1. Restricted The Relationship Between Bank Lending Rates, Policy Rates and Bank Funding Costs After the Global Financial Crisis by Anamaria Illes, Marco J Lombardi and Paul Mizen Society for Economic Measurement, July 22-24, 2015 The views expressed are solely those of the author and should not be attributed to the BIS

2. Restricted 2 Key Issues The global financial crisis prompted central banks in many countries to cut short-term policy rates to near zero levels after the Lehman collapse in September 2008 Based on the pre-crisis relationship between bank lending rates and policy rates, it would have been reasonable to expect lending rates to have fallen by similar amounts. But lending rates did not fall as much as policy rates did

3. Restricted 3 Short term lending and policy rates

4. Restricted 4 Long term lending and policy rates

5. Restricted 5 Developing a bank funding cost We construct a weighted average cost of liabilities for banks, which reflects the increase in the cost of funds that they have experienced. It comprises a volume-weighted average of the rates at which banks can obtain finance: where r ijt are the rates on the different component liabilities that the banks use to provide funds, and w ijt are the weights on those rates based on the component share in total liabilities for the banks in each country.

6. Restricted 6 Options for developing WACL Option 1: Outstanding cost of funding Option 2: Marginal cost of funding assuming that they keep composition of the balance sheet unchanged Option 3: Marginal cost of funding Our baseline approach is option 2.

7. Restricted 7 Types of WACL Short-term assuming a maturity of 1-year or less Long-term assuming maturities longer than 1-year, with an average of 5-years Banks fund themselves through:  Deposits  Securities  Central bank operations  Covered bonds  Others..

8. Restricted 8 Weights Based on MFI balance sheets data Weights: short-term WACLWeights: long-term WACL Deposits by MFIs less than 1-year in maturity Deposits by MFIs more than 1-year in maturity Deposits by private non-financial sector less than 1-year maturity Deposits by private non-financial sector more than 1-year maturity Short-term securities other than shares Long-term securities other than shares Central bank operations Covered bonds

9. Restricted 9 Weights of the short term WACL

10. Restricted 10 Weights of the long term WACL

11. Restricted 11 Data description Banks have a substantial deposit base in most countries, over 90 percent of total funding in the short-term and 70 percent of total long-term funding (Table 1). Conventional bonds comprise a small share of short-term funding, while they account for approximately 15-30 percent of funding in the long term. The volume of covered bonds outstanding is not large, but has grown since the crisis. Exceptions are Germany and Spain. Equity issuance is excluded

12. Restricted 12 Adjusted WACL Adjusted stocks-based WACL. Option 2 methodology is adjusted for the long-term WACL calculation by shifting weight of long-term securities to covered bonds.

13. Restricted 13 Interest rates Interest rates used are based on new transactions data. Interest rates: short-term WACLInterest rates: long-term WACL Interbank overnight ratesInterbank 1-year rate Deposit rates with maturities up to 1-year for the private non-financial sector Deposit rates with maturities over 1-year for the private non-financial sector 1-year interest rate swap + 1-year CDS of selected banks 5-year interest rate swap + 5-year CDS of selected banks Central bank policy rateCentral bank policy rates Covered bonds yield to maturity

14. Restricted 14 Bank funding costs and policy rate

15. Restricted 15 Bank funding costs and policy rate

16. Restricted 16 Estimation by pooled mean group (PMG) estimator We take y it, (the lending rate), and x it, (the driver of the lending rate i.e. policy rate or the WACL). Assuming an ARDL model we rewrite as a stacked set of N individual equations relating y it ­ and x it for groups i = 1,2,…N over the time period t=1,2,…T as where Y i = (y i1, … y iT )’, X i = (x i1, … x iT )’, 1 = (1,…, 1)’, e i = (e i1, … e iT )’ are all T x 1 vectors of observations, ones and residual errors, and  is the first difference operator.

17. Restricted 17 Results Data are non-stationary according to Im-Pesaran-Shin tests. Cointegration relationships exist for the full sample for the WACL(stock) and WACL(flow) vs lending rates with a few exceptions. NO cointegration exists for policy rates vs lending rates on the full sample. In the case of policy rates the null hypohtesis of no cointegration is rejected more often PMG estimates imply  100bp ↑ WACL(stocks) funding cost for banks  85-89bp ↑ in lending rates.  long-term lending rates are consistently less responsive than short-term lending rates.

18. Restricted 18 Results – Main messages (continued) Results for WACL(flows) are not statistically different from those reported for WACL(stocks) despite different weights. Taking all 11 countries PMG estimates imply  short-term lending rates and WACL(stocks) relationships are fairly similar across the three groups of countries Hausman test

19. Restricted 19 PMG estimates of pass through – WACL (stock)

20. Restricted 20 PMG estimates of pass through – WACL (flow)

21. Restricted 21 PMG estimates – policy rates

22. Restricted 22 Results – contrast with policy rates Pass-through estimates with policy rates: 100bp ↑ in policy rates  52-96bp ↑ in lending rates. The range of values is wider. In many cases the pass through estimate is lower. There is a structural break in policy rates and cointegration breaks down even when we allow for a structural break. A Hausman test rejects the null of equality of coefficients across countries (rejecting the PMG restriction).

23. Restricted 23 Robustness checks Findings are robust for: Different weighting schemes Different country sub-samples Different times sub-samples

24. Restricted 24 Questions?