A. Javadekar () September 13,2013

  4. Main Findings of thepaper CRA’s incentive to ’lie’ ? - CRA acts ’truthfully’, given the signal Stable ratings quality over business cycles?- Counter Cyclical- Lower inexpansions Incentives of reputed firms to maintain quality? - Optimal quality decreases after a reputational level ’Reputation Cycles’ ?- Yes! Reputation is built in recessions and consumed inexpansion Implications andReasons: ’Ratings inflation’ problem replaced by ’Ratings deflation’ Why? Costly to make an error in a badstate A. Javadekar () September 13,2013

  5. Literature Review Isaac and Shapiro (2012) - Cyclicality result under issuer pay model; Considers cyclical issue ignoring opportunisticbehavior Mathis et al (2009) - Reputational concerns work only when other income isdominant Skreta, veldKamp (2008) - Higher asset complexity induces ratings Shoppings Bolton, Frexias, Shapiro (2012) - ratings shopping + naive investors implies ratings inflation Becker, Milbourn (2010) - Competition reduced the quality of ratings post entry of FITCH What this paper adds? First model to analyze Investor pay system Cyclicality result under very weak assumption as compared to Isaac and Shapiro (2012) Sheds light on possible problems with investor pay model A. Javadekar () September 13,2013

  6. Model TimeLine At the beginning of period t State at time t isrealized st ∈ {g,b} follows a Markov process (could be persistent orIID) Projectarrives good with probability of λst Returns = π > 1 if good, 0 if bad, with probability1. λg > λb - only distinguishing feature between a good and a bad state Investor pays a non-contingent fee toCRA CRA choses the effort level to identify quality of the project: E={e1,e2,...,en} Efforts are costly and costs could be statedependent CRA receives a ’noisy signal’ about quality of the project. A(.) is the accuracy of the signal. 1 2 3 4 5 1 1√ A(et)=2+2et (1) CRA rates the product truthfully according to the signal and 6 investment takes place if rating is good A. Javadekar () September 13,2013

  7. Model TimeLine At the end of period t Project success or failure is known publicly zt ∈ {S, F,N} Beliefs about CRA’s accuracy are updated based on the outcome Belief at t - A distribution φt over E Expected Accuracy of ratings or reputation 1 2 A(φ)=.φ(e)A(e) e∈E Update - φt+1 =B(φt, st, zt ) Example: Bayesian Update if A(e) isknown (2) t λ φ (e)A(e, s) s t φt+1(e)|(zt =S) = . (3) t φ (e )A(e , s)λ t rr t s e∈E t t If A(.) is unknown - Arbitrary rules to update the reputation Example: z = S ⇒ reward, z = F ⇒ penalty s.t reputation hits lower bound (zero fees), z = N ⇒ noupdate A. Javadekar () September 13,2013

  8. Properties of Beliefupdates BayesianUpdate Reward for success and penalty for failure is same irrespective of the state for any given belief Badratinginbadtimes⇒upwardupdate,Badratingingoodtimes ⇒ downwardupdate Arbitrary Rules - Designed to follow similar patterns, but may have higher penalties and counter cyclical rewards Example of Arbitrary Rule - Lower Bound Penalty, Grim - Trigger Strategy (Abreu 1986, Isaac Shapiro (2012)) A. Javadekar () September 13,2013

  9. Equilibrium Fee and EquilibriumConcept Risk neutral investors operate in a competitivemarkets Equilibrium fee is such that given the beliefs, expected profits net of fees are zero forinvestors f(st,φt)=λs (π−1)A(φt)−(1−λs )(1−A(φt)) t t Fee is increasing in reputation and higher in a good state Definition (4) Given φ0 ∈ Φ, an equilibrium with Bayesian update for this economy isa ∞ t t sequence of fee schedules, optimal efforts {f (s , φ ), e(s ,φ )} anda t t t=0 ∞ t sequence of beliefs{φ} , such that for everyt, t=0 f (st, φt ) is determinedcompetitively e(st, φt ) solves the revenue maximization program of theCRA. φt+1=B(φt,st,zt). 1 2 3 A. Javadekar () September 13,2013

  10. Value functions Maximization Problem for CRA ∞ max(U)=maxEt.βtf(st,φt) (5) et t=0 subject to φt+1 =B(φt, st, zt ). Starting from a good state, after earning the current fee Vg(φt)=max(−c(et,g)+βpggEt(Vg(φt+1)+f(φt+1,g)) et +β(1−pgg)Et(Vb(φt+1)+f(φt+1,b))) Starting from a bad state after earning the current fee (6) Vb(φt)=max(−c(et,b)+β(1−pbb)Et(Vg(φt+1)+f(φt+1,g)) et +βpbbEt(Vb(φt+1)+f(φt+1,b))) (7) Expectation is with respect to future beliefs - choice of e induces a distribution overφt+1 A. Javadekar () September 13,2013 10 /15

  11. Optimal Policy Under BayesianBeliefs Figure :Optimal Policy Under Bayesian Update A. Javadekar () September 13,2013

  12. Simulation Results Under Lower BoundPenalty Table :Simulation Results: Lower BoundPenalty A. Javadekar () September 13,2013

  13. Discussion ofResults CyclicalAsymmetry Results from asymmetric cost of making errors Bad State - Most likely Error ⇒ bad project rated as good ⇒ high penalty toreputation Good State - Most likely Error ⇒ good project rated as bad ⇒ lower penalty toreputation ⇒ higher incentive in a bad state to keep quality of ratings high Non-Monotone effort choice in a good state Good State - Cost of error lower + Marginal gain low at higher levels of reputation ⇒ Decreasing efforts inreputation Bad State - Cost of error high enough to keep quality high even when marginal gain from higher quality is limited Ratings Deflation - Good projects turned down inexpansion A. Javadekar () September 13,2013

  14. Extensions Introducing other Income - Not much impact (already solved) Analytical results - Solving two period problem (obtained cyclical results) Competition - Horse race between Issuer pay and Investor pay model to acquire market share - cyclical credibility of each business model could bedifferent Long Lived Projects - Endogenous upgrades and downgrades A. Javadekar () September 13,2013

  15. Acknowledgments Thank You for coming ! A. Javadekar () September 13,2013

More Related