Impact of Quantitative Easing
By Sheryl Dong, Birce Akay, Daniel Carey and Dillon Oppon-Ferguson
Desired Outcomes of QE
Channels through which QE operates:
Quantitative Easing (QE) was first introduced to the UK in 2009 as a way to combat the depressing effects of the financial crisis in the UK economy, in more ways than just conventional monetary policy. At the time, its main aim was to provide confidence to issuers and investors about selling off their assets because of ineffective policy rates that were at the effective lower-bound, so the bank intervened to plug the gap of the dried up liquidity in the markets between investors and financial institutions. The hope was that the influence of QE would lead investors and issuers to not incur excessive price discounts. However, as the effects of the financial crisis remained, QE has been sustained as a monetary stimulus to the macroeconomy and financial markets, in the hopes of aiding the increase in economic growth and rising prices.
In theory, QE impacts the macroeconomy in the following three ways: the portfolio balance channel, the policy “signalling” channel and the market liquidity channel.
When the Bank of England conducts large-scale asset purchases, they tend to buy longer-term assets from the private sector. In doing so, this causes the supply of shorter-term assets to increase relative to longer-term assets and thus causes a restructuring of the private sector’s portfolio balance. This leads to a fall in relative returns, as the supply of short-term assets increases to the public, bringing forward future consumption by lowering the incentive to invest for the long-run. In some cases, due to the massive quantities of longer-term assets purchased by the BoE, some investors may wish to demand certain assets for specific reasons but by doing so increases the cost to adjust their portfolio. This leads to the determination of asset prices as investors re-adjust their portfolio structure, for the bank to purchase.
The policy “signalling” channel refers to the way in which investors believe that through the actions of asset purchasing by the central bank, they are upholding their policy commitment to sustain low rates for a longer period of time. This implies a fall in expected shorter-term rates and borrowing costs, giving way to increased levels of aggregate spending and thus an expansion of the macroeconomy via rising prices and economic growth.
Lastly, QE affects market liquidity by exchanging private sector assets for liquidity. By purchasing assets, the central bank provides liquidity to financial markets in times of financial tightening; when monetary conditions are somewhat disrupted, the providence of cash from the central bank helps to loosen market conditions. By doing so, financial institutions are more able to maintain/increase their levels of investment leading to further spending in the macroeconomy, leading to similar effects of the other channels.
The desired outcomes of QE can be best represented by figure 1, whereby portfolio rebalancing affects wider asset prices, policy signalling leads to changes in the exchange rate and bank funding costs, and market liquidity influences the yields on gilts (as gilts are the most targeted financial asset in the process of QE). These channels then increase wealth whilst reducing the cost of borrowing, leading incomes to rise, expenditure to increase and the macroeconomy to expand with rising prices.
On balance, the literature provides evidence of ongoing QE transmission to the macroeconomy in later rounds of purchases and is supportive that at least some of the associated transmission channels are persistent. However, this need not suggest that all QE channels are equally powerful and persistent in all states of the world. Unsurprisingly, the evidence is not clear-cut on precisely which QE transmission channels have operated in which circumstances. There are only a relatively small number of observed QE actions, making it hard to identify the overall impacts of those actions and harder still to ascertain the precise channels of transmission. This perhaps explains why different authors have placed differing weights on the relative importance of the main QE transmission channels.
Desired outcomes of QE in the macroeconomy and financial markets:
The UK’s QE portfolio currently amounts to around 30% of nominal UK GDP and that share will continue to rise as the latest round of purchases are completed over the course of this year.
The first stage of the transmission mechanism from asset purchases to the real economy occurs through their impact on financial market prices. A number of empirical studies have looked quantitatively at the impact of QE on a range of asset prices: government bond yields; exchange rates; equity prices; volatility; and corporate bond spreads. A summary is shown below.
Joyce et al (2011) estimate that the Bank of England’s first wave of asset purchases from March 2009 to January 2010, which involved purchasing a cumulative £200 billion of medium- to long-term UK government bonds, led to an average fall in 5 to 25-year gilt yields of about 100 basis points. Looking across the first £375bn of Bank of England QE, Meaning and Warren (2015) estimate that QE reduced yields by around 25bps.
QE announcements are, in general, associated with higher equity prices. The reaction is, however, far from uniform across different QE interventions. For example, in the UK the FTSE index rose following the May, August and November 2009 QE announcements, but fell after the February and March 2009 events. Summing across all events, the FTSE index on average fell by 3%. It is possible any significant positive impact on equity prices may have taken longer to feed through.
In terms of the macroeconomy, work by Bank of England economists suggest that the initial £200bn of QE in the UK may have pushed up on the level of GDP by a peak of 1.5%-2% and on inflation by 0.75%-1.5% (Joyce, Tong and Woods, 2011).
Assessing the Impacts of QE
Estimating the effects of QE with event studies and VARs:
Credibly interpreting QE effects as causal is a challenge because the counterfactual is unobserved and other events (e.g., international events) and policy measures are likely to be present at the same time as QE announcements. Therefore, the “event study” approach is usually applied in the empirical studies, to estimate the initial impact of QE on interest rates, financial market variables and asset prices. (“Normal” variables without QE are estimated based on the event window and they are compared with the “actual” variables, immediate reactions after QE announcements.) However, this approach is less well-suited for persistent effects of QE such as its macroeconomic effects.
To produce estimates of QE’s impact on economic activity and inflation, two methods are favoured: (i) a Two-step procedure and (ii) a Structural Vector Auto-Regression (SVAR). In a two-step procedure, firstly the impact of QE on asset prices is analysed using an event study method and subsequently asset price changes are applied to a macroeconomic model in order to determine the impact on GDP and inflation.
However, this procedure gives a snap-shot measure of the QE’s effect on asset prices, while the effects are observed over an extended period in practice. Thus, some researchers prefer an alternative methodology that is provided by Weale and Wieladek (2016). Introducing asset purchase announcements directly into a SVAR and imposing restrictions with the help of economic theory enable the impact of QE to be identified, though this method is far from perfect as well. Structural breaks around QE announcements should be considered and due to a lack of consensus in identifying QE shocks, different identification schemes may be used. Using these approaches, some empirical evidence of the QE’s macroeconomic effect is found, though estimates are very uncertain; according to Williams (2013), the degree of uncertainty regarding the macroeconomic effects of asset purchases is at least twice as large as that for conventional monetary policies. Hence, the accuracy of the estimates should be carefully assessed.
Another difficulty for researchers is isolating the impact of QE on the behaviour of the domestic non-bank financial sector (e.g. insurance companies and pension funds). Joyce, Liu, and Tonks (2014) adopted a counterfactual analysis of what would happen to investors’ asset allocation in the absence of QE, and ex-ante and ex-post measures were reported. (ex-ante impact: the difference between the predictions for investment with and without QE, ex-post impact: comparing the actual with what would have been expected by the model estimated over pre-crisis period to predict investment over the QE period.)
Distinguishing the effects of QE during normal times and during crises:
State contingency:
QE is state contingent if its transmission mechanisms depend on the state of the economy. The effectiveness of QE would then be contingent on the position of the economy in the business cycle.
Recent research has highlighted the possible state-contingency of Q.E.’s effect. Baily et al. (2020) make three observations on when/how Q.E. is more effective. They emphasise the importance of ‘going big’ and ‘going fast’ at times of significant market dysfunction while also speculating on the significance of sources of liquidity stress in the system.
A correlation between the size of purchases and the effect on interest rates is not controversial in the conventional understanding of Q.E. transmission. As such, the critical part of the first observation is the suggestion that QE is more effective during periods of market dysfunction (in the form of liquidity and/or confidence shocks, likely as it is the two are intertwined). The originality of Baily et al’s analysis lies in the suggestion that violations of the EMH, defined as (i) constraints on arbitrage (i.e. liquidity) and (ii) an increase in risk aversion (of investors) are increasing with respect to the level of market dysfunction, from which it follows transmission through the portfolio balance channel is state-contingent.
It is less arduous to understand the speculated dependencies of other channels. Central to the state contingency of the market liquidity mechanism is liquidity premia on private assets. During crises, such as the GFC and March 2020, liquidity can dry up, causing these premia to increase. Hence the effect of QE in increasing liquidity, decreasing premia, and cutting the relative price of illiquid assets to liquid ones, is strongest during dysfunctional periods (crises) when liquidity is shortest. As for the signalling effect, it is amplified whenever monetary policy is less accurately predicted by market participants. Baily et al. highlight QE1 as an example when the novelty of the policy contributed to its effect. D’Amico and Kaminska (2020) also provide some evidence for a state-dependent market liquidity channel. During periods of market distress (and not in stable times) government bond purchases have lowered corporate yields more than gilt yields. This uniqueness during periods of dysfunctionality suggests QE could have acted through the market liquidity channel to decrease premia in corporate markets during the stressed periods, something it could not do if markets were calm. The market liquidity effect is visible on the graph below from Baily et al. (2020), where the bid-ask spread on 10-year government bonds (a measure of illiquidity) is plotted over time and the points of QE1 through QE5 are indicated. The two most effective programs (1 and 5) occurred at times of illiquidity, and QE5 certainly coincided with a large increase in liquidity.
Another possible contingency on the effectiveness of QE transmission is the origin of stress in the markets. Baily et al. assert QE1 occurred during a structural shift in banks demand for liquidity, due to both an endogenous realisation by banks that they lacked high-quality liquid assets, and regulations. Thus, the new reserves created by bond purchases helped satiate this demand. QE5 (starting in March 2020) was during the previously mentioned ‘dash for cash,’ which Baily et al. hold was somewhat driven by margin calls. Hence by getting deposits to gilt market investors, the CB allowed them to meet these calls without the need for fire sales of other assets. In both cases, the design of the bond-purchases programs got liquidity to those in need. However, if, for example, liquidity issues originated in foreign funding markets, QE would not be able to ease the tension directly. It follows the origin of liquidity stress is important for the effectiveness of bond purchases.
The amplified effect of ‘going fast’ is more original to Baily et al.’s analysis, insofar as little to no other popular papers have dealt with the issue before. Born out of the Bank’s economists experience of the ‘dash for cash,’ in March 2020, it relies on a specific dynamic only present during times of heightened risk aversion and liquidity shortage. Increased risk aversion creates a demand from investors to rebalance portfolios. If CB purchases do not occur at a quick enough pace such that immediate demands are not satiated, the remainder will be made up for with fire sales causing asset prices to fall further and creating a positive feedback effect. With these dynamics in mind, the MPC conducted their March 2020 QE at the ‘maximum achievable pace consistent with improved market functioning, purchasing £13.5bn worth of gilts a week (two times more per week than QE1). The additional £100bn of purchases in June 2020, once markets had relaxed, was reverted to a standard pace.
It is worth noting, a not insignificant section of the specifically pre-Covid literature objects to the claim QE had been state contingent. Bernanke (2020) reviews the literature and suggests much of the evidence cited for ineffective asset purchases is actually a result of accurate and unaccounted for market anticipation. If the actual purchases had not occurred, yields would have risen as investors shifted expectations and portfolio positions.
The ECB’s first significant dalliance into QE (starting in March 2015) is generally considered to have been well-anticipated and occurred at a time of standard market functioning (specifically no serious liquidity shortages). To account for the full effect of the purchases, De Santis et al (2019) use mentions of the program in Bloomberg articles as a proxy for market expectations. They find that, following a speech by Draghi on 22nd August 2014, expectations of ECB asset purchases increased significantly. The total estimated effect up until October 2015 is an average 72bp fall in the 10-year sovereign yield, with 89% of the effect occurring between September 2014 and February 2015. The index certainly indicates the important role of expectations, and how they can perhaps be misconstrued as evidence of state-contingency.
Effects of QE on foreign economies: Financial spillovers
Bond purchases in large economies, or economies home to internationally significant financial markets, may have international spillovers through demand, exchange rate and financial channels. As the largest economy and owner of the global reserve currency, most work on the subject has focused on the US.
As a side effect of the mechanisms that lead to the portfolio balance channel, it is unlikely investors want to hold onto the deposits gained through CB asset purchases. As these are swapped for corporate bonds, the quantity of domestic currency in forex markets expands, leading to deprecation, and an expansion of net exports in the medium to long term (accounting for any J curve effects). This negative spillover to foreign markets is, however, offset by the expansion of imports in the domestic country resulting from an increase in the components of domestic demand.
More significant in the evidence, according to Hadane et al (2016), is a powerful international portfolio rebalancing effect. Ex-post any QE announcement (without loss of generality assume this is for the US) and subsequent revision of exchange rate expectations, but ex-ante any portfolio swaps, international investors holding dollar-denominated assets see a fall in the relative return of their assets to, otherwise identical, foreign-denominated assets. Arbitrage implies portfolio switching amongst these investors away from dollar assets and towards foreign ones. As a result, yields on said foreign assets fall similar to if foreign countries had themselves engaged in QE. There is empirical evidence of this positive spill-over. Rogers, Scotti and Wright (2014) find that US monetary policy surprises which cause a 25bps fall in the US 10-year yield, cause roughly a 10bps fall in 10-year rates in the UK and the Euro area. A similar UK monetary policy surprise has a smaller but statistically significant effect on US yields, and equivalent surprises in the Euro area and Japan have no statistically significant effect on US yields. Fratzscher et al. (2013) find that US QE1 triggered a portfolio rebalancing out of emerging market economies (EMEs) to the US, but QE2 triggered an opposite rebalancing. This suggests a pro-cyclical effect of US QE on EMEs. When markets are stressed, the Fed’s actions reassure risk-averse investors that US assets are relatively safer than those of EMEs. When markets are less stressed, the expansion of dollar capital encourages US investors to look to higher-return emerging markets.
Unintended Consequences of QE
Despite QE supposedly causing an expansion of the macroeconomy, many have criticised QE for having various unintended consequences, one of which is the “crowding out effect” which occurs when some types of investments could potentially be compromised as a result of the central bank choosing other types of investments that would be incentivised and promoted. For instance, in the US, lending to corporations was largely neglected as the Fed mainly focused on buying treasuries and mortgage-backed securities in QE.
Furthermore, persistently low interest rates have caused problems for pension funds hunting for returns, incentivising them to move into riskier areas such as illiquid assets and grapple with rising debt levels and asset prices. Even though the QE continued to suppress volatility and prevent a big market correction, it also distorted asset values and sidelined the conventional investment rules on diversification, risk-return relationship and risk-free assets. It gave governments an excuse to backslide on essential growth-friendly supply-side reforms and overinflated pension liabilities via zero-bound interest rates.
Another concern relates to the claim that QE may increase the wealth of those already wealthy. This can shortly be explained by the fact that the majority of the wealthy hold a large sum of the assets purchased by the BoE when they exercise QE, so when asset prices increase, so too does the wealth of those who hold these assets. However, there is also room to argue that the bottom half of households benefit greatly from the use of QE, and so decreases income inequality.
These ideas were analysed by an independent think-tank, Resolution Foundation, which believed that by looking at the income effect one can claim that QE has positive distributional effects on the macroeconomy. The income effect is explained by rising employment, which is induced once the output gap decreases and the economy is producing more goods. Alongside this, the tightening of labour market conditions causes wages to increase for workers as more workers are demanded to aid the increase in output, leading to higher incomes for households and increased consumption levels. What is positive to note is that the combination of rising employment and wage growth has had the most effect on households centred towards the lowest income decile. This can be seen in the figure below by which the average real change in annual labour and benefit income (new income generated) seems to have a progressive effect if we look the 10th decile down to the 1st (richest to poorest). We see that as the income decile progresses, so too does the effect of QE on income. This leaves reason to suggest that income inequality is somewhat reduced by QE as those of the poorest household gain a greater income from QE than those of a richer household. Though this may be true when presented, it is important to realise that these gains via the income effect, in monetary terms, are still substantially different. Data presented by the ONS shows that the top 10% had, on average, final income of £73,682 at the end of 2014 and the bottom 10% received final incomes of £12,050 (2013–14). When you assess the gains in income from this data compared to that which is proportional to GDP, the top 10% still hold greater shares of income relative to the bottom 10%. Therefore, though the poorest decile may receive higher gains in income, it does not reduce the huge income gap between them and the top 10%; rather, it increases it.
The wealth effect derives from the purchase of assets, which causes rising asset prices thus greater wealth generated for those holding such assets. Changes in property wealth are also said to be a part of the wealth effect as the effect of rising financial asset prices tends to spill over into the housing sector. However, the purchase of bonds, equity and portfolio rebalancing is a far more targeted channel by QE. Since the ownership of wealth is already heavily skewed towards the top 10% of households in the UK, this means that changes in asset prices majorly affect the 10% whilst minimally serving the wealth of the lower-income households. Bearing this in mind, the think tank finds that around 40 percent of the aggregate boost to wealth from changes in financial asset prices, property prices, and inflation went to families in the highest wealth decile, while only 12 percent of the benefit went to the bottom half of the distribution, suggesting that the criticisms of many do hold. Note that the type of financial assets held by households also contributes to the aggregate impact as households holding more financial assets are more affected by QE than households holding other assets like property. Furthermore, upon assessing the UK Lorenz curve between 2012–14 cited from the ONS, the top 10% own a greater proportion of total income in the economy compared to the bottom 10% (10% and 1%, respectively). Ultimately, this goes to show how the wealth effect far outstrips the income effect since the top 10% will inevitably get bigger gains from changes in wealth prices compared to households at the bottom 10%.
In the staff working paper no.720 of the Bank of England, Bunn, Pugh and Yeates (2018) investigated the distributional impact of UK monetary policy between 2008 and 2014. They calculated the difference between the actual Gini coefficient and the counterfactual to estimate the effect of monetary policy on income and wealth inequality. In this analysis, the pre-crisis and pre-QE trends in inequality have been considered and QE is found to have a modest effect on Gini coefficients. Combined with a lower Bank Rate, QE is estimated to decrease the net wealth Gini, compared to what would otherwise have happened. When this effect is decomposed, it can be seen that financial wealth and pension wealth growth on account of monetary policy actually had a positive impact on the wealth Gini, however, the effects of inflation and of housing wealth were more dominant (housing wealth is a very important source of wealth for the middle of the distribution) and helped reduce the Gini coefficient.
On the one hand, the paper implies that the income effect helps to compensate for the wealth gap since we are analysing distributional effects and so are more concerned with the relative effects across different parts of the economy rather than the macroeconomic impact. On the other hand, the think tank does not place relative weights of each effect on the macroeconomy. Even though the income effect targets more parts of the economy than the wealth effect, if the income effect is minimal compared to the wealth effect, then it is the wealth effect that will have greater implications on the macroeconomy and further enhance wealth inequality in the UK, whilst income inequality remains constant. However, if the income effect is greater than the wealth effect, then income inequality surely reduces and potentially wealth inequality too, depending on how households in the lowest deciles spend their income. Furthermore, gains in income still deepen income inequality despite the relatively higher gains by the bottom decile compared to the top, because the top decile still owns a greater proportion of the economy. Therefore, it can still be asserted that QE aids the wealthy more than the poorest households and is a measure that continues to increase income and wealth inequality in the UK.
Nevertheless, since the distribution was already highly skewed towards the wealthiest prior to the Great Recession and therefore prior to the QE, the marginal gain of wealthier households which possess more housing wealth and pension wealth from monetary policy is estimated to be much larger in cash terms. The marginal increase in the measured real wealth of the bottom 10% is estimated to be approximately £3,000 between 2006–2008 and 2012–2014 whereas it is £350,000 for the top 10%. Nevertheless, the increase in measured wealth, especially in housing and pension wealth, does not necessarily enable households to consume more, i.e. make them better off. When interpreting these figures, it is also important to consider that they are the marginal contributions of monetary policy; over the given period, the decline in real house and equity prices would have a disproportionate negative impact on real wealth of richer households, so these marginal gains reduced the extent to which asset holders were made worse-off, instead of making them better-off overall.
When the distributional effects of QE by age are analysed, it can be concluded that monetary policy narrowed the absolute gap in income between the young and elderly and mitigated the loss of younger households, which did much worse than older households during the financial crisis period overall. Younger people are more likely to be employed so they have benefited the most from a stronger economy, which leads to a lower unemployment rate and higher wages. Moreover, since recessions tend to affect the employment prospects of younger people more adversely, the labour income effects are skewed towards younger households. Unlike younger households, the older people are more likely to be retired/not working and savers, so they are estimated to lose out on savings income because of lower interest payments on their savings. Between 2008 and 2014, monetary policy is estimated to have decreased the incomes of older households (where the head is over the age of 65) by 20% and increased the annual income of younger households (where the head is aged under 40) by approximately 60% of a year of income in total.
This article is written by: Sheryl Dong (Head of UK Monetary Policy Research), Birce Akay (Research Associate, UK Monetary Policy Research), Daniel Carey (Research Associate, UK Monetary Policy Research) and Dillon Oppon-Ferguson (Research Associate, UK Monetary Policy Research). This article is reviewed by Jason Jia (VP, Monetary Policy Research).