Mahmoud Fatouh
Quantitative easing (QE) involves creating new central bank reserves to fund asset purchases. Deposited in the reserves account of the seller’s bank, these reserves can have implications for banks’ asset mixes. In our paper, we use balance sheet data for 118 UK banks to empirically investigate whether the asset compositions of banks involved in the UK QE operations reacted differently in comparison to banks not involved in the initial rounds of QE between March 2009 and July 2012.
QE and central bank reserves
QE includes the creation of central bank reserves (reserves hereafter), which considerably increases the size of central bank balance sheet. Chart 1 shows that reserves in the UK increased significantly after the introduction of UK QE.
Chart 1: Bank of England liabilities and capital (£million, weekly)
Source: Bank of England
When the Bank of England conducts QE, reserves are credited to the reserves account of the seller’s bank, and that bank then credits the seller’s deposit account with the same amount. Hence, banks involved in QE operations (QE banks) initially receive additional liquidity (as reserves), while other banks (non-QE banks) do not.
However, the distribution of the additional reserves between QE banks and non-QE banks could change. Some of the additional liquidity could ‘leak’ to non-QE banks as the sellers (mostly non-bank financials in the UK) invest their money in other assets, resulting in their deposits moving around the banking system. The magnitude of this ‘leak’ depends on whether those deposits end up with other QE banks or not
Ultimately, we argue that while some of the additional reserves would leak to non-QE banks, most would stay with QE banks. This is because non-bank financials mostly do business with a small subset of banks who are also participants in the Bank’s QE operations. Due to the dynamic nature of reserves and the several factors affecting them, it is hard to isolate the impact of QE on the distribution of additional reserves between QE banks and non-QE banks. Yet, comparing the stock of reserves before and after QE introduction supports our argument (see Chart 2).
Chart 2: Changes in lending and securities – QE banks versus non-QE banks
QE and bank lending
Conventionally, in the presence of reserve requirements, any increase in reserves can potentially increase total credit offered by the banking system by several multiples of the initial increase in reserves (money multiplier effects). This would be the case if there are no other constraining factors such as capital and liquidity requirements (that are more binding than reserve requirements), or low demand for credit. There are no reserve requirements in the UK (ie the money multiplier is undefined), meaning that the supply of credit is mainly driven by banks’ ability and/or incentives to lend.
While the availability of liquidity and capital positions govern banks’ ability to lend, their incentives are largely determined by the regulatory framework in operation especially in downturns. Post-crisis, the additional reserves from QE purchases increased the availability of liquidity which alleviated liquidity constraints. However, banks had depleted capital positions and operated in a regulatory framework that assigns low risk weights to investment in government securities and much higher weights to lending to the real economy. As a result, increased reserves may or may not have led to higher lending, and hence the Monetary Policy Committee did not emphasise the transmission of QE impact via the bank lending channel (BLC).
In other words, QE improves banks’ ability to lend when implemented in a liquidity-scarce environment, but other factors, such as bank capital positions and risk weighting, can affect the impact of QE on bank lending.
What we do
We are interested in understanding the impact of QE, while controlling for other developments that could also affect bank lending, such as the post-crisis Basel III reforms. These reforms tightened capital requirements, requiring weakly capitalised banks to raise additional capital resources. To isolate the impact of QE, we construct a difference in differences (DiD) approach, in which we compare changes in balance sheets of QE banks (treated group) to those of similar non-QE banks (control group). We determine QE banks using a confidential Bank of England’s data set, which shows which banks received reserves through UK-QE operations and the size of the additional reserves. Our sample includes balance sheet data for 118 UK banks from 2000 to 2018.
To draw meaningful conclusions about the role played by QE, we need to assess the (statistical) significance of the differences between the two groups, isolate the impact of other factors, and ensure any effects we observe are specific to the UK-QE period.
It is important to note that our analysis covers the impact of the early QE rounds, ie those until July 2012. The effects of the Brexit and Covid-19 rounds are covered in another paper assessing the interaction between QE and the government lending support schemes during the Covid stress.
QE banks versus non-QE banks
QE banks are on average bigger and hold relatively more securities than non-QE banks. Hence, without any modifications, our results would be prone to selection bias, in the sense that any differences detected between QE banks and non-QE banks could be due to differences in bank characteristics rather than QE treatment. To alleviate the effects of selection bias on the validity of results, we use a propensity score matching methodology to eliminate average differences between the treatment and control groups, and hence improve the validity of our conclusions based on comparisons between the two groups. Propensity scores rely on different bank characteristics, and are used to create a matched set from non-QE banks for each QE bank, based on a certain matching ratio (1:5 in our baseline setup). That is, each QE bank is assigned a matched set that includes five non-QE banks with propensity scores most similar to it. Matching would be successful if it manages to eliminate pre-matching differences between the treatment and control groups. We check that by regressing a treatment status dummy on variables reflecting the size, profitability and balance sheet profile, before and after matching. As Table A illustrates, without matching, QE banks, on average, are larger and hold more securities, and these differences are statistically significant. Post matching, average differences between QE banks and the (matched) control group become statistically insignificant. We run our matching using alternative matching ratios from 1:1 to 1:8, and observe similar patterns.
Table A: Propensity score matching
Notes: Probit regressing the treatment on bank characteristics in 2008h2. The dependent variable is the bank treatment status. The independent variables are size as the natural log of total assets, equity as total assets minus total liabilities, return on assets (ROA), total securities over total assets and net interest income over total assets. Model (1) reports the pre-matching results while model (2) reports the post matching results with matching ratio 1:5. Coefficients and standard errors are reported for each variable. Standard errors are clustered at the bank level and reported in brackets, * p<0.10 ** p<0.05 *** p<0.01.
DiD Results
Our DiD model controls for differences in size (total assets), level of leverage (equity to total assets), profitability (return on assets), and securities to total assets and net interest income to total assets (differences in business models).
Bank lending
Table B presents the treatment coefficients for lending DiD regressions. In line with the average trends in Chart 2, treatment coefficients for all (but one) lending regressions are statistically insignificant. That is, we find no evidence of the alternative bank lending channel (BLC); the additional liquidity did not incentivise QE-bank to increase lending, relative to the control group. There is no evidence suggesting that these results were driven by changes in relative demand for loans the two groups faced.
Table B: Treatment coefficients for DiD lending regressions
Notes: Treatment status (Treatedi) equals to 1 for QE banks and 0 for non-QE banks. Controls are size as log of total assets, equity over total assets, return on assets (ROA), securities over total assets and net interest income over total assets. The reported p-values test the coefficient inequality between QE1 and QE2. Standard errors are clustered at the bank level and reported in brackets, * p<0.10 ** p<0.05 *** p<0.01.
Securities and other assets
Table C shows the DiD coefficients for other bank assets. Relative to the control group, QE banks increased reserves and reduced lending to other banks after QE1. They also increased holdings of government securities, especially after QE2. This suggests that QE banks reallocated their resources from lending towards government securities with low risk weights.
Table C: Treatment coefficients for DiD regressions for other assets
Notes: Treatment status (Treatedi) equals to 1 for QE banks and 0 for non-QE banks. Controls are size as log of total assets, equity over total assets, return on assets (ROA), securities over total assets and net interest income over total assets. The reported p-values test the coefficient inequality between QE1 and QE2. Standard errors are clustered at the bank level and reported in brackets, * p<0.10 ** p<0.05 *** p<0.01.
Conclusion
We test whether the additional reserves created via QE led to an increase in bank credit. We do so by comparing the behaviour of QE banks’ assets, compared to other banks, using a DiD approach.
We find no evidence of transmission via the BLC. We think that the optimisation of regulatory capital motivated QE banks to invest the additional liquidity in high-yield low risk weight sovereigns rather than business loans that attract higher risk weights.
Mahmoud Fatouh works in the Bank’s Prudential Framework division.
If you want to get in touch, please email us at bankunderground@bankofengland.co.uk or leave a comment below.
Comments will only appear once approved by a moderator, and are only published where a full name is supplied. Bank Underground is a blog for Bank of England staff to share views that challenge – or support – prevailing policy orthodoxies. The views expressed here are those of the authors, and are not necessarily those of the Bank of England, or its policy committees.
