The Run for Safety: Financial Fragility and Deposit Insurance

Size: px

Start display at page:

Download "The Run for Safety: Financial Fragility and Deposit Insurance"

Olivia Carson
5 years ago
Views:

1 The Run for Safety: Financial Fragility and Deposit Insurance Rajkamal Iyer- Imperial College, CEPR Thais Jensen- Univ of Copenhagen Niels Johannesen- Univ of Copenhagen Adam Sheridan- Univ of Copenhagen

2 Motivation Fragility inherent in demand deposit contracts offered by banks (Diamond and Dybvig, 1983; Calomiris and Kahn, 1991). Sudden concerns about safety by uninsured depositors can lead to runs, destabilizing the financial system. In the U.S. 9 trillion dollars are held by households in bank deposits. Deposit Insurance has been used by most countries to quell concerns about safety. In the current crisis, countries around the globe, including U.S., every country in the European Union increased coverage limits.

3 Motivation Very little micro evidence on the effectiveness of increasing deposit insurance limits during a crisis. How does it effect depositor behavior? Does the reaction differ across different categories of banks? Estimate the magnitude of deposits at risk.

4 Questions What is the potential liquidity shocks that can arise due to sudden withdrawals of uninsured deposits? Are there any real effects in terms of lending?

5 Empirical Setting Detailed data on deposit balances for every bank account in Denmark for the period between Individual Characteristics: demographics, relationships with bank, balance sheet etc. Denmark has a similar deposit insurance system like the U.S. Denmark had a huge housing market collapse (20% drop in house prices) and large drop in the GDP of 8%

6 Identification Exogenous change in the deposit insurance threshold. Initial threshold DKK ($50000). Post Lehman-unlimited Insurance. Threshold reduced to DKK ($125000) in 2010 due to a European Union legislation. Left significant fraction of deposits uninsured. Had to reduce deposit insurance threshold in the middle of the sovereign debt crisis. Spreads on sovereign CDS were at their highest, similar to when Lehman had failed.

7 FIGURE 2 - EVOLUTION OF THE UNINSURED SHARE

9 Impossible d afficher l image. FIGURE 1A BUNCHING IN THE DEPOSIT DISTRIBUTION AROUND 750,000 DKK

10 FIGURE 3A BUNCHING IN THE DEPOSIT DISTRIBUTION AROUND 750,000 DKK, SYSTEMIC VS. NON-SYSTEMIC BANKS

11 TABLE 1 SUMMARY STATISTICS, BANK-LEVEL VARIABLES All banks Non-systemic banks Systemic banks Mean SD Mean SD Mean SD Share of total deposits > 750,000 DKK Deposit growth, Lending growth, Total assets ( 000,000s) , , , Equity-to-assets ratio Loans-to-assets ratio Observations

12 Preliminary Evidence Graphical evidence suggests that depositors withdrawing deposits above the insurance threshold. More prominent for non-systemic banks. More pronounced in 2011 (haircut of 40% on uninsured deposits when the 11 th largest bank failed). Moody s downgraded credit ratings of the largest banks, systemic support from high to low. Next, we consider broader ranges to estimate the overall impact on bank liquidity.

13 Estimation Equation Deposits: Range between DKK and DKK Above: Indicator for the range above the deposit insurance threshold of DKK After: Indicator for years 2010 and 2011 Above x After: captures the relative growth in the mass of deposits above the the deposit insurance threshold as compared to the mass below.

14 TABLE 2 REGRESSION RESULTS, DEPOSIT GROWTH AND CHANGES TO THE DEPOSIT INSURANCE LIMIT Log of deposits in 50,000 DKK ranges, from 500,000 1,000,000 DKK Log of deposits in 50,000 DKK ranges, from 50, ,000 DKK (1) (2) (3) (4) Above insurance limit *** *** *** *** (0.0283) (0.0165) (0.0262) (0.0275) After insurance change 0.316*** 0.351*** 0.110*** 0.110*** (0.0393) (0.0385) (0.0199) (0.0204) Above insurance limit After insurance change *** *** *** *** (0.0520) (0.0404) (0.0227) (0.0229) Systemic bank 3.512*** 3.513*** (0.557) (0.592) Equity-to-assets ratio *** *** (2.548) (2.560) Loans-to-assets ratio 3.088*** 3.371*** (0.909) (0.902) Constant 12.96*** 13.99*** 14.42*** 15.53*** (0.839) (0.0151) (0.880) (0.0153) Bank fixed effects No Yes No Yes Observations 4,376 4,376 1,832 1,832 R-squared Sample period 2007: : : :12

15 Interpretation After: Insured deposits increased by 37% after Above x After: Uninsured deposits decreased by 31% percent relative to this trend. Estimates robust to inclusion of Bank fixed effects. Robust to different choice of ranges.

16 TABLE 3 REGRESSION RESULTS, DEPOSIT GROWTH AND CHANGES TO THE DEPOSIT INSURANCE LIMIT: TIMING OF EFFECTS Log of deposits in 50,000 DKK ranges, from 500,000 1,000,000 DKK (1) (2) (3) (4) (5) Above insurance limit *** *** *** *** *** (0.0415) (0.0339) (0.0443) (0.0360) (0.0702) Above insurance limit *** *** *** *** *** (0.0754) (0.0542) (0.0804) (0.0572) (0.187) Above insurance limit Systemic (0.0350) (0.0244) (0.0562) Systemic bank (0.0734) (0.0729) (0.101) Systemic bank (0.0858) (0.0863) (0.148) Above insurance limit Systemic *** 0.185*** 0.203** (0.0466) (0.0389) (0.0718) Above insurance limit Systemic *** 0.343*** 0.466** (0.0870) (0.0663) (0.190) Bank fixed effects No Yes No Yes Yes Observations 4,376 4,376 4,376 4, R-squared

17 Beliefs about Implicit Guarantees 2010: 23% relative decrease above insurance limit. 2011: 39% relative decrease above insurance limit. For non-systemic banks: For Systemic banks 2010: 24% 2010: 8% 2011: 40% 2011: 20% No difference in deposit growth between systemic and non-systemic banks for insured deposits. Similar results if we limit the sample to the largest 12 banks.

18 Are there differences in growth rates across ranges? FIGURE 4A DIFFERENCES-IN-DIFFERENCES ESTIMATES OF THE DEPOSIT SHOCK

19 Are there differences across bank types? Above x Year x Non-systemic FIGURE 4B DIFFERENCES-IN-DIFFERENCES-IN-DIFFERENCES ESTIMATES OF THE DEPOSIT SHOCK

20 Potential concern Double counting: deposits from above the limit get reallocated to another bank below. could inflate the estimated effects. Fixed sample with balance above DKK in Sample just considering new accounts in year t excluding all individuals who held an account over DKK in the year t-1

21 TABLE 4 REGRESSION RESULTS, DECOMPOSITION OF EFFECTS FOR NEW AND EXISTING ACCOUNTS Panel A: Existing Accounts Panel B: New Accounts (1) (2) (3) (4) (5) (6) (7) Above insurance limit *** *** ** *** *** *** *** (0.0267) (0.0286) (0.0918) (0.0553) (0.0594) (0.102) (0.0689) Above insurance limit *** *** ** *** *** *** *** (0.0513) (0.0551) (0.206) (0.0631) (0.0688) (0.141) (0.0744) Above insurance limit Systemic (0.0256) (0.0612) (0.0349) (0.0400) (0.0430) Systemic bank * (0.0371) (0.0835) (0.114) (0.174) (0.272) Systemic bank (0.0545) (0.123) (0.0832) (0.192) (0.159) Above insurance limit Systemic ** 0.256** (0.0361) (0.0947) (0.0726) (0.111) (0.127) Above insurance limit Systemic *** 0.439* 0.235*** 0.493*** 0.333** (0.0675) (0.210) (0.0818) (0.148) (0.143) Bank fixed effects Yes Yes Yes Yes Yes Yes Yes Observations 4,232 4, ,949 3, ,588 R-squared

22 Results For the largest 12 banks the magnitudes similar to those estimated before. Fixed accounts New accounts 2010: 23% 2010: 25% 2011: 43% 2011: 51% Similar magnitudes using different variation. New money flowing into the banking system also is placed just below the threshold deposit insurance allows the banking system to create safe assets.

23 FIGURE X CUMULATIVE NEW ACCOUNT DEPOSIT MASS, SYSTEMIC VS. NON- SYSTEMIC BANKS

24 Do banks experience funding shocks and real effects? Banks lose uninsured but gain insured deposits from reallocations from other banks. Share of deposits above DKK in 2007 (before the shock). Aggregate growth in deposits between 2007 and 2011 at the bank level. Lending growth between 2007 and 2011 at the bank level.

25 TABLE 5 REGRESSION RESULTS, BANK DEPOSIT GROWTH, LENDING GROWTH AND EXPOSURE TO THE INSURANCE CHANGE Deposit growth, Lending growth, (1) (2) (3) (4) (5) (6) Instrumented deposit growth 0.381*** 0.344** (0.137) (0.165) Share of deposits over 750,000 DKK *** *** ** * (0.412) (0.374) (0.222) (0.283) Systemic bank (0.181) (0.126) (0.157) Equity-to-debt ratio (0.989) (0.598) (0.615) Loans-to-assets ratio * (0.301) (0.224) (0.257) Constant 0.576*** *** (0.0764) (0.341) (0.0607) (0.228) (0.0377) (0.221) Observations R-squared

26 Results so far Up to 50% of uninsured deposits get withdrawn for non-systemic banks. Differential response for systemic and non-systemic banks. Significant liquidity shocks which affect lending. Are there any differences in bank liquidity risk arising due to differences in depositor characteristics? Prior work: not during a systemic crisis. Balance sheets of individuals. Demographics (education, wealth etc..), Relationships Regulatory implications. Definition of mover: limited to uninsured accounts. Account decreases from above the threshold to below. Amount deposited in another account netting out DKK (accounting for average annual adjustments).

27 FIGURE 7B - AVERAGE NUMBER OF ACCOUNTS BY RANGES OF TOTAL INDIVIDUAL DEPOSIT HOLDINGS

28 FIGURE 7B SHARE OPENING NEW ACCOUNTS BY RANGES OF TOTAL INDIVIDUAL DEPOSIT HOLDINGS

29 TABLE 6, PART 1 SUMMARY STATISTICS, INDIVIDUAL AND ACCOUNT-LEVEL CHARACTERISTICS OF MOVERS AND NON-MOVERS Movers Non-Movers Mean SD Mean SD Age Female Married Education, short Education, medium Education, long Retired Self employed Unemployment, 24m After-tax income ( 000s) SD of income ( 000s) Bank debt ( 000s) , Total debt ( 000s) , , , Debt-to-income ratio Observations 1,280 8,405

30 TABLE 6, PART 2 SUMMARY STATISTICS, INDIVIDUAL AND ACCOUNT-LEVEL CHARACTERISTICS OF MOVERS AND NON-MOVERS Movers Non-movers Mean SD Mean SD Total assets ( 000s) 3, , , , Percentage change in assets Stock market participation Value of stock holdings ( 000s) , Loan linkage Account age Systemic bank Account balance ( 000s): , , , , , Total deposits ( 000s): , , , , , , , , , , , , Observations 1,280 8,405

31 Movers versus Non-Movers When people move they often they go below the threshold and not just up to the limit. Liquidity coverage ratios assume that the amount at risk is only up the uninsured amount. Noise: moving for liquidity needs or other reasons (other definitions)

32 TABLE 7 REGRESSION RESULTS, DETERMINANTS OF RESPONSE TO THE CHANGE OF THE DEPOSIT INSURANCE LIMIT (PART 1) Mover definition 1 Mover definition 2 (1) (2) (3) (4) Age *** *** *** *** ( ) ( ) ( ) ( ) Female ( ) ( ) ( ) ( ) Married *** *** *** *** ( ) ( ) ( ) ( ) Education, short *** *** *** ** (0.0137) (0.0138) (0.0138) (0.0147) Education, medium *** *** *** ** (0.0104) (0.0106) (0.0106) (0.0114) Education, long *** *** *** *** (0.0111) (0.0115) (0.0115) (0.0125) Retired (0.0121) (0.0121) (0.0121) (0.0129) Self-employed ** * * ( ) (0.0101) (0.0102) (0.0108) Unemployment, 24m ** ** * ** (0.0401) (0.0403) (0.0401) (0.0428) Debt-to-income ratio *** *** *** *** ( ) ( ) ( ) ( ) After-tax income ( ) ( ) ( ) ( ) SD of income *** *** *** *** ( ) ( ) ( ) ( )

33 TABLE 7 REGRESSION RESULTS, DETERMINANTS OF RESPONSE TO THE CHANGE OF THE DEPOSIT INSURANCE LIMIT (PART 2) Mover definition 1 Mover definition 2 (1) (2) (3) (4) Total assets ** ** ** * ( ) ( ) ( ) ( ) Percentage change in assets ( ) ( ) ( ) (0.0105) Stock market participation (0.0282) (0.0284) (0.0284) (0.0306) Value of stock holdings ( ) ( ) ( ) ( ) Loan linkage ** ** ** *** ( ) ( ) ( ) ( ) Account age *** *** *** ** ( ) ( ) ( ) ( ) Account balance *** *** *** *** ( ) ( ) ( ) ( ) Systemic bank *** *** *** ( ) ( ) (0.0102) Municipality fixed effects No Yes Yes Yes Bank fixed effects No No Yes No Observations 9,113 9,113 9,113 9,113 R-squared

34 Share opening other new accounts in addition to accounts over the limit.

35 Main results More likely to move Educated (proxy for financially sophistication). Income risk: Unemployed in the past. Higher standard deviation of income Self Employed Less likely to move Loan linkages (even after netting off loans) Accounts with systemic banks Debt-to-Income (because they pay down debt) Results controlling for wealth, stock market participation etc. Correlations important for the design of liquidity requirements that take into account depositor characteristics (Basel III)

36 Policy Implications For non-systemic bank run-off rates of up to 50% on uninsured deposits. Aggregate effects Relationship lending Relocation in favor of systemic banks of large deposits. Fraction of uninsured deposits high: Difficult for banks to withstand shocks without guarantees. Transaction costs associated with splitting large sums, funds move to systemic banks. Current liquidity requirements propose a run-off rates on uninsured deposits of 10%. Seem insufficient given the estimated magnitudes. Guarantees of some form seem inevitable.

37 What can be done? Moral hazard concerns, sowing the seeds for the next crisis. Provide guarantees, increase regulation (opt-in). Difficulties with regulation of systemic banks, regulatory capture. Narrow banking (close to having very high liquidity requirements) or breaking up large banks (Cochrane, 2014). Whether illiquid assets can be financed by a less fragile form of financing. Long-term bonds or equity without significantly raising the cost of financing. Important to address these questions to understand the optimal design of the financial system. Less fragile without compromising on liquidity creation.

38 Thank you.

39 FIGURE 3B BUNCHING IN THE DEPOSIT DISTRIBUTION AROUND 300,000 DKK, SYSTEMIC VS. NON-SYSTEMIC BANKS

40 TABLE AX INTEREST RATE REGRESSIONS Imputed interested rate (1) (2) Above insurance limit (0.0842) (0.0968) Above insurance limit ** ** (0.0785) (0.0884) Above insurance limit Systemic (0.0663) Systemic bank (0.2060) Systemic bank (0.1580) Above insurance limit Systemic (0.1090) Above insurance limit Systemic (0.1460) Bank fixed effects Yes Yes Observations R-squared

41 TABLE AX, PART 1 SUMMARY STATISTICS, CHARACTERISTICS OF NEW ACCOUNTS IN RANGES K AND 750K+ Year: Deposit range: k 750k k 750k k 750k k 750k k 750k+ Mean Age Female Married Education, short Education, medium Education, long Retired Self-employed Unemployment 24m After-tax income ( 000s) Observations

42 TABLE AX, PART 2 SUMMARY STATISTICS, CHARACTERISTICS OF NEW ACCOUNTS IN RANGES K AND 750K+ Year: Deposit range: k 750k k 750k k 750k k 750k k 750k+ Mean Bank debt ( 000s) Total debt ( 000s) Total assets ( 000s) Total deposits ( 000s) Debt to Income Deposits to Assets Deposit balance ( 000s) Stock participation Stock holdings ( 000s) Observations

43 TABLE AX, PART 1 SUMMARY STATISTICS, INDIVIDUAL AND ACCOUNT-LEVEL CHARACTERISTICS OF Deposit range: SAMPLE OF ACCOUNTS SPLIT BY 2009 DEPOSIT SIZE 1-300k 300k-500k 500k-750k 750k-1000k 1000k+ Mean Age Female Married Education, short Education, medium Education, long Retired Self employed Unemployment, 24m After-tax income ( 000s) Bank debt ( 000s) Total debt ( 000s) , Observations 520,062 20,614 9,681 3,906 5,914

44 TABLE AX, PART 2 SUMMARY STATISTICS, INDIVIDUAL AND ACCOUNT-LEVEL CHARACTERISTICS OF SAMPLE OF ACCOUNTS SPLIT BY 2009 DEPOSIT SIZE Deposit range: 1-300k 300k-500k 500k-750k 750k-1000k 1000k+ Mean Total assets ( 000s) , , , , Stock market participation Value of stock holdings ( 000s) Loan linkage Account age Systemic bank Total deposits ( 000s): , , , Deposits-to-assets: Observations 520,062 20,614 9,681 3,906 5,914

Understanding Bank Runs: Do Depositors Monitor Banks? Rajkamal Iyer (MIT Sloan), Manju Puri (Duke Fuqua) and Nicholas Ryan (Harvard)

Understanding Bank Runs: Do Depositors Monitor Banks? Rajkamal Iyer (MIT Sloan), Manju Puri (Duke Fuqua) and Nicholas Ryan (Harvard) Bank Runs Bank Runs Bank runs were a prominent feature of the Great