Do Changes in the Timeliness of Loan Loss Recognition Affect Bank Risk Taking?

Transcription

1 Do Changes in the Timeliness of Loan Loss Recognition Affect Bank Risk Taking? Manuel Illueca a, Lars Norden b, and Gregory F. Udell c a Universitat Jaume I b Rotterdam School of Management, Erasmus University c Kelley School of Business, Indiana University SESAMI/MEW Seminar

2 Motivation (I) Risk taking is the core of banks business Clear intertemporal link between bank lending and loan losses Tension between different views of bank regulators vs. accountants Banking regulation: reduce systemic risk micro- prudential considerations macro-prudential considerations intertemporal issues Increasing regulatory discretion (Fair Value) Accounting: Timely loss recognition, realistic image of assets, liabilities, liquidity, profitability and solvency point-in-time perspective, reducing discretion 2

3 Motivation (II) The recent/current crisis has put this issue to the forefront of the academic debate on accounting standards in the banking industry (e.g., Laux & Leuz, JEP 2010) Policy interest in pro-cyclical banking/lending behavior even preceding crisis, e.g., Borio, Furfine and Lowe (2001) Borio (2002) Horvath (2002) Berger and Udell (2004) Kashyap and Stein (2004) 3

4 Motivation (III) Policy interest in countercyclical policy reactions, e.g., Basel III counter-cyclical capital buffers Jokipii and Milne (2008) Basel Committee on Banking Supervision (2011) LTV ratios Stein (1995) IMF (2011) Almeida, Campello, and Liu (2006) Goodhart, Kashyap, Tsomocos and Vardoulakis (2012) Ono, Uchida, Udell and Uesugi (2013) Dynamic loan loss provisioning, e.g., Saurina (2009) 4

5 Motivation (III) Policy interest in countercyclical policy reactions, e.g., Basel III counter-cyclical capital buffers Jokipii and Milne (2008) Basel Committee on Banking Supervision (2011) LTV ratios Stein (1995) IMF (2011) Almeida, Campello, and Liu (2006) Goodhart, Kashyap, Tsomocos and Vardoulakis (2012) Ono, Uchida, Udell and Uesugi (2013) Dynamic loan loss provisioning, e.g., Saurina (2009) 5

6 Motivation (IV) Our question: What happens if we require banks to recognize loan losses in a more timely manner? Risk taking? Lending volume and standards? Accounting behavior of bank borrowers? Our laboratory: The compulsory adoption of dynamic loan loss provisions in Spain in ) Large-scale field experiment: Exogenous change in LLP rules imposed on all banks by the Bank of Spain 2) First international attempt to reduce the cyclicality and increase the timeliness of loan loss recognition Theory: balance sheet (Basel III) vs. P&L statement (provisions) Here: Assessment for impairment among 6 broad loan categories with additional LLPs through the P&L statement 3) Banking system with substantial heterogeneity 6

7 The Spanish Context Dynamic provisioning would seem to have inoculated Spain from a banking meltdown including limiting effects in early crisis period (Saurina, 2009), but: Spanish banking system essentially imploded Implosion came in the cajas (or savings bank) sector of the banking system BBVA and Santander substantially immunized with global diversification Cajas 50% of Spanish banking system High risk lending prior to crisis Cajas had very different governance structure Deregulated Political capture Likely reacted differently to dynamic provisioning -- (2013) -- 7

8

9 Spain and the Cajas Cajas at center of bubble Cajas were nearly 50% of the Spanish banking system The cajas are a disaster The majority of the 2010 and 2011 European stress test failures were cajas banks Failed cajas were bailed out and force merged, e.g., Bankia a conglomeration of 7 cajas 43 of 45 cajas (in existence in 2009) restructured (only 2 small ones remain independent) All SBs had to convert to banks Gov t hoped for synergies illusory Continued revelation of massive losses swamp any synergies Estimated total SB losses range as high as 153 billion euros Bankia former board members all prosecuted for fraud Now creates a TBTF problem

10 The Cajas The cajas are a very interesting experiment Governance is state centered (i.e., gov't governed) Significant deregulation: the final removal of branching restrictions in 1988 followed by aggressive expansion Also associated with a change in governance that affected political influence This is a common theme and even applies to the U.S. Branching and product deregulation of the S&L industry Spatial deregulation and product deregulation of credit unions Congress currently debating increasing business lending restrictions (including CRE) from 12.5% to 27.5% of assets!

11 The Cajas and Deregulation in Spain Pre Crisis History Existed for 100 years Established by local governments churches and/or welfare societies Private foundations with no owners either retain profits, or pay social dividends Purpose: promote savings by middle and working class people provide lending to small businesses from the same city or province

12 The Cajas and Deregulation in Spain II Governance of Cajas Cajas initially muncipal level institutions In 1985 national legislation gave control to: Depositors (44%) Local governments (40%) Founders (11%) Employees (5%)

13 The Cajas and Deregulation in Spain III Deregulation In 1975 spatial scope extended to provincial level After 1975, extended to regional level usually at the expense of local government s voting rights Although all stakeholders were represented in the board, not all of them had the ability to influence the bank s management Then in 1988 extended nationwide Expansion (Figs 1 & 2) Explosive growth in lending to real estate and construction firms

14 17 regions, 52 provinces

15 Deregulation Led to Shift in Governance Depositors Local governments Founders Employees Depositors Regional governments Local governments Founders Employees

16 Deregulation Led to Shift in Governance Diffuse Depositors Local governments Founders Employees Consolidation of control: for 50% of SBs, one regional government had a stake on average of 20% Depositors Regional governments Local governments Founders Employees

17 Cajas Branches

18 Cajas Lending to Constr. and RE firms

19 (2013) (Cajas Paper) Key Results After deregulation, cajas expanded rapidly Expansion was associated with a significant increase in risk in lending portfolios Weak governance played a role Facilitated by political influence Increased political influence resulted in: higher ex ante risk higher ex post risk (i.e., default)

20 Results Ex Ante Risk Taking Analyze firms that start borrowing from SBs from other provinces Focuses on expansion behavior Univariate analysis Ex ante risk greater ZSORE riskier than those that never borrow from SB EQ/TA ratio lower for never group Risk highest when regional gov t has stake in SB and host and home are the same political party Multivariate analysis Confirms uivariate analysis

21

22 Risk: higher for firms borrowing from SB out of home region and higher yet if SB politically influenced

23 (Back to) Dynamic Provisioning: Pre-adoption (old) loan loss provisioning system Establishes a minimum LLP Minimum LLP linked to actually NPL, e.g., 5% of NPL if between 90 days and 179 days 25% of NPL if between 180 days and 359 days, etc. Banks have discretion beyond this The use of the discretion will determine whether bank is a: Smoother Backward looking Forward looking (conditionally conservative) 23

24 The new loan loss provisioning system (I) During the quarter. ALL loans classified into 6 risk categories Risk categories based on historical loan losses Zero risk: Government debt Low risk: Housing Collateral Medium low risk: Other collateral Medium risk: Uncollateralized loans to entrepreneurs Medium high risk: long term consumption goods High risk: Credit cards At the end of the quarter, banks compute incurred loan losses and expected loan losses INCURRED LOAN LOSSES => SPECIFIC LLP) EXPECTED LOAN LOSSES Dependent on specific borrower creditworthiness + current economic conditions IASB+FASB framework 24

25 The new loan loss provisioning system (II) 25

26 The new loan loss provisioning system (II) Completely Mechanical 26

27 The new loan loss provisioning system (III) 0.06 Pre adoption period Post adoption period NPL / Lagged total loans Additional dynamic component of LLP LLP / Lagged total loans No crowding out of specific and general LLP NPL / Lagged total loans [Specific+General LLP] / Lagged total loans Total LLP / Lagged total loans 27

28 The new loan loss provisioning system (IV) Dynamic loan loss provisions increased unconditional accounting conservatism. (news independent-ex ante-recognition of loan impairment, Beaver & Ryan, RAS, 2005) At the expense of reducing conditional accounting conservatism (i.e., forward looking LLP) (news dependent-ex post-recognition of loan impairment) We examine hypothesize that dynamic loan loss provisions created an incentive to reduce borrower screening and monitoring within loan categories leads to increase in ex ante risk-taking & higher default risk 28

29 Related literature (I) Accounting quality and investment efficiency Biddle & Hilary, (TAR, 2006): Firms with higher quality financial reporting exhibit lower investment-cash flow sensitivity Biddle et al (JAE, 2009): High-quality financial reporting reduces under- and over-investment Francis & Martin (JAE, 2010) : Conditional accounting conservatism increases post acquisition operating performance Bushman et al (JBFA, 2011): Effects of accounting conservatism on corporate behavior are higher under weaker investment opportunities Biddle et al (2012): Accounting conservatism reduces bankruptcy risks In the banking industry Laeven & Majnoni (JFI, 2003): Banks delay loan loss recognition and thereby magnify economic cycles Beatty & Liao (JAE, 2011): Timeliness of loan loss recognition and lending behavior in the U.S. Vyas (JAR, 2011): Timeliness of loan loss recognition during the crisis is lower for more leveraged, constrained and less transparent banks in the U.S. Bushman & Williams (JAE, 2012): Cross-country analysis showing that risk taking incentives depend on whether LLPs aim at smoothing or timely loss recognition 29

30 Related literature (II) Regulation/deregulation and bank risk taking Laeven & Levine (JFE, 2009): Cross-country differences in regulation, focus on privately owned listed banks, standard corporate governance mechanisms at work Illueca, Norden & Udell (2011): Focus on the deregulation of state-owned banks in Spain: liberalization & non-standard governance & political influence increase in risk taking undesired outcomes! Dynamic loan loss provisions Saurina (2009; World Bank Note No. 7, July 2009) Peydro, Jimenez, Ongena & Saurina (2012): Positive effects on credit supply and performance of the Spanish corporate sector But: Risk taking and accounting behavior of banks, and borrowers? We don t know 30

31 Implications from Literature (I) IN OUR PAPER: We evaluate the effects of accounting conservatism on bank risk taking According to previous literature, AC should lead to a decrease in lending to more risky clients... BUT 1) AC could also lead to an increase in perceived lending costs, which would eventually increase the interest rates charged to firms Entrepreneurs might engage in riskier activities (Stiglitz and Weiss, 1981), increasing their default probability

32 Implications from Literature (II) IN OUR PAPER: We evaluate the effects of accounting conservatism on bank risk taking According to previous literature, AC should lead to a decrease in lending to more risky clients... BUT 2) Loan loss provisions might be perceived by managers as an increase in capital buffers Managers woud have an incentive to shift towards riskier clients within the same risk category (Koehn and Santomero 1980; Berger, Bouwman, 2010; Allen, Fulghieri and Mehran, 2011)

33 Implications from Literature (III) IN OUR PAPER: We evaluate the effects of accounting conseervatism on bank risk taking According to previous literature, AC should lead to a decrease in lending to more risky clients... BUT ULTIMATELY THIS IS AN EMPIRICAL QUESTION!!!

34 Hypotheses Hypothesis 1: The decrease of conditional accounting conservatism after the adoption of the dynamic provisioning system led to an increase in risk taking Hypothesis 2: The increase in risk taking is more pronounced for commercial banks than savings banks Hypothesis 3: The accounting quality of bank borrowers decreases after the introduction of the dynamic LLPs (earnings management, switch from international to local auditor) 34

35 Effects of Dynamic LLP on Borrower Monitoring Might Differ Across Banks The banks whose LLPs were more related to current NPLs in the pre-adoption period have less incentive to screen and monitor loans in the post-adoption period To identify these banks, we look at the empirical relationship between SPECIFIC LLP and a) current increase in NPL, b) past increase in NPL, and c) current earnings before LLP (EBLLP) The higher the correlation between specific LLP and current increase in NPL (forward-looking banks), the higher the negative effect of dynamic LLP on loan screening and monitoring The higher the correlation between specific LLP and a) current earnings (earnings smoothing banks) or b) past increase in NPL (backward-looking banks), the lower the negative effect of dynamic LLP on loan screening and monitoring 35

36 How to identify the effects of the dynamic LLP on borrower monitoring (II) Three time-series regressions by bank for (58 banks) SMOOTHING: One for a bank whose Adj. R 2 (1) is above median, zero otherwise BACKWARD: One for a bank whose Adj. R 2 (2) Adj. R 2 (1) is above median, zero otherwise FORWARD: One for a bank whose Adj. R 2 (3) Adj. R 2 (2) is above median, zero otherwise (Nichols et al, RAS 2009) Results robust with continuous variables instead of dummies 36

37 The data Large and unique dataset from Spain ( ) Matched bank-borrower data (623,870 bank-firm-year obs.) Firm data: SABI database audited firms, firm year-observations Balance sheet + Income Statement + General information (auditor identification, location, age, ) Matching with bank Key firm variable: Altman Z-Score for private firms (Altman, 2000) Bank data: 58 banks (17 commercial banks + 41 savings banks) BANKSCOPE database: Balance sheet + Income Statement Key bank variables: General+Specific LLP, Dynamic LLP (hand collected) 37

38 Empirical results (I) Hypothesis 1: Model estimated for the period Zscore t = f(zscore t-1, EVENT, EVENT ACC, controls, fixed effects) EVENT: Equals to one after 1999, zero otherwise ACC = {FORWARD, BACKWARD, SMOOTHING} Controls = Bank variables: ROA, Loans to total assets, Equity to total assets, Deposits to total assets fixed effects: unobserved heterogeneity at the bank-firm level 38

39 Empirical results (II) Our findings are consistent with Hypothesis 1 increased unconditional conservatism => more risk taking Predicted signs All clients (1) (2) (3) Z-SCORE ijt Z-SCORE ijt Z-SCORE ijt Z-SCORE ijt *** *** *** (0.000) (0.000) (0.000) EVENT ijt ** *** *** (0.026) (0.000) (0.000) FORWARD EVENT ijt (-) *** (0.000) SMOOTHING EVENT ijt (+) *** (0.000) BACKWARD EVENT ijt (+) *** (0.000) COMM_BANK ijt BANK_LOANTA ijt *** *** *** (0.000) (0.000) (0.000) BANK_DEPTA ijt *** *** *** (0.000) (0.000) (0.000) BANK_EQTA ijt *** *** *** (0.000) (0.000) (0.000) BANK_ROA ijt *** *** *** (0.000) (0.000) (0.000) Firm fixed effects YES YES YES Industry fixed effects NO NO NO Number of observations 324, , ,549 Adjusted R 2 Number of instruments Hansen Test H0: Exogeneity of instruments subset (0.939) 0.42 (0.999) (0.941) 0.40 (0.999) (0.940) 0.42 (0.999) Ex-ante risk taking: Higher decrease in Z- score after adoption 39

40 Empirical results (II) Our findings are consistent with Hypothesis 1 increased unconditional conservatism => more risk taking Predicted signs All clients (1) (2) (3) Z-SCORE ijt Z-SCORE ijt Z-SCORE ijt Z-SCORE ijt *** *** *** (0.000) (0.000) (0.000) EVENT ijt ** *** *** (0.026) (0.000) (0.000) FORWARD EVENT ijt (-) *** (0.000) SMOOTHING EVENT ijt (+) *** (0.000) BACKWARD EVENT ijt (+) *** (0.000) COMM_BANK ijt BANK_LOANTA ijt *** *** *** (0.000) (0.000) (0.000) BANK_DEPTA ijt *** *** *** (0.000) (0.000) (0.000) BANK_EQTA ijt *** *** *** (0.000) (0.000) (0.000) BANK_ROA ijt *** *** *** (0.000) (0.000) (0.000) Firm fixed effects YES YES YES Industry fixed effects NO NO NO Number of observations 324, , ,549 Adjusted R 2 Number of instruments Hansen Test H0: Exogeneity of instruments subset (0.939) 0.42 (0.999) (0.941) 0.40 (0.999) (0.940) 0.42 (0.999) Ex-ante risk taking: Higher decrease in Z- score for borrowers of banks with more forward-looking specific LLP, less for backward looking specific LLP, and lower degree of earnings smoothing Ex-post risk: Results hold for borrowers that went bankrupt during the period

41 Empirical results (III) Hypothesis 2: Model estimated for the period Zscore t = f(zscore t-1, EVENT, COMM_BANK, EVENT ACC, EVENT COMM_BANK, EVENT ACC COMM_BANK, controls, fixed effects) EVENT: Equals to one after 1999, zero otherwise ACC = {FORWARD, BACKWARD, SMOOTHING} COMM_BANK: Dummy variable, equals to one if commercial bank, zero otherwise Controls = Bank variables: ROA, Loans to total assets, Equity to total assets, Deposits to total assets fixed effects: unobserved heterogeneity at the bank-firm level First-difference GMM estimator, based on orthogonal deviations (Arellano-Bover, 1995) 41

42 Empirical results (IV) Our findings are compatible with hypothesis 2 Predicted (1) (2) (3) signs Z-SCORE ijt Z-SCORE ijt Z-SCORE ijt Z-SCORE ijt *** *** *** (0.000) (0.000) (0.000) EVENT ijt *** *** *** (0.000) (0.000) (0.000) COMM_BANK EVENT ijt *** *** (0.000) (0.000) (0.252) FORWARD EVENT ijt *** (0.000) COMM_BANK BACKWARD EVENT ijt (-) *** (0.000) SMOOTHING EVENT ijt (0.401) COMM_BANK SMOOTHING EVENT ijt (+) *** (0.000) BACKWARD EVENT ijt *** (0.005) COMM_BANK BACKWARD EVENT ijt (+) ** (0.025) BANK_LOANTA ijt *** *** *** (0.000) (0.000) (0.000) BANK_DEPTA ijt *** *** *** (0.000) (0.000) (0.000) BANK_EQTA ijt *** *** *** (0.000) (0.000) (0.000) BANK_ROA ijt *** *** *** (0.000) (0.000) (0.000) N 324, , ,549 Number of instruments Hansen Test 1.77 (0.940) 1.75 ( (0.940) Decrease in conditional accounting conservatism: Stronger effect for commercial banks 42

43 Empirical results (V) Further evidence on the decrease in monitoring effort after the adoption of dynamic LLP (hypothesis 3) Lower demand for accounting quality: AQ t = f(event, EVENT ACC, controls, fixed effects) AQ = Dummy variables: {Reporting small losses, Auditor downgrading} EVENT: Equals to one after 1999, zero otherwise ACC = {FORWARD, BACKWARD, SMOOTHING} Controls (1)= Firm variables: Leverage, Growth, Number of bank relationships, ROA, Big Auditor Controls (2)= Bank variables: ROA, Loans to total assets, Equity to total assets, Deposits to total assets fixed effects: unobserved heterogeneity at the bank-firm level We estimate a conditional logit model for each AQ-ACC combination 43

44 Empirical results (VI) The results support Hypothesis 3 Predicted Signs Dep. Var.= SMALL_PROF ijt (1) (2) (3) EVENT ijt *** *** (0.184) (0.000) (0.000) FORWARD x EVENT ijt (+) ** (0.036) SMOOTHING x EVENT ijt (-) (0.170) BACKWARD x EVENT ijt (-) (0.146) FIRM_LEVERAGE ijt ** ** ** (0.027) (0.027) (0.026) FIRM_GROWTH ijt *** *** *** (0.000) (0.000) (0.000) FIRM_NREL ijt (0.837) (0.850) (0.845) FIRM_BIG4 ijt (0.425) (0.414) (0.401) BANK_LOANTA ijt (0.680) (0.699) (0.795) BANK_EQTA ijt (0.414) (0.349) (0.371) BANK_DEPTA ijt *** *** *** (0.001) (0.001) (0.001) BANK_ROA ijt (0.918) (0.806) (0.749) Bank-Firm fixed effects YES YES YES N 30,356 30,356 30,356 Pseudo R Borrowers of banks with more forward looking LLP in the pre-adoption period are more likely to report small profits 44

45 Empirical results (VII) The results support Hypothesis 3 Predicted Signs Dep. Var.= DOWNGRADING ijt (4) (5) (6) EVENT ijt *** *** *** (0.001) (0.000) (0.000) FORWARD x EVENT ijt (+) (0.234) SMOOTHING x EVENT ijt (-) *** (0.010) BACKWARD x EVENT ijt (-) *** (0.000) FIRM_LEVERAGE ijt *** *** *** (0.003) (0.003) (0.003) FIRM_GROWTH ijt (0.407) (0.414) (0.404) FIRM_NREL ijt (0.132) (0.129) (0.124) FIRM_ROA ijt ** ** ** (0.037) (0.038) (0.039) BANK_LOANTA ijt 3.305*** 3.318*** 3.270*** (0.000) (0.000) (0.000) BANK_EQTA ijt 4.966*** 4.391** 4.372** (0.004) (0.011) (0.011) BANK_DEPTA ijt *** *** *** (0.000) (0.000) (0.000) BANK_ROA ijt 59.57*** 57.97*** 57.81*** (0.000) (0.000) (0.000) Bank-Firm fixed effects YES YES YES N 44,591 44,591 44,591 Pseudo R Borrowers of banks which exhibit a) less backward looking LLPs and b) higher degree of earnings smoothing in the preadoption period: more likely to switch from an international auditor (Big4) to a local auditor (Non-Big4) 45

46 Conclusion The introduction of the dynamic LLP system significantly increased the risk taking of banks that exhibit a more timely loan loss recognition in the pre-adoption period (more forward-looking LLPs) less earnings smoothing in the pre-adoption period a weaker link between their LLPs and past economic conditions. Our findings are consistent with reduced incentives for monitoring efforts in the banking sector Especially for banks that exhibit a forward-looking and nonsmoothing financial reporting behavior 46