February 27, The Development of Securities Markets: Trends, Risks and Policies Università Bocconi

Transcription

1 February 27, 2015 The Development of Securities Markets: Trends, Risks and Policies Università Bocconi

2 Motivation Credit risk is a significant factor in the determination of the market liquidity. At the same time, liquidity concern is closely connected to credit risk of issuers as well as market participants. It is complex relationship. Our study is unique: it presents empirical evidences on the dynamic relationship between credit risk and market liquidity, exploiting the time series evolution of credit risk, rather than cross sectional differences in credit ratings other studies focused on. In addition, we study the effects of the ECB interventions on this relation. The Euro zone sovereign crisis provides us with an unusual laboratory to study the interaction between credit risk and illiquidity.

3 Italian Sovereign Bond market We focus on the Italian sovereign bond market which faces the Euro zone crisis in July Italy has the largest sovereign bond market in the Euro zone (and the third largest in the world after the US and Japan). It has a large number of bond issues with a wide variety of characteristics. Hence, the Italian sovereign bond market is best suited to an in depth analysis of the illiquidity effects of the Euro zone crisis We employ the time series of various liquidity metrics, and CDS spreads as a measure of credit quality, to analyze the liquidity of Italian sovereign bonds during the period from June 1, 2011 to December 31, 2012.

4 ITALIAN AND GERMAN 10-YEAR BOND YIELD AND SPREAD Time-series of the Italian and German bond yields spread for 10-year maturity, the Italian CDS spread and BTP-10Y yield. Our analysis period covers the two highest spikes in the CDS spread and BTP-Bund spread pattern. 2/23

5 EVOLUTION OF THE BID-ASK AND CDS SPREAD Spikes in the quoted bid-ask spread (blue) overlap with spikes in the CDS spread (Red). 3/23

6 EVOLUTION OF THE BID-ASK AND CDS SPREAD Spikes in the quoted bid-ask spread (blue) overlap with spikes in the CDS spread (Red). 3/23

7 Hypothesis H1: Credit risk is a significant factor in the determination of the market (A) Liquidity of Italian sovereign bonds. The dynamic relationship between credit risk and market liquidity Asymmetric information channel Inventory channel Internal and BIS risk management the market makers perception of credit risk = counter party risk (B) The relation is non linear in the creditworthiness of the Government the market makers perception of credit risk = counter party risk H2: The interventions of the central bank affect the dynamic relationship between credit risk and market liquidity. H3: Both global systemic risk factors and the funding liquidity of the primary dealers have an effect on the market liquidity of the bonds, after controlling for credit risk,.

8 Findings: Contribution H1: A strong and dynamic relationship between Italian sovereign credit risk and market liquidity in the secondary sovereign bond market. The change in credit risk leads changes in market liquidity, not vice versa. The relationship is stronger when the CDS spread is above 500 bp. H2: The strength of the relationship diminishes after the announcement of the LTRO by the ECB on December 8, H3: Other global market and funding liquidity risk factors affect market liquidity. Funding costs specific to the primary dealers adversely affects market liquidity

9 Literature The existing literature has highlighted the theoretical relationship between credit risk and market liquidity, as well as that between funding liquidity and market liquidity (see Brunnermeier and Pedersen (2009)) in a general setting. Theoretical Background: Inventory models, Amihud and Mendelson '80, Ho and Stoll '80; Funding liquidity: Brunnermeier and Pedersen '09. Papers on the market liquidity of the US Treasury bond market: Fleming, Remolona '99, Fleming '03, Goyenko, Subrahmanyam, and Ukhov '11 Paper on the liquidity of the corporate bond market: Friewald, Jankowitsch,Subrahmanyam '12, Dick Nielsen, Feldhuetter, and Lando '12 Papers on Euro zone sovereign bonds: Beber, Brandt, Kavajecz '08, Cheung, de Jong, Rindi '05

10 THE DATA GRANULARITY AND MARKET STRUCTURE MTS, Mercato dei Titoli di Stato, is an Electronic, Inter-Dealer market. Large share of interdealer market Dealer to Retail client: not covered. Primary Dealers and Price takers From June 2011: Trade-by-Trade data. Order-by-Order data, uniquely linked to the trades. Every quote, every update, un-netted. Until June 2011: Trade-by-Trade, best 3 quotes prices and quantities. We calculate a series of liquidity measures. Quoted Spread: Best ask-best bid per 100e of face value. Quoted Quantity: Quantity quoted at any level of the bid and ask, in Million e. Lambda: How much a trader would move the best price when trading 15 Me Eective Bid-Ask Spread: 2* Share-weighted average price - relevant best price. 7/23

11 THE DATA BONDS COMPOSITION AND MARKET VOLUME We cover all 152 Italian sovereign bonds between June 11 and December 12 (=406 days). Variable Mean 5th Pct Median 95th Pct Trades Volume(be) Quoted Spread (e/100e) Italian CDS (bps) Median market daily volume is 2 billion e. US treasury market: 500 Billion$. US muni: 15 Billion$. Similar to US securitized xed income and corporate bond markets. We employ several liquidity measures. This presentation: Quoted Spread. CDS spread increased nearly threefold in the sample. 8/23

12 THE DATA BONDS COMPOSITION AND MARKET VOLUME We cover all 152 Italian sovereign bonds between June 11 and December 12 (=406 days). Variable Mean 5th Pct Median 95th Pct Trades Volume(be) Quoted Spread (e/100e) Italian CDS (bps) Median market daily volume is 2 billion e. US treasury market: 500 Billion$. US muni: 15 Billion$. Similar to US securitized xed income and corporate bond markets. We employ several liquidity measures. This presentation: Quoted Spread. CDS spread increased nearly threefold in the sample. 8/23

13 THE DATA BONDS COMPOSITION AND MARKET VOLUME We cover all 152 Italian sovereign bonds between June 11 and December 12 (=406 days). Variable Mean 5th Pct Median 95th Pct Trades Volume(be) Quoted Spread (e/100e) Italian CDS (bps) Median market daily volume is 2 billion e. US treasury market: 500 Billion$. US muni: 15 Billion$. Similar to US securitized xed income and corporate bond markets. We employ several liquidity measures. This presentation: Quoted Spread. CDS spread increased nearly threefold in the sample. 8/23

14 THE DATA BONDS COMPOSITION AND MARKET VOLUME We cover all 152 Italian sovereign bonds between June 11 and December 12 (=406 days). Variable Mean 5th Pct Median 95th Pct Trades Volume(be) Quoted Spread (e/100e) Italian CDS (bps) Median market daily volume is 2 billion e. US treasury market: 500 Billion$. US muni: 15 Billion$. Similar to US securitized xed income and corporate bond markets. We employ several liquidity measures. This presentation: Quoted Spread. CDS spread increased nearly threefold in the sample. 8/23

15 H1: CREDIT RISK AFFECTS LIQUIDITY THE DEVELOPMENT AND TESTING OF THE HYPOTHESIS Credit risk to aect liquidity through two channels: Microstructure: The higher the risk of an asset, the greater its illiquidity Asymmetry of information Inventory models VaR: Internal and external risk control constraints The model in He and Milbradt (2014) shows that credit risk could depend on market liquidity. In order to establish the causality direction(s) in the relationship we perform a Granger-causality analysis: ( QSt ) ( ) ( KQS a111 a 121 = + CDS t K CDS a 211 a ( a113 a 123 a 213 a 223 ) ( ) QSt 3 CDS t 3 We consider log-changes of the variables. ) ( ) ( QSt 1 a112 a CDS t 1 a 212 a 222 ) ( ) QSt 2 CDS t 2 ( ) ( ) ( ) a11p a 12P QSt P ɛqst a 21P a 22P CDS t P ɛ CDSt 9/23

16 H1: CREDIT RISK AFFECTS LIQUIDITY THE GRANGER-CAUSALITY TEST Variable BA t CDS t Intercept BA t *** CDS t *** 0.266*** BA t * CDS t * BA t *** CDS t Granger Causality Tests BA GC CDS CDS GC BA 5.189***. Strong statistical support for one-directional relationship. 10/23

17 H1: CREDIT RISK AFFECTS LIQUIDITY THE IMPULSE RESPONSE FUNCTIONS Impulse Response from BidAskSpread Impulse Response from CDS CDS CDS BidAskSpread BidAskSpread % Bootstrap CI, 5000 runs 95 % Bootstrap CI, 5000 runs 11/23

18 H1: CREDIT RISK AFFECTS LIQUIDITY THE LIQUIDITY REGRESSION Given the VAR results, we decide to focus on the liquidity regression and its dynamics. M N BASpread t = α 0 + α i BASpread t i + β j CDS t j + ɛ t i=1 j=0 Variable Intercept BA t *** *** *** *** CDS t *** *** ** ** CDS t *** *** *** CDS t Adj R Like in the VAR analysis, strongly signicant statistical and economical eect. 12/23

19 H1: CREDIT RISK AFFECTS LIQUIDITY THE LIQUIDITY REGRESSION Given the VAR results, we decide to focus on the liquidity regression and its dynamics. M N BASpread t = α 0 + α i BASpread t i + β j CDS t j + ɛ t i=1 j=0 Variable Intercept BA t *** *** *** *** CDS t *** *** ** ** CDS t *** *** *** CDS t Adj R Like in the VAR analysis, strongly signicant statistical and economical eect. 12/23

20 H1: CREDIT RISK AFFECTS LIQUIDITY THE NONLINEARITY IN THE CDS LEVEL We expect a structural shift in the relation between credit risk and liquidity: Margins in the REPO market depend on credit quality Dierent credit rating means dierent investors Similar to grouping corporate bonds by credit rating We estimate the threshold ˆγ 0 that minimizes the sum of squared residual of the following regression: BASpread t =α 0 + α 1 BASpread t 1 + β 0 CDS t + β 1 CDS t 1 +I [CDS γ 0 ] ( α 0 + α 1 BASpread t 1 + β 0 CDS t + β 1 CDS t 1 ) + ɛ t The test HT (γ, ˆγ 0 ) developed by Hansen (2001) has a pivotal non-standard distribution which allows us to calculate condence intervals for ˆγ 0. We estimate ˆγ 0 and calculate HT (γ, ˆγ 0 ), for the null H 0 : γ = ˆγ 0, which is drawn in the next slide as a function of γ. 13/23

21 H1: CREDIT RISK AFFECTS LIQUIDITY THE NONLINEARITY IN THE CDS LEVEL Hansen Test: Modication of the Likelihood Ratio Test Presence of a threshold at 500bp in the CDS spread: Statistical (see Hansen 2001): bp is the 5% condence interval Investment Grade: 500bp is the CDS level for junk-bonds also for CCP's Margins: LCH.Clearnet documents cite 500 bps. Market participants' comments 14/23

22 H1: CREDIT RISK AFFECTS LIQUIDITY THE NONLINEARITY IN THE CDS LEVEL Hansen Test: Modication of the Likelihood Ratio Test Presence of a threshold at 500bp in the CDS spread: Statistical (see Hansen 2001): bp is the 5% condence interval Investment Grade: 500bp is the CDS level for junk-bonds also for CCP's Margins: LCH.Clearnet documents cite 500 bps. Market participants' comments 14/23

23 H1: CREDIT RISK AFFECTS LIQUIDITY THE NONLINEARITY IN THE CDS LEVEL - TWO SUBSETS CDS t Below 500bp Intercept BA t *** *** *** *** CDS t 0.72 ** ** CDS t *** *** *** CDS t Adj R CDS t Above 500bp Intercept BA t *** *** *** *** CDS t 3.64 *** 3.78 *** *** *** CDS t CDS t Adj R Relationships between changes in the CDS spread and liquidity are dierent below and above 500bp: Larger economic impact: a 10% increase in CDS Five-fold increase Below 500bp: 7% increase in Quoted Spread Above 500bp: 36% increase in Quoted Spread Dierent dynamics Below 500bp: Lagged reaction Above 500bp: Contemporaneous reaction 15/23

24 H1: CREDIT RISK AFFECTS LIQUIDITY THE NONLINEARITY IN THE CDS LEVEL - TWO SUBSETS CDS t Below 500bp Intercept BA t *** *** *** *** CDS t 0.72 ** ** CDS t *** *** *** CDS t Adj R CDS t Above 500bp Intercept BA t *** *** *** *** CDS t 3.64 *** 3.78 *** *** *** CDS t CDS t Adj R Relationships between changes in the CDS spread and liquidity are dierent below and above 500bp: Larger economic impact: a 10% increase in CDS Five-fold increase Below 500bp: 7% increase in Quoted Spread Above 500bp: 36% increase in Quoted Spread Dierent dynamics Below 500bp: Lagged reaction Above 500bp: Contemporaneous reaction 15/23

25 H1: CREDIT RISK AFFECTS LIQUIDITY THE NONLINEARITY IN THE CDS LEVEL - TWO SUBSETS CDS t Below 500bp Intercept BA t *** *** *** *** CDS t 0.72 ** ** CDS t *** *** *** CDS t Adj R CDS t Above 500bp Intercept BA t *** *** *** *** CDS t 3.64 *** 3.78 *** *** *** CDS t CDS t Adj R Relationships between changes in the CDS spread and liquidity are dierent below and above 500bp: Larger economic impact: a 10% increase in CDS Five-fold increase Below 500bp: 7% increase in Quoted Spread Above 500bp: 36% increase in Quoted Spread Dierent dynamics Below 500bp: Lagged reaction Above 500bp: Contemporaneous reaction 15/23

26 H2: CENTRAL BANK INTERVENTIONS AFFECT THE CREDIT RISK - MARKET LIQUIDITY RELATIONSHIP Euro-zone nancial regulators took actions which we expect to aect the credit-liquidity relationship EU bail-outs. ECB Monetary interventions: SMP, LTRO, OMT. Draghi's The ECB is ready to do whatever it takes to preserve the Euro and, believe me, it will be enough. The ECB can aect market liquidity through 3 channels: Purchase of bonds on the market Provide liquidity to the banks Moral Suasion 16/23

27 H2: CENTRAL BANK INTERVENTIONS AFFECT THE CREDIT RISK - MARKET LIQUIDITY RELATIONSHIP Structural break for: BA t = CDS t BA t = CDS t 1 Evidence of structural break on December 8, 2011 Statistical (Chow (1960) test) Announcement date of LTRO I Sample split 17/23

28 H2: CENTRAL BANK INTERVENTIONS AFFECT THE CREDIT RISK - MARKET LIQUIDITY RELATIONSHIP Estimating the same equation for the three subsamples: Before 8/12/2011 Before 8/12/2011 After 8/12/2011 Variable CDS<500bp CDS>500bp Intercept BA t ** *** CDS t *** CDS t *** ** 0.553** Adj R Dierent economic impact: a 10% increase in CDS Below 500: 18% increase in Quoted Spread Above 500: 58% increase in Quoted Spread In 2012: 5% increase in Quoted Spread 18/23

29 H2: CENTRAL BANK INTERVENTIONS AFFECT THE CREDIT RISK - MARKET LIQUIDITY RELATIONSHIP Estimating the same equation for the three subsamples: Before 8/12/2011 Before 8/12/2011 After 8/12/2011 Variable CDS<500bp CDS>500bp Intercept BA t ** *** CDS t *** CDS t *** ** 0.553** Adj R Dierent economic impact: a 10% increase in CDS Below 500: 18% increase in Quoted Spread Above 500: 58% increase in Quoted Spread In 2012: 5% increase in Quoted Spread 18/23

30 H3: MARKET LIQUIDITY IS DRIVEN BY MARKET AND FUNDING RISK FACTORS. Do global factors have an eect on the liquidity of the Italian bond market? Global systemic factors impact liquidity through risk-aversion and inventory concerns: Cost and availability of funding liquidity (Euribor-Eonia, Eonia-DeTBill, CCBSS) Appetite for risk (USVIX) Funding Liquidity: DiEuribor: FLrt,τ = Mi=1 r i,t,τ M ˆr t,τ Can we disentangle market-wide liquidity concerns and market-maker specic liquidity concerns? Brunnermeier and Pedersen (2009) show how funding liquidity aects market liquidity. Higher margins will have market-wide eect on market liquidity. 19/23

31 H3: MARKET LIQUIDITY IS DRIVEN BY MARKET AND FUNDING RISK FACTORS. To test the funding liquidity hypothesis, we replicate the previous regression, including a cohort of macro variables and proceed in a general-to-specic fashion. Variable Below 500, 2011 Above 500, Intercept BA t *** *** BA t *** BA t *** *** CDS t *** CDS t *** ** ** Euribor-Eonia t **. CCBSS t ** *** DiEuribor t **.. Adj R /23

32 ROBUSTNESS CHECKS We verify the robustness of our results through a series of robustness checks: Go Go Go Go Go Other Liquidity Measures: Repeat the analysis with alternative liquidity measures. Yield volatility as driving force: MMS literature indicates volatility as a fundamental determinant of liquidity, we include it in the VAR. Contemporaneous volatility: The eect could be on the contemporaneous volatility vs. liquidity. Unknown break date: The Chow test refers to a specic break date, the supwald is more suitable for unknown break date. Longer time series: Other breakpoints could be present if a longer time series was considered. 21/23

33 THE CONCLUSIONS Strong negative eect of credit risk on market liquidity (with no feedback eect). CDS-Liquidity linkage is stronger and faster when the CDS level rises above 500 bp. The relationship is weaker after the Central Bank intervention. 22/23

34 THE CONCLUSIONS Policy implications: Euro-zone national treasuries: to understand the dynamic nature of the relationship between credit risk, global risk factors and market liquidity, which has strong consequences for the pricing of their issues in the auctions as well as in secondary markets. ECB: to understand the impact of the unconventional instruments of new monetary policy and focus on the market's perceptions of sovereign credit risk. The introduction of the LTRO program, providing short-term liquidity to banks, shows that the channel from bank bailout to sovereign risk could be reversed: oering liquidity to banks may improve the market liquidity of sovereign bonds and also indirectly reduce sovereign risk! Market regulators (the national central banks or European market regulators such as ESMA): to identify the main factors that aect sovereign bonds' market liquidity in the Euro area. Bank regulators: to improve their tools for monitoring both bank capital adequacy and liquidity risk. Changes in bank regulation impact market liquidity, close coordination between regulators may prevent strong negative externalities. 23/23

35 Thank you for your attention!

36 ROBUSTNESS CHECKS OTHER LIQUIDITY MEASURES Back The results are robust to the choice of liquidity variable. Variable All Sample Below 500 Above 500 <500bp >500bp T=2011 T=2012 Dependent Variable: Quoted Quantity, QQ t Intercept QQ t *** *** *** ** ** *** QQ t *** *** * *** QQ t *** ** ** *** CDS t * *** ** CDS t * * R Dependent Variable: Lambda, λ t Intercept λ t *** *** *** *** ** *** λ t *** *** ** *** λ t *** *** *** *** *** λ t ** *** *** *** CDS t *** *** CDS t ** 1.542** ** * R

37 ROBUSTNESS CHECKS OTHER LIQUIDITY MEASURES Back The results are robust to the choice of liquidity variable. Variable All Sample Below 500 Above 500 <500bp >500bp T=2011 T=2012 Dependent Variable: Quoted Quantity, QQ t Intercept QQ t *** *** *** ** ** *** QQ t *** *** * *** QQ t *** ** ** *** CDS t * *** ** CDS t * * R Dependent Variable: Lambda, λ t Intercept λ t *** *** *** *** ** *** λ t *** *** ** *** λ t *** *** *** *** *** λ t ** *** *** *** CDS t *** *** CDS t ** 1.542** ** * R

38 ROBUSTNESS CHECKS RESULTS WITH VOLATILITY Back Only the credit risk Granger-causes the other measures. Variable BA t CDS t σ 2 t Intercept BA t *** ** CDS t *** *** σ 2 t *** BA t ** CDS t * σ 2 t *** BA t * CDS t σ 2 t *** BA t CDS t * σ 2 t *** Granger Causality Tests BA GC CDS + σ CDS GC BA + σ *** σ 2 GC CDS + BA 0.34

39 ROBUSTNESS CHECKS RESULTS WITH VOLATILITY Back The parameters' magnitude is widely unchanged by the inclusion of the contemporaneous yield volatility. Variable Below 500, 2011 Above 500, Intercept CDS t ** 2.82 *** CDS t *** ** *** BA t *** BA t *** BA t *** ** σ 2 t *** *** *** Adj R-Sq

40 ROBUSTNESS CHECKS SUP-WALD STRUCTURAL BREAK Back The supwald test indicates December 8th as breakpoint. A second breakpoint can be found when the CDS crosses the 500bp threshold. F statistics F statistics AveF= 17.05*** P=0.000 Time Time SupF = 23.39** P=0.013

41 ROBUSTNESS CHECKS LONGER TIME SERIES Our results are robust to using a longer time series (July 10-December12): The lag-selection leads the same result. The Granger causality is just as sigicant. We nd the same structural break. Back