Liquidity Risk and Bank Stock Returns Yasser Boualam (UNC) Anna Cororaton (UPenn) June 16, 2017 1 / 20
Motivation Recent financial crisis has highlighted liquidity mismatch on bank balance sheets Run on repo markets, freeze of interbank lending Significant spillovers to real economy 2 / 20
Motivation Recent financial crisis has highlighted liquidity mismatch on bank balance sheets Run on repo markets, freeze of interbank lending Significant spillovers to real economy Risk coming from liquidity and maturity transformation is inherent to a bank s business model Short-term/liquid liabilities long-term/illiquid assets 2 / 20
Motivation Recent financial crisis has highlighted liquidity mismatch on bank balance sheets Run on repo markets, freeze of interbank lending Significant spillovers to real economy Risk coming from liquidity and maturity transformation is inherent to a bank s business model Short-term/liquid liabilities long-term/illiquid assets Current literature: Literature on financial fragility stemming from liquidity risk is mainly theoretical: Diamond and Dybvig (1983) But empirical research looking at modern institutions is relatively new: Berger and Bouwman (2009), Krishnamurthy et al (2016) However, little is known about how market participants perceive liquidity risk 2 / 20
This Paper Takes an empirical asset-pricing approach 3 / 20
This Paper Takes an empirical asset-pricing approach Research question: Does liquidity risk explain the cross-section of bank stock returns? Do stock returns of banks command risk premia for liquidity risk? 3 / 20
This Paper Takes an empirical asset-pricing approach Research question: Does liquidity risk explain the cross-section of bank stock returns? Do stock returns of banks command risk premia for liquidity risk? Contribution: Empirical banking on liquidity risk Asset pricing for financial firms Policy discussion on Liquidity Coverage Ratio (LCR) under Basel III 3 / 20
Preview of Results Banks with higher exposures to liquidity risk have lower expected returns Long-short portfolio delivers statistically significant alpha of 6 percent annually Results seems to be stronger for small/medium sized banks, and highly leveraged banks Results are overall robust to: Alternative asset pricing specifications Alternative definitions of liquidity mismatch Sample periods Bank characteristics/controls: bank size, leverage, profitability 4 / 20
Outline Describe data and liquidity risk measure Empirical results Discuss potential explanations 5 / 20
Measuring Liquidity Risk Liquidity Gap(LG) = Volatile Liabilities Liquid Assets Total Liabilities 6 / 20
Measuring Liquidity Risk Liquidity Gap(LG) = Volatile Liabilities Liquid Assets Total Liabilities Captures potential inability to service liabilities Higher LG more exposed; lower LG less exposed Calculation depends on availability of data: Use Y-9C reports to the Federal Reserve (quarterly, 1991 Q1-2014 Q4) Consolidated financial statements for bank holding companies Includes balance sheet, income statement, detailed supporting schedules, off balance-sheet items 6 / 20
Measuring Liquidity Risk Liquidity Gap(LG) = Volatile Liabilities Liquid Assets Total Liabilities Volatile liabilities are calculated based on our sample For each type of liability, calculate the standard deviation of the 4-quarter growth rates for each bank over time, then calculate the average standard deviation across banks Rank the different types of liabilities 7 / 20
Definition of Volatile Liabilities Item Std Dev Mean VW Share Trading Liabilities 0.616 0.126 0.044 Other Borrowed Money 0.556 0.101 0.119 Deposits, Foreign 0.537 0.100 0.111 Federal Funds Purchased and Repos 0.516 0.074 0.084 Volatile Liabilities 0.439 0.106 0.327 Other Liabilities 0.414 0.083 0.054 Subordinated Notes and Debentures 0.256 0.052 0.024 Non-interest Bearing Deposits, Domestic 0.204 0.095 0.125 Equity 0.150 0.080 0.103 Interest Bearing Deposits, Domestic 0.127 0.071 0.369 Non-volatile Liabilities and Equity 0.112 0.076 0.673 Deposits in non-commercial banks 0.338 0.049 0.026 Deposits in commercial banks 0.120 0.077 0.488 Off-Balance Sheet 0.740-0.068 0.033 This table presents the volatility of flows of major liability categories in the Y-9C Federal Reserve quarterly reports from 1991-2014. (1) is the average of the standard deviation of the cross-section 4-quarter growth rates; (2) is the average 4-quarter growth rates; (3) is the value-weighted share of each liability category. 8 / 20
Measuring Liquidity Risk Liquidity Gap(LG) = Volatile Liabilities Liquid Assets Total Liabilities Liquid assets: Cash and balances due from other institutions including reserves at the central bank All securities and trading assets Federal funds sold and securities purchased under agreements to resell 9 / 20
Aggregate Liquidity Gap (in Bil. $) 1,500 Panel A. Vola+le Liabili+es Less Liquid Assets, Aggregate Bil $ 1,000 500 0-500 - 1,000-1,500-2,000 1991 1994 1997 2000 2003 2006 2009 2012 Without off balance sheet items With off balance sheet items 10 / 20
Data We extend the sample back to 1974 using COMPUSTAT Merge to CRSP monthly stock returns Final Sample: 148,347 BHC-month observations, 1,092 unique BHC s, average of 300 per year Follow Fama and French (1993): sort stocks from July year t-june year t + 1 using LG in December year t 1, and rebalance annually 11 / 20
Factor Regressions Er p,t+1 = E[β f t+1 ] CAPM Fama-French 3-factor model 4-factor FF with financial sector size factor (Gandhi and Lustig (2015)) Low (2) (3) (4) High Low-High E[r e p,t+1 ] 0.103 0.083 0.080 0.085 0.061 0.042 (3.40) (2.72) (2.32) (2.52) (1.55) (1.94) CAPM alpha 0.045 0.028 0.016 0.018-0.012 0.057 (2.05) (1.22) (0.67) (0.68) (-0.44) (2.72) 3-factor alpha 0.011-0.008-0.030-0.024-0.059 0.071 (0.62) (-0.39) (-1.46) (-1.01) (-2.64) (3.21) 4-factor alpha 0.024 0.003-0.007 0.007-0.026 0.050 (1.22) (0.15) (-0.31) (0.29) (-1.04) (2.26) N 492 492 492 492 492 492 t statistics in parentheses. p < 0.05, p < 0.01, p < 0.001 12 / 20
Liquidity Risk Anomaly? 6% 4% Annualized alpha, in percent 2% 0% -2% -4% 0020 2040 4060 6080 8000-6% Mean excess return CAPM alpha 3-factor alpha 4-factor alpha 13 / 20
Factor Loadings Regression coefficients for Fama-French + Financial Size (4-factor) model Low (2) (3) (4) High Low-High β M 0.758 0.735 0.799 0.793 0.880-0.121 (15.80) (11.30) (12.42) (10.15) (12.63) (-2.50) β smb 0.331 0.289 0.330 0.206 0.229 0.101 (6.10) (3.78) (5.49) (1.86) (2.89) (1.50) β smbf -0.214-0.198-0.390-0.532-0.574 0.360 (-4.14) (-3.08) (-5.60) (-8.14) (-9.30) (6.06) β hml 0.463 0.520 0.655 0.621 0.698-0.235 (6.82) (8.13) (10.77) (7.21) (8.76) (-3.50) N 492 492 492 492 492 492 t statistics in parentheses. p < 0.05, p < 0.01, p < 0.001 14 / 20
Controlling for Bank Characteristics Do other bank characteristics correlated with LG explain results? Portfolio Lo 2 3 4 Hi Mean Std Dev Balance Sheet Items Assets, Bil. $ 4.36 4.85 12.13 26.69 63.69 20.49 122.65 Deposits/Assets, % 82.20 80.32 78.90 76.04 68.60 77.59 9.19 Equity/Assets, % 9.55 9.55 9.20 9.12 8.74 9.25 2.32 Loans/Assets, % 54.61 63.89 67.91 70.63 71.20 65.45 11.86 Income Statement Items Return on Assets, % 0.94 0.88 0.84 0.71 0.73 0.82 1.56 Charge-offs/TA, % 0.31 0.41 0.44 0.52 0.64 0.46 0.85 Asset Risk Characteristics Tier-1 Capital/Assets, % 9.11 9.12 8.78 8.57 8.23 8.78 2.18 Std Dev of ROA 0.44 0.47 0.50 0.54 0.63 0.51 1.18 Bank Organization Characteristics Complex 0.13 0.17 0.25 0.37 0.49 0.27 0.45 15 / 20
Bank Characteristics: Double Sorting Portfolio Low (2) (3) (4) High Low-High Panel A. Size Small 0.046 0.010 0.013-0.028-0.050 0.096 (2.34) (0.49) (0.60) (-1.07) (-2.25) (4.83) Medium 0.015-0.003-0.005-0.001-0.034 0.048 (0.66) (-0.12) (-0.18) (-0.02) (-1.2) (3.14) Big 0.019 0.008-0.001-0.019-0.029 0.048 (0.91) (0.30) (0.05) (-0.75) (-1.06) (1.98) Panel B. Leverage Low 0.014-0.016-0.017-0.013-0.006 0.020 (0.67) (-0.56) (-0.72) (-0.57) (-0.24) (0.82) Medium 0.028 0.014 0.014 0.003-0.004 0.032 (1.29) (0.60) (0.52) (0.09) (-0.15) (1.43) High 0.040-0.004 0.001-0.028-0.027 0.067 (1.52) (-0.16) (0.02) (-0.84) (-0.86) (2.03) t statistics in parentheses. p < 0.05, p < 0.01, p < 0.001 16 / 20
Bank Characteristics: Double Sorting Underperformance of high liquidity risk stocks considerably stronger for Small to medium size banks High leverage banks LG appears to be unrelated to profitability, non-interest income share, and charge-offs Portfolio Low (2) (3) (4) High Low-High Panel C. Profitability Low 0.048-0.018-0.016 0.014-0.022 0.070 (1.74) (-0.62) (-0.52) (0.46) (-0.66) (2.03) Medium 0.017 0.03-0.039 0.002-0.030 0.047 (0.83) (1.21) (-1.60) (0.06) (-1.20) (1.83) High 0.028-0.006-0.017-0.008-0.026 0.054 (1.41) (-0.18) (-0.72) (0.34) (-0.97) (2.10) t statistics in parentheses. p < 0.05, p < 0.01, p < 0.001 17 / 20
Discussion Robustness With/without off balance sheet items Excluding sample period using projected LG measure: 1991 onwards only Exclude crisis and post-crisis periods Correlation between LS2080 portfolio returns and changes in VIX or TED spread is small Result appears to be counterintuitive as it casts doubt on the notion of a market premium for liquidity risk Robustness 18 / 20
Potential explanations Risk-based, rational explanations: Result analogous to the distress risk puzzle for nonfinancials Endogenous sorting due to heterogeneity in risk management and liquidity distress costs? Mispricing due to imperfect information: Institutions that engage in high liquidity risk activities are highly complex Fed categorizes bank as complex when they have material credit-extending activities, high-risk non-banking activities (such as securities broker/dealer activities, insurance underwriting), large public debt issuance, complex management factors Low (2) (3) (4) High Low-High Not Complex 0.067 0.007 0.075 0.035 0.041 0.026 (1.69) (0.20) (2.00) (1.00) (1.32) (0.80) Complex 0.054 0.035 0.020 0.029 0.020 0.034 (1.84) (1.06) (0.55) (0.86) (0.47) (1.20) t statistics in parentheses 19 / 20
Conclusion Balance sheet measures of liquidity risk explain the cross-section of bank stock returns However, banks who have higher liquidity risk have lower expected stock returns Long-short portfolio delivers statistically significant alpha of 6 percent annually Given LG LCR, investors perception of bank liquidity risk may not be in line with the regulators 20 / 20
Robustness Results are robust to alternative definitions, sample period, ex-dividend returns Low (2) (3) (4) High Low-High Panel A. Use Measure with Off Balance Sheet Items 0.025 0.013-0.005-0.008-0.025 0.050 (1.36) (0.64) (-0.20) (-0.31) (-1.08) (3.11) Panel B. Expanded Volatile Liabilities 0.016 0.021 0.005 0.000-0.031 0.047 (0.83) (1.03) (0.24) (0.01) (-1.46) (2.94) Panel C. Sample period excluding financial crisis (1974-2007) 0.032 0.017 0.000 0.017-0.022 0.054 (1.55) (0.79) (0.00) (0.73) (-0.91) (3.05) Panel D. Using ex-dividend returns -0.012-0.034-0.039-0.034-0.084 0.073 (-0.63) (-1.78) (-1.92) (-1.64) (-3.81) (4.17) t-statistics in parentheses. Standard errors are Newey-West corrected with 6 lags. Back 21 / 20