Very much anticipated: ECB QE had a big impact on asset prices, even before it was officially announced

Menno Middeldorp.

What did markets price in before the European Central Bank (ECB) started purchasing government bonds? This is a difficult question, but using the frequency of news articles related to ECB QE can offer an answer. Regression analysis using news flow suggests that substantial price moves in the bond, equity and foreign exchange markets in the euro area, UK and, in some cases, the US anticipated ECB QE.

The ECB and quantitative easing

On March 9th the ECB started buying sovereign and other public sector bonds in its Public Sector Purchase Programme (PSPP), although market participants generally prefer the term quantitative easing (QE). Even prior to the official announcement, there was widespread discussion among market participants and in the financial press about the possibility of ECB QE. Chart 1 shows the relative frequency of Bloomberg news articles that contain the terms “QE” or “quantitative easing” along with “ECB”. The flow of QE news picked up almost a year before PSPP started.

Chart 1: ECB QE related news flow on Bloomberg and major ECB announcements

2015_24_chart1

(a)           ECB discusses QE
(b)           Preparation for asset-backed purchase programme
(c)           Draghi Jackson Hole Speech
(d)           Covered bond and asset-backed purchase progs announced
(e)           Covered bond purchase programme starts
(f)            Asset-backed purchase programme starts
(g)           ECB discusses QE
(h)           Public Sector purchase programme announced
(i)            Public Sector purchase programme starts

Source: Bloomberg, ECB

The power and pitfalls of news flow analysis

Given that market participants are forward-looking it is likely that asset prices reflected ECB QE before purchases began. One of the few ways of identifying the market’s anticipation of QE is to use the news flow data in Chart 1. Bloomberg news will typically report not only on statements by ECB officials but also on the views of economists, strategists and market participants who comment on QE, including forecasts of QE and opinions about whether market moves are due to QE expectations. So, if asset price moves coincide with spikes in QE related news flow, then it makes sense to conclude QE expectations were a factor. Straightforward regressions of changes in various asset prices on news flow about ECB QE show that such moves did indeed coincide.

However, there are at least two potential problems with news flow analysis. First, the news might just not be relevant: there could be a lot of articles about QE without there actually being any fresh information that might influence expectations; market commentators may incorrectly attribute some unrelated asset price changes to QE; and despite the fact that the ECB’s covered bond and asset-backed security programs are much smaller, some of the news could be solely about them, without implications for PSPP. Second, the sign of the relationship between QE news and asset prices could change. While news flow will probably rise whenever there is something QE-related to report, that does not tell us whether this news made QE more or less likely. Therefore, this approach will be most helpful when QE expectations are predominantly heading in one direction. This argues against using the news flow variable after purchases actually started because of the decline and then reversal of government bond yields. Prior to PSSP’s implementation, however, QE news generally pointed in the same direction, suggesting the news flow variable is appropriate.

In the end, the proof of the pudding is in the eating. The regression results indicate a clear relationship between QE news flow and changes in several asset prices, to a degree that is statistically very unlikely to be due to mere coincidence. So it appears to be a very useful measure indeed.

Using news flow in asset price regressions

The approach used here is to regress changes in various asset prices on news flow and control variables. The asset prices included here are 10-year sovereign bond yields and equity indices for the EA, UK and US, plus the spread of Italian sovereign yields over Germany and the exchange rate of the euro against the dollar and sterling (See Table 1 for more detail).

Table 1: Regression specification summary

2015_24_table1

Of course, there are many other things that could affect asset prices that are not related to QE. Although it is too ambitious to try and define complete models of the changes in all these asset prices, I do introduce some control variables to try and account for this.

10-year overnight interest rate swap (OIS) rates –The 10-year OIS rate should reflect the expected path of overnight money market rates, which are essentially controlled by the ECB. This future path of overnight interest rates should also be priced into government bond yields. However, bond yields will reflect other factors too, including the ‘scarcity’ of these bonds, which is directly affected by asset purchases. So including OIS rates in the regression should control for the movement in interest rates not directly attributable to asset purchases. The disadvantage is that QE might also impact OIS rates indirectly because it is interpreted as a signal about future monetary policy or affect aggregate duration risk in the private market, which could mean the coefficient on the news flow variable might not capture the full QE effect. Nevertheless, Joyce, Tong and Woods (2011) found that the Bank of England’s QE programme had a substantially bigger effect on UK gilt yields than OIS. All the regressions use domestic OIS, apart from the exchange rate regressions which use the spreads between the euro and foreign OIS rates. In most cases the inclusion of OIS rates as a control variable does not affect the main results substantially, except for government bond rates. There the effects are bigger (but less certain) if we don’t control for OIS, so I also show these results Table 2.

Economic surprises – These measure the difference between economist expectations of macroeconomic releases and the actual outcomes. Because US macro news has been widely found to impact UK and EA markets, I include US surprises. Here too, the control variable may capture some of the effect of QE, to the extent that QE expectations respond directly to the economic news.

US equity prices – US asset prices capture both domestic and global factors that influence asset prices, and thus can be used to control for the latter. Of course, this is problematic if US asset prices are impacted by ECB QE. I find evidence of this for US government bonds, so I only use this control for the EA and UK equity regressions

The sample runs from 2 January 2013 to 6 March 2015 (566 observations), using two-day changes for all asset prices and two-day totals for the news flow variables, which has three benefits. First, it allows for news that is released after market hours to impact prices on the next trading day. Second, it gives potentially less liquid markets more time to respond to QE news. Third, it reduces the mismatch between asset prices that trade in different time-zones – this is particularly relevant when using the US as a control variable.

Regression results and calculating total effects

The majority of asset prices examined here have a statistically significant relationship with the news flow variables (Table 2).

Table 2: Summary of regression results

2015_24_table2

Using the news flow data and the coefficients on the QE variable I can estimate total effects of ECB QE on each asset price along with cumulative standard errors. The results are reported in the last two columns of Table 2. Charts 2-4 shows a similar counterfactual exercise, where the cumulative effect of QE has been subtracted from the current asset prices (solid lines) to show where they would have been if QE had not been anticipated (dashed lines).

Asset prices before ECB QE – actual vs. counterfactual (without ECB QE)

Chart 2: Sovereign Bonds

BU 2014_24 Chart 2 Sovereign bonds

Chart 3: Euro exchange rates(a)

2015_24_chart3

Chart 4: Equity indices(a)

2015_24_chart4
(unbroken lines) Actual      (broken lines) Counterfactual
(xx%) Absolute difference between actual and counterfactual.
(a) Differences are in %-points, while total cumulative effect reported in Table 2 is percent change since 31/03/2014
Source: Datastream, Bloomberg, Bank of England yield curves, Author’s calculations

Conclusion

The results indicate that there were statistically significant and economically meaningful effects from anticipation of ECB QE across a variety of asset classes in the EA and UK and to a lesser extent the US. While it is difficult to be precise about the exact scale of these effects and while the news flow and control variables have drawbacks, there are few alternative ways to estimate QE effects. Furthermore, the overall effects quantified here match the widespread perception among financial market participants that the anticipation of ECB QE had a major impact on asset prices.

Menno Middeldorp works in the Bank’s Macrofinancial Analysis Division.

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.

If you want to get in touch, please email us at bankunderground@bankofengland.co.uk 

Comments Off on Very much anticipated: ECB QE had a big impact on asset prices, even before it was officially announced

Filed under Financial Markets, International Economics, Macroeconomics, Monetary Policy

Comments are closed.