Bank funding costs and solvency

Abstract

This paper investigates the relationship between bank funding costs and solvency for a large sample of euro area banks using two proprietary ECB datasets for both wholesale funding costs and deposit rates. In particular, the paper studies the relationship between bank solvency, on the one hand, and senior bond yields, term deposit rates and overnight deposit rates, on the other. The analysis finds a significant negative relationship between bank solvency and the different types of funding costs. It also shows that this relationship is non-linear, namely convex, for senior bond yields and term deposit rates. It also identifies a positive realistic solvency threshold beyond which the effect of an increase in solvency on senior bond yields becomes positive. The paper also finds that senior bond yields are more sensitive to a change in solvency than deposit rates. Among the deposit rates, the interest rates of the overnight deposits are the least sensitive. Banks’ asset quality, profitability and liquidity seem to play only a minor role in driving funding costs while the ECB monetary policy stance, sovereign risk and financial markets uncertainty appear to be material drivers.

KEYWORDS:

• Banks
• solvency
• funding costs

JEL Classifications:

• G15
• G21

Acknowledgements

We thank seminar participants at the European Central Bank, the Bank of England, the 34th Annual Congress of the European Economic Association, the 24th Spring Meeting of Young Economists, the 9th International Conference of the Financial Engineering and Banking Society, the 2019 Singapore Economic Review Conference, the 2021 ASSA, and in particular, an anonymous referee for helpful comments and advice. We, however, are solely responsible for any errors that remain. The findings, views and interpretations expressed herein are those of the authors and should not be attributed to the Bank of England, the Eurosystem, the European Central Bank, its Executive Board, or its management.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Notes

1 While this paper focuses on the relationship between banks’ funding costs and characteristics, it is important to acknowledge that banks generally consider the Weighted Average Cost of Capital (WACC) as a relevant metric when they decide to raise new capital. The WACC includes not only debt funding costs but also the cost of equity.

2 However, the share of total assets covered by the types of liabilities considered in this work exhibits material heterogeneity across the banks in the sample. This feature points to important differences in banks’ funding mix. Moreover, over time the sample is characterized by a clear decreasing trend for the share of wholesale funding over total assets and an increasing trend for the share of deposits over total assets.

3 The analysis carried out in this paper would be useful to inform the calibration of second round price effects. However, to compute second round losses in a stress test featuring a dynamic balance sheet, it would also be necessary to take into account the volume effects and the possible changes in banks’ liability structure. This would not be required in a stress test conducted under a static balance sheet assumption.

4 Bonfim and Santos (2004).

5 Schmitz, Sigmund, and Valderrama (2017).

6 Bank level yields are computed relying on data for both investment grade as well as sub-investment grade bonds. Various types of bonds (floating rate, callable, etc.) are used to build the yield index but neither convertible bonds nor structured notes are employed. The minimum required outstanding amount of a bond to enter the database is EUR 150 mn, all bonds must have a remaining time to maturity of at least one year at the index calculation date, and there must be at least 3 bonds on the market to compute the bank level index since bonds with the highest and the lowest yield are not considered. Further information on the senior bond yields used in this analysis are reported in Table A3.

7 Term deposits refer to deposits with an agreed maturity.

8 The data extracted from the iBoxx dataset comprise a sample of 57 banks of which 30 are in common with the other two sets of data. Additionally, 45 banks are included both in the sample for overnight deposit rates and in the one for senior bond yields, 30 banks are available both for term deposit rates and senior bond yields and finally 69 banks are available both for overnight deposit rates and term deposit rates.

9 Common equity Tier 1 comprises bank’s core capital and includes common shares, stock surpluses resulting from the issue of common shares, retained earnings, common shares issued by subsidiaries and held by third parties, and accumulated other comprehensive income. The common equity Tier 1 ratio is computed as the ratio between common equity Tier 1 and risk weighted assets.

10 The SNL data providing information on bank characteristics were matched with the data on cost of funding from iBoxx and IMIR relying on two sets of bank identifiers, i.e. the Monetary Financial Institution (MFI) identifiers and Legal Entity Identifiers (LEI).

11 Among the many credit ratings provided by different agencies, we rely on the information provided by S&P as, according to ESMA (2017), they have the largest market share in terms of annual turnover generated from credit rating activities and the ratings provided by the other agencies cover a smaller portion of our sample of banks.

12 The models estimated for term deposit rates and overnight deposit rates include both the first and the second lag of the dependent variable.

13 The models estimated for senior bond yields also include as control the remaining time to maturity while the models estimated for term deposit rates also include the average original maturity.

14 A similar empirical strategy is adopted by Aymanns et al. (2016) and Gambacorta and Shin (2018).

15 While we base our empirical strategy on the well-established system GMM estimator, which addresses the potential endogeneity concerns by using internal instruments, another possible empirical strategy would be to estimate a system of simultaneous equations as in Schmitz, Sigmund, and Valderrama (2017) which would allow investigating the two-way interlinkages between funding costs and solvency. We exploit a system GMM estimator due to the challenges related to the identification of reliable exogenous external instruments which are needed to identify the endogenous variables in a simultaneous equation framework. Indeed, in this context, the identification strategy requires finding exogenous sources of variation of bank solvency and funding costs and relying on two-stage-least squares or three-stage-least squares estimators. On the other side, however, the system GMM approach does not allow analysing the two-way interaction between funding cost and solvency.

16 In Appendix A, in Table A1, we report also the long-term effects (and the related statistics of significance) of the CET1 ratio on senior bond yields and term deposit rates respectively. These long-term effects are computed relying on the results of the regressions reported in column 8 in Tables 4 and 5. More specifically, the long-term effects are computed as the ratio of the estimated coefficients of the CET1 ratio over one minus the estimated coefficients of the autoregressive terms. The same results are not reported for overnight deposit rates as the estimated coefficient of the CET1 ratio is not significant in the baseline regression reported in column 8 in Table 6.

17 Using data for the Italian banking system’s costs of funding, Albertazzi et al. (2014) find a similar result, i.e. the sovereign spread (in their work the BTP-Bund spread) has no significant effect on the overnight deposit rates. In line with our results, they also find that there is a positive significant relationship between the sovereign spread, the term deposit rates and the bond yields and the effect of the spread is larger on the bond yields than on the term rates.

18 The turning point can be identified by setting the total effect of CET1 ratio on funding costs to zero, that is, by taking β1 + 2 * β3 *CET 1ratio = 0, and solving for the value of CET1 ratio for which the relationship holds.

19 In Appendix A, in Figure A1, we report also the plots of the non-linear long-term effects of the CET1 ratio on senior bond yields and term deposit rates respectively. These effects are computed relying on the results of the regressions reported in column 1 in Tables 7 and 8. More specifically, the long-term effects are computed as the sum of the estimated coefficients on the CET1 ratio and the first derivatives of the related squared terms over one minus the estimated coefficients of the autoregressive terms. The same results are not reported for overnight deposit rates as the estimated coefficients of the CET1 ratio are not significant in the regression reported in column 1 in Table 9. In Table A2, we also show the joint significance of the non-linear long-term effects and the boundaries of the confidence intervals around the turning points both for senior bond yields and term deposit rates. Overall, it can be noticed that the slope of the non-linear long-term effects is steeper than that of the non-linear short-term effects. This implies that as the CET1 ratio moves away from the turning points, the long- term impacts of a given CET1 ratio on the senior bond yields and term rates are stronger than the short-term impacts.

20 The shadow rate is a concept originally introduced by Black (1995) and is a yardstick which measures the monetary policy stance when nominal interest rates hit the zero lower bound.

21 The BRRD, which was adopted in 2014 (EU Directive 2014/59 of the European Parliament and Council) and intended inter alia to weaken the sovereign-bank nexus, introduced amongst other things the bail-in tool. This tool provides the resolution authority with the possibility to write down and convert into equity some part of the bank’s liabilities to set off banks’ losses and restore regulatory capital. Indeed, under the new BRRD framework, banks’ losses will first be borne by those that have invested in the risk capital (shareholders) and then by those who have financed the bank. The BRRD sets a creditor hierarchy of liabilities that fall within the bail-in scope. Equity instruments are the first to be called in case of a bank’s distress. If the bail-in of these instruments is not sufficient to cover losses, subordinated debts and senior unsecured debt will be called upon to cover losses. At the bottom of the creditor hierarchy, there are eligible deposits greater than 100,000 euro (i.e. the threshold established by the deposit insurance). For further details see https://eur-lex.europa.eu/legal-content/EN/TXT/?uri=celex%3A32014L0059 .

22 As the new bail-in rules enacted by the BRRD became effective in January 2016 and our analysis covers the period between 2005 and 2017, our BRRD dummy takes a value equal to 1 in 2016 and 2017.