How Much Should Creditors Worry About Operational Risk? The CDS Spread Reaction to Operational Risk Events

How Much Should Creditors Worry About Operational Risk? The CDS Spread Reaction to Operational Risk Events CFS Research Conference on Operational Risk March 22 nd, 2013 House of Finance, Frankfurt Department of Banking // 1

Motivation and research question Growing (quantitative) interest in operational risk with Basel II but empirical work on operational risk remains relatively scarce At the same time, rating agencies increasingly gave importance to operational risk (Moody s 2003; Ferry 2003; Fitch 2004) Operational risk events can trigger rating actions (Fitch 2008; S&P 2008; Moody s 2008, 2011) If operational losses are relevant for the evaluation of a company s default risk they should be reflected in CDS and bond prices. Changes in credit ratings are highly anticipated, i.e. CDS spreads increase before ratings are adjusted (Norden/Weber 2004) 2

Literature and contribution The literature examining the market reaction to operational losses has so far only looked at stocks (Cummins et al. 2006; Gillet et al. 2010; Sturm 2013) with one exception analyzing bonds (Plunus et al. 2012) Looking at CDS spreads addresses a different question (shareholders vs. debt holders) Looking at CDS allows for an analysis of a different set of banks (i.e. non-listed banks) CDS event studies have several advantages over bond event studies. From a regulatory point of view the effects on the default risk of banks are probably more important than shareholder wealth effects. 3

Hypotheses Hypothesis 1: Operational loss events have an impact on CDS spreads of the bank incurring the loss (i.e., CDS spreads increase). Hypothesis 2: The CDS market reaction differs depending on event characteristic (i.e., relative loss size, loss amount and event type). Hypothesis 3: The CDS market reaction differs depending on company characteristics (i.e., leverage, credit rating, and firm size). 4

Data and sources Data on 99 operational loss events from European financial institution between 01/2004 and 9/2010 (with threshold 1 million Euro) Sources: Loss events: ÖffSchOR (VÖB), LexisNexis CDS spreads: CMA Datavision (via Datastream), Bloomberg Credit ratings: Standard & Poor s, Moody s Balance sheet information: Bankscope For each loss two different event dates are defined: 1. The date of the first news article (first press date) 2. The settlement date (if different from first press date) CDS data: 5 yr senior (MMR) CDS (available starting January 2004) itraxx Europe 5 yr senior (benchmark) 5

Description of Data Summary statistics for the sample of 99 loss events Mean Median Std Dev. Min Max Operational losses (in million Euro) 205.06 45.00 577.63 1.10 4900.00 Total assets (in million Euro)* 1,016,378 987,064 558,714 43,911 2,465,660 Total equity (in million Euro)* 33,093 29,936 17,287 3,846 90,130 Total liabilities (in million Euro)* 983,285 961,169 547,530 40,065 2,375,530 Total liabilities to total assets (%)* 96.31 96.56 1.47 91.24 98.55 RLS 1: Relative loss size (as % of total equity*) 0.82 0.12 2.04 0.00 15.67 RLS 2: Relative loss size (as % of total assets*) 0.03 0.00 0.07 0.00 0.46 Credit rating (from AAA = 1 to BBB = 9) 4.6 4 = AA- 1.4 2 = AA+ 8 = BBB+ * Total assets, total liabilities, total liabilities to total assets of financial institution affected by the loss are reported as of December 31st preceding the date of the initial news article. 6

Description of Data Loss events by business line and event type Business lines Internal fraud External fraud Employment practices and workplace safety Clients, products & business practices Damage to physical assets Business disruption and system failures Execution, delivery & process management Total across event types Corporate finance 5.1% 2.0% 0.0% 9.1% 0.0% 0.0% 0.0% 16.2% Trading and sales 7.1% 3.0% 1.0% 12.1% 0.0% 0.0% 1.0% 24.2% Retail banking 4.0% 2.0% 0.0% 12.1% 0.0% 0.0% 0.0% 18.2% Commercial banking 0.0% 3.0% 0.0% 2.0% 0.0% 0.0% 1.0% 6.1% Payment and settlement 0.0% 0.0% 0.0% 3.0% 0.0% 0.0% 0.0% 3.0% Agency services 0.0% 0.0% 0.0% 0.0% 0.0% 0.0% 0.0% 0.0% Asset management 2.0% 6.1% 0.0% 3.0% 0.0% 0.0% 0.0% 11.1% Retail brokerage 0.0% 0.0% 0.0% 0.0% 0.0% 0.0% 0.0% 0.0% No business line information 9.1% 3.0% 3.0% 5.1% 0.0% 1.0% 0.0% 21.2% Total number of events 27 19 4 46 0 1 2 99 Total across business lines 27.3% 19.2% 4.0% 46.5% 0.0% 1.0% 2.0% 100.0% 7

Methodology Abnormal spread changes ASC it = (CDS it CDS it 1 ) (I it I it 1 ) (as in Hull et. al. 2004; Norden/Weber 2004; Galil/Soffer 2011) Abnormal relative spread changes ARSC it = CDS it CDS it 1 CDS it 1 itraxx t itraxx t 1 itraxx t 1 (as in Micu et al. 2006; Jacobs 2010; Burghof et al. 2012) 8

CASCs (bps) CARSCs (%) The CDS Spread Reaction to Operational Risk Events Univariate Results (first press date) Cumulative abnormal absolute and relative spread changes 4 CASCs - cumulative abnormal spread changes 3 CARSCs - cumulative abnormal relative spread changes 2 1 6.00% 4.00% 2.00% 0-20 -19-18 -17-16 -15-14 -13-12 -11-10 -9-8 -7-6 -5-4 -3-2 -1 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20-1 0.00% -2.00% -2-3 -4.00% -4-6.00% Figure displays cumulative abnormal absolute and relative CDS spread changes from day [-20] to day [+20] around the first press date. 9

CASCs (bps) CARSCs (%) The CDS Spread Reaction to Operational Risk Events Univariate Results (settlement date) Cumulative abnormal absolute and relative spread changes 8 CASCs - cumulative abnormal spread changes 6 CARSCs - cumulative abnormal relative spread changes 4 2 0-20 -19-18 -17-16 -15-14 -13-12 -11-10 -9-8 -7-6 -5-4 -3-2 -1 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20-2 -4-6 -8 15.00% 12.00% 9.00% 6.00% 3.00% 0.00% -3.00% -6.00% -9.00% -12.00% -15.00% Figure displays cumulative abnormal absolute and relative CDS spread changes from day [-20] to day [+20] around the settlement date. 10

Univariate Results (settlement date) Cumulative abnormal absolute and relative spread changes N Window Mean CASC (bps) t-value % ( > 0 ) N Window Mean CARSC (%) t-value % ( > 0 ) 59 (0,0) -0.95-0.85 42.37 59 (0,0) 0.54 0.60 50.85 59 (-1,+1) 3.36 3.35 *** +++ 57.63 59 (-1,+1) 2.58 2.56 ** ++ 62.71 59 (-2,+2) 5.51 3.50 *** +++ 61.02 59 (-2,+2) 3.81 2.82 *** ++ 66.10 59 (-3,+3) 5.65 3.77 *** +++ 67.80 59 (-3,+3) 4.40 3.57 *** +++ 66.10 59 (-5,+5) 6.91 4.46 *** +++ 72.88 59 (-5,+5) 6.77 4.31 *** +++ 66.10 59 (-10,+10) 4.56 2.59 ** +++ 66.10 59 (-10,+10) 8.77 4.90 *** +++ 71.19 59 (0,+1) 2.61 3.00 *** +++ 67.80 59 (0,+1) 1.71 1.91 * ++ 66.10 59 (-1,+2) 5.56 3.48 *** +++ 61.02 59 (-1,+2) 3.21 2.38 ** ++ 64.41 59 (-1,+3) 6.07 3.76 *** +++ 67.80 59 (-1,+3) 3.77 2.66 ** +++ 66.10 59 (-1,+4) 4.85 3.76 *** +++ 67.80 59 (-1,+4) 3.56 2.60 ** +++ 69.49 59 (-1,+5) 5.67 4.37 *** +++ 74.58 59 (-1,+5) 5.26 2.53 *** +++ 76.27 59 (-1,+10) 4.51 2.59 ** ++ 61.02 59 (-1,+10) 5.58 3.05 *** +++ 64.41 This table displays cumulative abnormal spread changes (CASC) and cumulative abnormal relative spread changes (CARSC) for different event windows around the first press date. *** / ** / * (+++ / ++ / +) indicate significance at the 1%-/5%-/10%-level according to the cross-sectional t-test (Wilcoxon signed-rank test). 11

Multivariate Results (settlement date) OLS regressions of CASCs on loss event and firm characteristics RLS 1: Relative loss size (as % of total equity) Internal fraud (yes=1/no=0) External fraud (yes=1/no=0) CPBP (yes=1/no=0) Total assets (in million Euro) Total liabilities to total assets (%) Credit rating Constant Mean CASC Mean CASC Mean CASC Mean CASC Mean CASC (-1,+1) (-2,+2) (-3,+3) (-1,+2) (-1,+3) 1.1592 ** 2.8396 *** 3.1719 *** 2.4538 *** 2.8772 *** (2.37) (3.75) (4.49) (3.07) (3.62) -6.6099-5.6741-4.2353-7.1316-7.1041 (-1.47) (-0.87) (-0.72) (-1.10) (-1.00) 1.5450 6.7157 3.8489 4.7537 1.2477 (0.31) (0.84) (0.50) (0.58) (0.14) -1.2487 1.7852 0.6582-1.5889-3.8451 (-0.27) (0.27) (0.11) (-0.24) (-0.54) 0.0000 0.0000 0.0000 0.0000 0.0000 (-0.43) (-0.20) (0.43) (0.49) (1.28) 1.4912-40.1705 26.8187-83.3798-1.0734 * (0.03) (-0.76) (0.36) (-1.41) (-1.81) -0.7110-1.8389 ** -1.6236 * -1.9845 ** -2.0759 ** (-1.28) (-2.27) (-1.87) (-2.2) (-2.19) 6.9725 49.9865-16.0209 93.5139 117.3952 * (0.15) (0.95) (-0.22) (1.59) (1.96) R 2 0.2234 0.2892 0.2679 0.2477 0.2231 Prob(F -Statistic) 0.0075 0.0014 0.0016 0.0176 0.0169 N 59 59 59 59 59 ***/**/* indicate statistical significance at the 1%-/5%-/10%-level. All regressions are estimated with robust standard errors. 12

Multivariate Results (settlement date) OLS regressions of CARSCs on loss event and firm characteristics RLS 1: Relative loss size (as % of total equity) Internal fraud (yes=1/no=0) External fraud (yes=1/no=0) CPBP (yes=1/no=0) Total assets (in million Euro) Total liabilities to total assets (%) Credit rating Constant Mean CARSC Mean CARSC Mean CARSC Mean CARSC Mean CARSC (-1,+1) (-2,+2) (-3,+3) (-1,+2) (-1,+3) 0.3105 1.8501 * 2.1096 ** 1.3692 * 1.7526 * (0.44) (1.9) (2.28) (1.69) (1.84) -0.1247 ** -0.1436 ** -0.1199 * -0.1240 ** -0.1368 * (-2.40) (-2.10) (-1.83) (-2.16) (-1.91) -0.0552-0.0477-0.0435-0.0449-0.0459 (-1.06) (-0.68) (-0.64) (-0.73) (-0.64) -0.0870-0.0593-0.0673-0.0790-0.1024 (-1.66) (-0.85) (-1.02) (-1.42) (-1.49) 0.0000 0.0000 0.0000 0.0000 0.0000 (-0.36) (-0.57) (0.01) (0.33) (0.83) -0.1074-0.3293 0.3822-0.3412-0.1356 (-0.26) (-0.67) (0.79) (-0.59) (-0.22) 0.0029-0.0126-0.0098-0.0061-0.0035 (0.45) (-1.53) (-1.10) (-0.59) (-0.36) 0.2017 0.4848-0.2257 0.4466 0.2431 (0.49) (1.00) (-0.47) (0.77) (0.39) R 2 0.2270 0.2895 0.2939 0.1723 0.2186 Prob(F -Statistic) 0.0380 0.0041 0.0455 0.0690 0.0731 N 59 59 59 59 59 ***/**/* indicate statistical significance at the 1%-/5%-/10%-level. All regressions are estimated with robust standard errors. 13

Summary of Results 1. Operational loss events have an impact on CDS spreads indicating an increase in risk for the debt position. 2. CDS market reaction is clearly influenced by the (relative) size of losses implying that the effect is not purely reputational. 3. The effect seems to be more pronounced for banks with a good credit rating while fraud events seem to be not particularly harmful. 14

Thank you. 15

backup 16

Description of Data Loss amounts (in million Euro) of loss events by business line and event type Business lines Internal fraud External fraud Employment practices and workplace safety Clients, products & business practices Damage to physical assets Business disruption and system failures Execution, delivery & process management Total across event types Corporate finance 2.5% 1.1% 0.0% 4.0% 0.0% 0.0% 0.0% 7.5% Trading and sales 35.9% 4.2% 0.0% 5.7% 0.0% 0.0% 0.5% 46.2% Retail banking 0.0% 0.1% 0.0% 9.3% 0.0% 0.0% 0.0% 9.4% Commercial banking 0.0% 3.9% 0.0% 0.1% 0.0% 0.0% 0.1% 4.2% Payment and settlement 0.0% 0.0% 0.0% 3.0% 0.0% 0.0% 0.0% 3.0% Agency services 0.0% 0.0% 0.0% 0.0% 0.0% 0.0% 0.0% 0.0% Asset management 0.0% 13.1% 0.0% 1.1% 0.0% 0.0% 0.0% 14.2% Retail brokerage 0.0% 0.0% 0.0% 0.0% 0.0% 0.0% 0.0% 0.0% No business line information 6.4% 0.1% 4.8% 3.1% 0.0% 1.0% 0.0% 15.5% Total loss amount (in million Euro) 9,097 4,568 984 5,322 0 205 125 20,301 Total across business lines 44.8% 22.5% 4.8% 26.2% 0.0% 1.0% 0.6% 100.0% 18

Univariate Results (first press date) Cumulative abnormal absolute and relative spread changes N Window Mean CASC (bps) t -value % ( > 0 ) N Window Mean CARSC (%) t -value % ( > 0 ) 99 (0,0) 0.31 0.73 49.49 99 (0,0) 0.58 1.14 49.49 99 (-1,+1) -0.62-0.94 41.41 99 (-1,+1) 0.06 0.09 49.49 99 (-2,+2) -1.34-1.41 41.41 99 (-2,+2) 0.48 0.47 50.51 99 (-3,+3) 1.45 1.23 50.51 99 (-3,+3) 1.00 0.68 51.52 99 (-5,+5) 2.68 1.58 + 59.18 99 (-5,+5) 3.32 2.14 ** ++ 60.61 99 (-10,+10) 4.86 1.30 51.09 99 (-10,+10) 5.19 2.81 *** +++ 59.60 99 (0,+1) 0.75 0.16 49.49 99 (0,+1) 0.64 1.04 56.57 99 (-1,+2) -1.01-1.21 43.43 99 (-1,+2) 0.76 0.79 53.54 99 (-1,+3) 0.27 0.31 50.51 99 (-1,+3) 0.71 0.62 58.59 99 (-1,+4) 1.45 1.58 48.48 99 (-1,+4) 1.43 1.37 48.48 99 (-1,+5) 2.02 1.75 * + 55.56 99 (-1,+5) 2.47 2.10 ** ++ 60.61 99 (-1,+10) 1.60 0.80 55.21 99 (-1,+10) 1.80 1.32 54.55 This table displays cumulative abnormal spread changes (CASC) and cumulative abnormal relative spread changes (CARSC) for different event windows around the first press date. *** / ** / * (+++ / ++ / +) indicate significance at the 1%-/5%-/10%-level according to the cross-sectional t-test (Wilcoxon signed-rank test). 20

[ ] 23