Concentrated Capital Losses and the Pricing of Corporate Credit Risk

Transcription

1 Concentrated Capital Losses and the Pricing of Corporate Credit Risk Emil Siriwardane 1,2 1 Harvard Business School 2 U.S. Treasury, Office of Financial Research (OFR) WFA 2016 Disclaimer: The views in this presentation are not necessarily a reflection of the Depository Trust and Clearing Corporation (DTCC), the U.S. Treasury, or the Office of Financial Research (OFR) 1/20

2 Introduction 1/20

3 Motivation A core assumption of neoclassical asset pricing is that capital can always flow frictionlessly to investment opportunities For many asset classes, there are barriers to capital entry: Investment requires specialized knowledge and technology Capital itself may be scarce due to agency problems Potentially exacerbated by internal capital market frictions In the face of these frictions, asset price dynamics may: Divert from neoclassical predictions, at least at higher frequencies Instead driven by capital of a small subset of trading desks 1/20

4 This Paper New empirical evidence of this story in the credit default swap market Proprietary dataset on credit default swap (CDS) positions Effectively covers entire U.S. market Credit default swap (CDS) definition/terminology: Seller insures against default of an underlying reference entity, r Buyer pays a premium, the CDS spread Main findings: Net sellers are highly concentrated Capital movements of sellers drive spread variation Degree of Segmentation + Internal Capital Market Frictions + Distribution Effects 2/20

5 Literature Review 1. Asset pricing w/ frictions - capital of specific agents matters Limits to Arbitrage - Shleifer and Vishny (1997); Kyle and Xiong (2001) Slow Moving Capital - Froot and O Connell (2008); Duffie (2010) Intermediary Asset Pricing - He and Krishnamurthy (2013); Adrian, Etula, and Muir (2014) What s New: Direct Evidence + Level of Segmentation 2. Pricing Credit Risk - credit spread changes hard to explain Collin-Dufresne, Goldstein, and Martin (2001); Ericsson, Jacobs, and Oviedo (2009); Chen, Cui, He, and Milbradt (2014) What s New: Data suggests limited capital models useful for credit risk 3. OTC Derivatives - we don t know much Oehmke and Zawadowski (2013); Atkeson, Eisfeldt, and Weill (2013) What s New: Stylized facts themselves are an important part of the debate 3/20

6 Data and Facts 3/20

7 Detailed Data on CDS Positions Live data from Depository Trust & Clearing Corp (DTCC) Starts in 2010, updated weekly 640+ million positions covering 5700 reference entities, 900 indices, 1700 counterparties Full information on every transaction if: 1. One of counterparties is U.S. registered, OR 2. Underlying reference entity is U.S registered One major contribution = map of the entire U.S. CDS market Capital fluctuations º profits and losses (P&L) of CDS desk 4/20

8 High Concentration, Especially Net Sellers Aggregate Market Share Jul 2010 Aggregate Share of Top 5 Sellers Aggregate Share of Top 5 Buyers Jan 2011 Jul 2011 Jan 2012 Jul 2012 Jan 2013 Jul 2013 Jan /20

9 Persistence in the Identities of Important Players Buyer Persistence (Count) Seller Persistence (Count) Jul 2010 Jan 2011 Jul 2011 Jan 2012 Jul 2012 Jan 2013 Jul 2013 Jan /20

10 CDS Pricing and Seller Capital 6/20

11 Framing the Question The challenge of linking capital fluctuations and spreads Check if capital of buyers/sellers impacts pricing Main focus is sellers, but also consider buyers Simplest way to ask question: How do CDS spreads change with capital fluctuations? Concern: do changes in capital impact prices or vice versa? 7/20

12 Addressing Identification Avoid reverse causality using OCF variables: Outside Capital Fluctuations for buyers and sellers Change in capital on positions taken outside of r s industry For example, explain Ford s CDS spread changes with seller gains/losses on positions taken outside of the auto-industry Let V c, r,t value of c s positions outside of r s industry Then, OCFrt s X Vc, r,t s, where Sr t 1 = r s net sellers c2s r t 1 OCFrt b = same thing for net buyers 8/20

13 Panel Regression Test (Weekly) Panel regression with 5-year CDS spreads: log(cds rt ) = a r + Ø 0 1 Z rt + Ø 0 2 M t + s OCFrt s + bocfrt b + " rt Reference entity controls, a r and Z rt : Firm Fixed Effect Moody s 5 year EDF and Markit s LGD Option-implied CDS spread and ATM Vol Macro-Controls, M t : A lot from theory/literature Can replace with industry time FE 9/20

14 Results: How Much Do Observables Explain? Dep. Variable log(cds rt ) 100 (1) (2) (3) (4) OCF s rt (0.28)** (0.25)** (0.27)** OCF b rt (0.33) (0.34) (0.35) (Firm + Macro Variables) t X X X X Industry Time FE X X Option-Implied Measures X R N 62,476 62,476 64,222 24,420 Observable Firm/Macro Factors Explain 15.7% of Spread Variation 10 / 20

15 Results: Seller Capital Explains Spread Variation Dep. Variable log(cds rt ) 100 (1) (2) (3) (4) OCF s rt (0.28)** (0.25)** (0.27)** OCF b rt (0.33) (0.34) (0.35) (Firm + Macro Variables) t X X X X Industry Time FE X X Option-Implied Measures X R N 62,476 62,476 64,222 24,420 Sellers Capital Fluctuations Account for 10% of Spread Variation 10 / 20

16 Results: Spreads Rise As Seller Capital Declines Dep. Variable log(cds rt ) 100 (1) (2) (3) (4) OCF s rt (0.28)** (0.25)** (0.27)** OCF b rt (0.33) (0.34) (0.35) (Firm + Macro Variables) t X X X X Industry Time FE X X Option-Implied Measures X R N 62,476 62,476 64,222 24,420 $1 bn Seller Loss Raises Spread Levels by 3.1%. æ(spreads)=6% 10 / 20

17 Results: Omitted Variable Problem Is Unlikely Dep. Variable log(cds rt ) 100 (1) (2) (3) (4) OCF s rt (0.28)** (0.25)** (0.27)** OCF b rt (0.33) (0.34) (0.35) (Firm + Macro Variables) t X X X X Industry Time FE X X Option-Implied Measures X R N 62,476 62,476 64,222 24,420 Unlikely to be contaminated by omitted variables 10 / 20

18 Additional Discussion What would you have to believe to invalidate identification? OCFrt s captures a factor that drive Ford s spreads, but in a way that isn t industry specific or captured by equity/options on Ford Implied segmentation? Suppose segmented at industry level: Spreads driven by sellers of firms in the same industry Industry Time FE would absorb this Suggests CDS is segmented at firm level (short horizons) 11 / 20

19 Internal Capital Markets Are Important Dep. Variable log(cds rt ) 100 (1) (2) OCF s rt (0.26)** (Leveragert s ) (0.04)** (0.02)** (Firm Variables) t X X Industry Time FE X X R N 64,006 64,006 Seller Leverage Does Not Drive Out CDS Desk Capital. Suggests Internal Capital Market Frictions Are Important. Robust to Subsamples 12 / 20

20 Effect Dissipates After 8 Weeks log CDS r,t+h 1 CDS r,t 1 = Controls s(h) OCF s rt + b(h) OCF b rt s(h) h (weeks) 13 / 20

21 Additional Ways Market Structure Impacts Pricing 13 / 20

22 Hedge Funds/Asset Managers Now Big Sellers Aggregate Market Share of Selling (%) Dealer HF/Asset Manager 10 Mar 2010 Sep 2010 Mar 2011 Sep 2011 Mar 2012 Sep 2012 Mar 2013 Sep 2013 Mar / 20

23 Institution Type Matters For Pricing Dep. Variable log(cds rt ) 100 OCFrt s (0.33)** OCFrt s HFSs rt (0.59)** (Firm Variables) t X Industry Time FE X R N 64,222 HFS s rt = share of net selling by HFAMs From 2010 to 2014, share of HFAM selling went from 15% to 55% Implies impact of a $1bn seller capital loss on spreads has changed from 2.31 to 3.47 percent over this same time 15 / 20

24 Why Does Concentration Matter? Thought experiment: within a segment, hold the total amount of risk bearing capacity constant But vary distribution of capital across sellers Not obvious from theory whether this should impact price dynamics. So why care about concentration? At least one reason is that concentration creates fragility An idiosyncratic capital shock to a key player doesn t wash out Popular concept in macroeconomics (Gabaix (2011)) 16 / 20

25 Testing the Importance of Concentration Re-construct OCF measures, but take equal-weighted and share-weighted averages SW-OCFrt s is share-weighted change in seller capital EW denotes the equal-weighted counterpart Run the following regression: log(cds rt ) = a r +a it + Ø 0 1 Z rt + SW s + EW s EW-OCFrt s + EW b EW-OCFrt b SW-OCFrt s + SW b SW-OCFrt b Checks whether accounting for concentration is important 17 / 20

26 Concentration Matters Dep. Variable log(cds rt ) 100 Standardized SWA-OCF s rt (0.11)** Standardized EWA-OCF s rt 0.61 (Firm Variables) t Industry Time FE (0.07)** R N 64,222 Capital loss at a large player has a bigger impact on spreads than a capital loss at a smaller player Only a handful of large sellers of default insurance, so their capital position has an outsized impact on pricing X X 18 / 20

27 Final Thoughts 18 / 20

28 What Else is in the Paper? Other ways to quantify concentration (e.g. market beta) Some alternative ways to emphasize link between concentration and fragility e.g. Volatility 2011 Japanese tsunami as a robustness check: Compare U.S. firms whose sellers were exposed to Japan vs. firms whose sellers were not Spreads rise for firms whose sellers were highly exposed Effect is not present in equities Additional evidence that distribution of capital shocks matters 19 / 20

29 Big Picture Conclusion Short-run capital market segmentation in CDS markets: Means capital of a few trading desks drives pricing High degree of segmentation (reference entity level) Internal capital markets also matter Distribution of limited capital is also important: Type of institution determines price impact of capital Concentration creates fragility Distribution effects known in banking, newer to capital markets 20 / 20

30 Thank You! 20 / 20

31 Tobias Adrian, Erkko Etula, and Tyler Muir. Financial intermediaries and the cross-section of asset returns. The Journal of Finance, page Forthcoming, Andrew Atkeson, Andrea Eisfeldt, and Pierre-Olivier Weill. The market for otc derivatives. Working Paper, url: Hui Chen, Rui Cui, Zhiguo He, and Konstantin Milbradt. Quantifying liquidity and default risks of corporate bonds over the business cycle. Working Paper, url: huichen/bondliquidity.pdf, Pierre Collin-Dufresne, Robert S. Goldstein, and J. Spencer Martin. The determinants of credit spread changes. The Journal of Finance, 56 (6): , ISSN doi: / URL Darrell Duffie. Presidential address: Asset price dynamics with slow-moving capital. The Journal of Finance, 65(4): , Jan Ericsson, Kris Jacobs, and Rodolfo Oviedo. The determinants of credit default swap premia. Journal of Financial and Quantitative Analysis, 44(01): , Kenneth Froot and Paul O Connell. On the pricing of intermediated risks: Theory and application to catastrophe reinsurance. Special Issue on Dynamics of Insurance Markets: Structure, Conduct, and Performance in the 21st Century Journal of Banking and Finance, 32 (1):69 85, Xavier Gabaix. The granular origins of aggregate fluctuations. Econometrica, 79(3): , ISSN doi: /ECTA8769. URL Zhiguo He and Arvind Krishnamurthy. Intermediary asset pricing. American Economic Review, 103(2):732 70, doi: /aer URL Albert S. Kyle and Wei Xiong. Contagion as a wealth effect. The Journal of Finance, 56(4): , Martin Oehmke and Adam Zawadowski. The anatomy of the cds market. SSRN: Andrei Shleifer and Robert W Vishny. The limits of arbitrage. The Journal of Finance, 52(1):35 55, / 20