1. Dynamic Effects of Idiosyncratic Volatility and Liquidity in Corporate Bond Markets Madhu Kalimipalli Subhankar Nayak M. Fabricio Perez Wilfrid Laurier University Waterloo, Canada CIGI, Oct 3, 2011

2. Agenda 2 Motivation and Introduction Literature and Data ResultsSummary

3. Figure 1 Monthly Industrial portfolio indices based on ratings (1994-2007) 3

4. Figure 2 Monthly Industrial portfolio indices based on maturity (1994-2007) 4

5. Figure 3 Monthly Industrial bond spreads, and aggregate bond and stock market variables (1994-2007 ) 5

6. Research Question  Does idiosyncratic risk subsume information in liquidity in explaining the time-series of corporate bond prices?  We examine this question by studying  the relative impact of idiosyncratic volatility and liquidity on corporate bond yield spreads over time, and  empirically disentangling both effects. 6

7. Objective  To identify principal drivers between volatility and liquidity in determining corporate bond spreads.  Shed more light on the interaction and potential substitution effects between volatility and liquidity on bond pricing over time. 7

8. Why is this important? -I Financial crisis 2007-9  Three phases of the credit crisis:  2006- Jul 2007  Delinquencies in MBS and related instruments  Aug 2007–Aug 2008  Marked by liquidity crisis and lending between banks ceased  Sept 2008-2009  Induced by Lehman crisis demonstrating an inconsistent and haphazard application of government intervention polices,  And evidence of shoddy underwriting and lack of due diligence that further hampers the working of the securitization market

9. Why is this important? –I contd. Credit and liquidity spreads: Subprime crisis (Brunnermeier, 2008) ABCP crisis,Sachsen, Northern Rock Fitch Bear Stearns Funding costs Credit spreads

10. Why is this important?-II 10  Volatility and illiquidity are both critical  An increase in idiosyncratic equity volatility  increases the ex-ante probability of firm default (or “distance-to- default” as in the KMV-Moody’s model),  thereby depressing corporate bond prices and  inflating bond spreads (Merton, 1974) - Campbell and Taksler (2003)  While the corporate debt constitutes a significant proportion of capital structure of firms (over 80%), the underlying market remains highly illiquid  Ignoring non-default sources of risk such as illiquidity can lead to structural models overpricing bonds, and resulting in the so-called “credit puzzle” (Covitz and Downing, 2007; Driessen, 2005 )

11. 11 ? Idiosyncratic volatility Bond market Liquidity Bond spreads Why is this important?-II contd.

12. Why is this important?–II contd.  While higher idiosyncratic equity volatility can imply higher ex- ante bond spreads, it is not obvious from the Campbell and Taksler (2003) results whether higher spreads are attributable to  higher equity volatility,  lower bond liquidity, or both.  Time-series:  Negative firm-specific news events leading to high underlying equity volatility can result in higher bond spreads as well as lower bond liquidity over time. 12

13. Why is this important? -III 13  Debt issuers need to evaluate the cumulative merit of volatility as well as liquidity effects while pricing and timing their bond issues.

14. Why is this important? -IV  Besides, Campbell and Taksler (2003) employ  an aggregate liquidity measure as opposed to individual bond liquidity measures.  aggregate debt turnover (defined as daily average of the total volume of dealer transactions in U.S. Government securities relative to marketable debt)  aggregate bond spreads  The Campbell and Taksler study is limited to the 1995-99 period, when the market experienced very high growth induced by the high-tech bubble. 14

15. Questions we ask 1. Can idiosyncratic volatility and bond liquidity explain the credit spreads?  Collin-Dufresene et al. (JOF, 2001) framework 2. Are volatility and liqudity effects on bond spreads conditional on the underlying regimes? 3. What are the dynamic effects of Idiosyncratic volatility and liquidity on corporate bond spreads?  Granger Casualty and VAR tests/impulse reaction functions 4. What are the information shares of Idiosyncratic volatility and liquidity in explaining corporate bond spreads?  Variance decomposition tests 15

16. Contributions  Our work is unique in that it focuses on a  large sample of corporate bonds over an extended period,  using an exhaustive list of volatility and liquidity variables, and  provides a comprehensive study of the volatility and liquidity effects over time.  Exhaustive data:  We employ about 196,000 secondary trades of option-free corporate bonds issued by 818 firms over the 14-year period, 1994- 2007  In addition, we employ a “ground-up” approach in building portfolio indices  In essence, the critical difference between earlier papers on volatility and liquidity and our work is that  prior papers exclusively focus on either of these variables,  whereas we emphasize joint focus on both variables in bond pricing. 16

17. Agenda 17 Motivation and Introduction Literature and Data ResultsSummary

18. Related Literature  Equity volatility studies include  Campbell and Taksler (2003) Cremers et al. (2008a, b), Alexander and Kaeck (2008) and Zhang et al. (2009).  Corporate bond liquidity studies include  Chen et al. (2007) and Houweling et al. (2005) and Mahanti et al. (2008)  Work on disentangling credit and liquidity risks from yield spreads  For e.g., Longstaff et al., (2005); Driessen, (2005); Covitz and Downing, (2007); Beber et al., (2009); Schwartz, (2010), Kalimipalli and Nayak (2011)  Modeling bond spreads  For e.g. Collin-Dufresne et al. (2001), Avramov et al (2007), Davies (2008)), Van Landschoot (2008) etc  Literature on information spillovers between (a) determination of bond spreads and (b) stocks and corporate bonds 18

19. 19 Data  NAIC corp. bond data:  1994-2007:14 year period spanning 168 months  Trades of insurance companies  FISD data  CRSP and COMPUSTAT  DATASTREAM (swap rates)  The final data file consists of NAIC straight bond trades linked to  respective FISD based issue-specific bond variables and  CRSP stock identifier variables for all firms that have public equity outstanding

20. Table 1 Summary statistics of underlying bonds used in the portfolio construction (1994-2007) 20 Source: NAIC

21. Form portfolios  eight bond portfolios:  two industries (Financials and non-Financials),  two ratings (high and low), and  two maturity (high and low) categories.  equally weighted  Bond Spreads  Underlying portfolio idiosyncratic volatility  Residuals from FF 3 (4) factor models  Underlying portfolio liquidity:  Liquidity factor  Details next slide 21

22. Table 2. Bond Liquidity measures 22 Trade variables 1.Trade size 2.Annual trading frequency Price Impact variables 1.Bond liquidity index 1: 2.Bond liquidity index 2: We employ a factor approach (Bai and Ng, 2002) to combine the four bond liquidity series for each portfolio.

23. Table 3: Summary statistics 23

24. Agenda 24 Motivation and Introduction Literature and Data ResultsSummary

25. Table 5: Monthly Collin-dufresne et al. bond-spread regressions(1994-2007) 25

26. Table 5: contd. 26

27. Sigma-shock Analysis 27  Both liquidity and volatility shocks however have negligible impact on bond spreads (1 and 2 bps respectively) for high- rated bonds.  Overall we find that  volatility has a first-order impact relative to liquidity on bond spreads, especially for low-rated bonds, and  the bond illiquidity significantly matters for pricing low-rated and short-term bonds. AggregateShort -termLow-rated 1 σ shock of volatility 16 bps18 bps23 bps 1 σ liquidity shock 6 bps11 bps9 bps

28. Table 6:Additional variables 28

29. Collin-Dufresne et al.(2001)puzzle 29  Does idiosyncratic volatility and illiquidity explain the systematic variation in bond spread residuals?  Principal component tests show:  that the idiosyncratic volatility and illiquidity variables mainly capture idiosyncratic or portfolio specific information, and  not the systematic variation that can address the Collin- Dufresne et al. (2001) puzzle

30. Figure 4 Underlying regimes in monthly Industrial bond spreads idiosyncratic volatility and bond liquidity (1994-2007) 30

31. Table 7 Monthly time-series regression tests with regime breaks for volatility and liquidity (1994-2007) 31

32. Table 8 Granger casualty tests for monthly bond spreads, volatility and liquidity 32

33. VAR system 33

34. Figure 5 Generalized impulse response functions for monthly credit spread, volatility and liquidity portfolios 34

35. Table 9 Cholesky variance decomposition of volatility and liquidity effects on bond spreads 35

36. Table 9: contd. 36

37. Agenda 37 Motivation and Introduction Literature and Data ResultsSummary

38. Summary of Results: (1) 38  First, both idiosyncratic volatility and liquidity effects jointly and significantly matter only for  the distress portfolios i.e. low rated and short-term bonds;  for other portfolios volatility subsumes liquidity.  The effect of 1 σ shock of volatility (bond illiquidity) on aggregate bond spreads is 16 bps, (6 bps), and increases to 23 bps (9 bps) for low-rated bonds.

39. Summary of Results: (2 & 3) 39  Second, there is a differential impact of volatility and liquidity on bond spreads, conditional on the underlying regime.  For example, the bond illiquidity effects mainly come from high-illiquidity regimes,  while volatility effects are significant even for low-volatility regimes.  Third, Granger-causality tests indicate a strong evidence for volatility and illiquidity influencing bond spreads for all distress portfolios.  Bond Illiquidity seems to exogenously determined, whereas spreads and volatility are mainly endogenously driven.

40. Summary of Results: (4) 40  Fourth, impulse response analysis shows that there is a differential impact between how volatility and liquidity impact bond spreads.  While the liquidity shocks are quickly absorbed into bonds prices, volatility shocks are more persistent and have a long-term effect.  Liquidity shocks can have significant trivial long-term effects for distress portfolios.

41. Summary of Results: (5) 41  Finally, variance decomposition tests show that  liquidity can explain a significant residual variance for the short-term portfolios, and  that effect is concentrated in the high-spread, high-volatility or high-illiquidity regime.  Volatility, however, has the dominant explanatory power for all the portfolios in both regimes, and under different orderings

42. Collectively 42  Our results imply that  the idiosyncratic volatility effect does not subsume the liquidity effect in explaining bond prices for distress portfolios,  unlike the findings in equity markets (Spiegel and Wang, 2005).  Our results, therefore, suggest that  both volatility and liquidity effects are important for bond pricing, and  can have differential short- and long- run impact on bond spreads conditioned on the underlying portfolios and regimes

43. Practical implications 43  Our findings can guide  fixed-income desks in building improved pricing models based on differential impact of volatility and liquidity;  bond traders better formulate their hedging and market timing strategies,  debt issuers better time their debt issuance in order to minimize the cost of borrowing, and  policy makers better address the impact of volatility and liquidity shocks on credit markets.