QE as a response to the Crisis
The Global Economic and Financial Crisis led central banks in advanced economies to adopt a variety of standard and non-standard measures to ease monetary conditions. In the US, the UK, and Japan the centrepiece of these non-standard measures has been large-scale purchases of financial assets, also known as quantitative easing (QE). The Bank of England began its programme of asset purchases financed through the creation of central bank reserves in March 2009. During the first wave of purchases from March 2009 to January 2010, the Bank purchased a cumulative total of £200 billion of medium- to long-term UK government bonds (gilts). In a subsequent wave of purchases that began in November 2011, it bought a further £175 billion of gilts, an overall amount equivalent to nearly 25% of annual GDP.
While the objectives of the Bank’s QE purchases were clear in terms of meeting its price stability mandate, there has been some debate over how the policy would work, i.e., what is the transmission mechanism. Policymakers in the UK and elsewhere have, however, consistently emphasised the role of the so-called portfolio balance channel as a key element in the expected transmission of asset purchases to the rest of the economy (see e.g., Bean 2011 for the UK, and Yellen 2011 for the US). According to this mechanism, purchases of financial assets financed by central bank money increase liquidity and push up asset prices, as those who have sold assets to the central bank rebalance their portfolios into riskier assets. This, then, stimulates expenditure by increasing wealth and lowering borrowing costs for households and companies. In the UK, policymakers were explicit in structuring their purchases with the aim of buying primarily from institutional investors, such as life insurance companies and pension funds (ICPFs), who are the traditional holders of long-term gilts.
The origins of the portfolio balance channel go back to the work of Tobin and others in the 1960s. However, the role of portfolio balance effects in modern macroeconomic theory remains a subject of some controversy, as most models imply that policies such as QE can only work if they change the private sector’s expectations of future policy rates, while the transferring assets between the private and public sectors under QE has no effect on the behaviour of the private sector per se (Woodford 2012). This neutrality result sits awkwardly with the large financial market reactions often observed after QE announcements, but other explanations may be possible and most of the empirical research on the topic has inferred the importance of this channel indirectly from the behaviour of government bond prices/yields and other asset prices (see Joyce et al. 2011 and Breedon et al. 2012 for evidence from the UK), rather than from direct evidence on the behaviour of investors.1 In this column, we address this gap in the literature by describing our recent research (Joyce et al. 2014) on the behaviour of insurance companies and pension funds, using both sectoral and micro-level data on individual institutions.
What is the sectoral evidence?
Figure 1 illustrates cumulative monthly changes in gilt holdings by different investor categories during 2009-2013, a period of strong gilt issuance. It can be seen that the Bank of England was a significant purchaser of gilts. Other monetary financial institutions were also net purchasers of gilts, as was the overseas sector. But, as Figure 1 shows, during this period the non-bank private sector, which includes insurance companies and pension funds, purchased unusually small quantities of gilts, which would be consistent with them selling gilts to the Bank of England. The same picture also emerges from ONS financial accounts data on ICPFs’ annual net investment flows.
Figure 1 Cumulative monthly changes in UK gilt holdings by category of investor
Source: Bank of England.
But the raw data can only tell us so much.
- If QE worked through a portfolio balance channel, then we would expect institutional investors to have reduced their holdings of UK government bonds (gilts) and to have increased their demand for riskier assets relative to what they would otherwise have done.
This requires addressing the difficult issue of the ‘counterfactual’ in the absence of QE.
Counterfactual analysis of investment behaviour
For counterfactual analysis, we need to make some allowance for other factors that may have been relevant in driving investors’ portfolio allocations. At the same time, allowing for the influence of other factors that may have been influenced by QE (e.g. domestic financial conditions) may lead to understating the potential effects of the policy (a switch into riskier assets will be attributed to better financial conditions rather than QE, even though the policy may have been behind the improvement). Following the approach advocated by Pesaran and Smith (2012), we deal with this issue by allowing only for factors that influence portfolio allocations, but at the same time are unaffected by the Bank’s purchases. More concretely, we run regressions explaining net investment by insurance companies and pension funds into different asset classes in terms of gilt issuance, various US financial control variables, and the amount of QE purchases. The size of the QE coefficient provides a metric of the impact of QE. For conciseness, rather than reporting the regressions results (see Joyce et al. 2014 for more details), we report the associated ex-ante and ex-post measures of the impact of QE on insurance companies and pension funds investment behaviour (Figure 2).
Figure 2 Impact of QE on ICPFs, ex-ante and ex-post QE effects, £ million
The ex-ante impact is measured by taking the difference between the model predictions for net investment with and without QE and the ex-post measure is based on comparing the actual (i.e., ex-post) out-turn with what would have been expected by using the model estimated over the pre-crisis period to predict net investment over the QE period. So, in each case a positive impact implies QE led to net investment being higher, and a negative impact implies it led to a lower impact.
- From the figure, it seems clear that both measures suggest that net investment in gilts fell as a result of QE.
- Moreover, there is evidence of rebalancing into corporate bonds, as net investment out-turns and predictions under QE were substantially greater than the model counterfactuals would suggest.
For net investment into equities and cash the pattern is much less clear cut, with the results, if anything, suggesting that net investment in equities was even lower than expected.
What is the micro evidence?
A similar approach can be applied to the micro-level data on individual life insurance companies (using annual data on life insurers provided by SynThesys for 1985-2012) and pension funds (using annual data provided in anonymised form by the Pension Protection Fund for 2005-2010). In this case, panel regressions are used to explain the annual portfolio share of each asset class in terms of a variety controls, including the characteristics of the respective insurance company or pension fund, and QE purchases.
- In each case, the results indicate that QE is associated with lower asset allocations to gilts and higher allocations to corporate bonds, while the results for equities tend to show lower allocations.
The main advantage of using the micro-data is that it enables us to examine how heterogeneous the responses to QE are across different types of institutions. To do this, we interact the QE purchase variable with a range of other individual variables to see whether this increases or reduces the effect. Table 1 summarises some of the results for pension funds from running separate sets of regression, of which each included a different QE interaction term. We show five interactions for:
- Fund risk-appetite (measured by being above or below the cross-sectional median allocation to equities);
- Pension fund size, measured by number of members in the scheme;
- Maturity of pension scheme, measured by the ratio of pensioners to all members;
- Whether the fund is in surplus or deficit, as measured by the funding ratio; and
- Whether the pension scheme was open or closed.
The results in Table 1 indicate that although there is some heterogeneity in the response to QE across different types of funds, there are also many similarities. The main heterogeneities are with respect to index-linked bonds, but reduced allocations to conventional gilts and increased allocations to corporate bonds seem to have been similar across most pension fund types. However, the relative size of the coefficients suggests that the switch out of gilts was more pronounced for those funds that were better funded and for those funds that were younger (with a lower ratio of pensioners).
Table 1 Asset allocation regressions for pension funds: summary of interaction effects with fund characteristics (2005-2010)
Note: Significance level: ***1%; **5%; *10%. Models are estimated using data provided by the Pension Protection Fund.
The evidence seems consistent with the hypothesis that the Bank of England’s QE policy resulted in some portfolio rebalancing behaviour by institutional investors. But it appears that portfolio rebalancing was limited to corporate bonds, with most of the findings suggesting that institutional investors moved out of equities during the period of QE purchases. This does not necessarily imply equity prices were not supported by portfolio reallocation behaviour, still less from QE, as the analysis only considers insurers and pension funds and does not consider the behaviour of other financial institutions such as mutual and hedge funds. The portfolio investment behaviour of these other financial institutions during the Crisis would be worthy of additional research.
Bean, C (2011), “Lessons on unconventional monetary policy from the United Kingdom”, Speech to the US Monetary Policy Forum, New York,
Carpenter, S, Demiralp, S, Ihrig, J and Klee, E (2013), “Analyzing Federal Reserve Asset Purchases: From whom does the Fed buy?” Finance and Economics Discussion Series 2013-32 Federal Reserve Board, Washington, D.C.
Breedon, F, Chadha, J and Meaning, J (2012), “The Financial Market Impact of UK Quantitative Easing”, Oxford Review of Economic Policy, Vol. 28 (4), 702-728.
Goodhart, C A E, and J P Ashworth (2012), “QE: a successful start may be running into diminishing returns”, Oxford Review of Economic Policy, Vol. 28 (4), 640-670.
Joyce, M, Lasaosa, A, Stevens, I and Tong, M (2011), “The financial market impact of quantitative easing in the United Kingdom”, International Journal of Central Banking, Vol. 7 (3), 113-61.
Joyce, M.A.S., Z. Liu, and I Tonks (2014), “Institutional investor portfolio allocation, quantitative easing and the global financial crisis”, Bank of England Working Paper No. 510, September.
Pesaran, H and Smith, R (2012), “Counterfactual Analysis in Macroeconometrics: An Empirical Investigation into the Effects of Quantitative Easing”, CESifo Working Paper Series 3879
Woodford, M (2012), “Methods of policy accommodation at the interest-rate lower bound”, Federal Reserve Bank of Kansas City Jackson Hole Symposium Conference Volume, August.
Yellen, J (2011), “Unconventional Monetary Policy and Central Bank Communications”, Remarks at the U.S. Monetary Policy Forum New York, February 25.
 There are some honourable exceptions. Goodhart and Ashworth (2012), for example, examine recent trends in the UK national accounts data on aggregate net investment behaviour by ICPFs, but they do not attempt to model investor behaviour to form a counterfactual. For the US, Carpenter et al. (2013) model sectoral flow of funds data for the US over the Crisis, but they do not look beyond the aggregate data.