‘Whatever it takes’ and the role of Eurozone news

ABSTRACT

This letter shows that the ‘Whatever it takes’ speech by ECB President Draghi on 26 July 2012 and the ensuing installation of the Outright Monetary Transactions’ framework are associated with a reduction in the domestic and cross-border effect of Eurozone news on absolute yield changes in Eurozone sovereign debt. These results are consistent with the popular view that these actions helped to avoid a collapse of the Eurozone.

KEYWORDS:

• Eurozone
• news
• break test
• crisis
• ‘Whatever it takes’

JEL CLASSIFICATION:

• E58
• E62
• G12
• G15
• H63

I. Introduction

On 26 July 2012, European Central Bank (ECB) President Mario Draghi stated: ‘Within our mandate, the ECB is ready to do whatever it takes to preserve the euro. And believe me, it will be enough’. Even though the Eurozone officially does not have a lender of last resort, Draghi signalled that the ECB would go as far as necessary within their mandate to save the Euro. Shortly after Draghi’s speech, on August 2, the Outright Monetary Transactions (OMT) framework was announced, which became operational on 6 September 2012. In the popular press it has been widely claimed that these actions have indeed calmed the Eurozone sovereign debt markets, a conclusion also supported, for example, by De Grauwe and Ji (2013) and Saka, Fuertes, and Kalotychou (2015).

Using an update of the news database by Bahaj (2014), this article revisits this claim empirically by testing whether Draghi’s speech and the ensuing OMT actions indeed reduced the impact of news on domestic and foreign changes in Eurozone sovereign debt yields. Our results do confirm a break in the transmission of news in this period. In particular, while highly significant before, the cross-border effects on absolute yield changes are absent in the period after.

The remainder of this article is as follows. Section II presents the data, Section III the empirical framework and Section IV the results. Section V concludes.

II. The data

We extend the news database constructed by Bahaj (2014), which is based on the EuroIntelligence news briefing from July 2009 until March 2013.¹ This briefing is released every weekday around 9 o’clock in the morning and contains the previous day’s most important economic and political news from the Eurozone. It focuses on news from the ‘crisis countries’ Cyprus, Greece, Spain, Ireland, Italy and Portugal. The timing of the news is checked by looking up the time at which the news appears for the first time on Bloomberg. Therefore, the EuroIntelligence news briefing is used as a filter to determine which news can be considered sufficiently important, while the timing of the news comes from Bloomberg. News is assigned to the date it first appears on Bloomberg (usually the day before it is reported on EuroIntelligence). News is only considered news if it falls into one of the following five categories created by Bahaj (2014): political events, foreign interventions (for example, statements by foreign politicians, bailout agreements, etc.),² technical events,³ fiscal data (in particular, news about data revisions, future fiscal projections, data quality, etc.) and instability. Besides these categories, news must also be ‘time-able’ in the sense that it is possible to isolate at what time the news event occurs. Furthermore, an event must be limited to a single country in the above group. Although more news may be available to investors, these categories capture a substantial fraction of the relevant news. Following Bahaj’s (2014) approach, we have extended his database until July 2016. For consistency reasons this was done by only one of the authors. Table 1 records the number of days that there is news about one of our six countries. Probably not surprisingly, the largest numbers of news days and news events are associated with Greece.

Table 1. Numbers of news days and news events, 1 July 2009–1 July 2016.

	Cyprus	Greece	Spain	Ireland	Italy	Portugal	All
Days with no news	1665	1368	1527	1599	1548	1558	969
1 event	29	261	158	84	130	125	456
2 events on same day	4	51	13	16	18	16	189
3 events on same day	0	17	3	2	3	2	52
4 events on same day	3	2	0	0	0	0	23
5 events on same day	0	1	0	0	2	0	9
6 events on same day	0	1	0	0	0	0	3
Total news days	36	333	174	102	153	143	732
Total news events	49	433	193	122	185	163	1145

Further, we take from Datastream the 10-year benchmark public debt closing yields for our crisis countries, plus the US and Germany. The dollar–euro exchange rate and the ECB main refinancing rate are also taken from Datastream. The S&P500-based Chicago Board Options Exchange Volatility Index (VIX) is the daily closing index taken from the Chicago Board Options Exchange. There are some days missing due to US bank holidays. Hence, out of a total of 1828 weekdays between 1 July 2009 and 1 July 2016, we are left with 1701 days in our sample.

III. Empirical framework

We estimate the following regression model:

(1)

where is the absolute change of the end-of-day benchmark 10-year government bond yield. We use the absolute yield change, because we do not distinguish between ‘good’ and ‘bad’ news and, hence, it is not a priori clear into which direction a news event should affect a yield. Moreover, in line with most of the relevant literature, we expect that more news as such produces more volatility in the financial markets. Unobserved country fixed effects are captured by . In our analysis below, we include five lags of the dependent variable, i.e. the number of lags corresponds to one work week. Because of the large number of observations in the time dimension, the standard bias associated with fixed-effects panel estimation with lags of the dependent variable as regressors will be negligible. Further, measures the number of news events specific to the country i. Hence, is foreign news to country i, i.e. the sum of the news events in the crisis countries other than country i in our sample. We thus allow for the possibility of a spillover of foreign news onto domestic absolute yield changes. is a dummy variable that is zero before the day of Draghi’s speech, 26 July 2012, and one from this day onwards (inclusive). We interact this dummy with the other variables to test for a difference in the coefficients before and after the speech.

The variable includes a set of control variables that are expected to be potentially relevant for changes in sovereign yields. These variables include the change in the VIX index, the change in the dollar–euro exchange rate, the changes in the German and US benchmark 10-year government debt yields and the change in the ECB main refinancing rate. We also allow for lagged effects of the controls. The VIX is based on market expectations of near-term stock price volatility implicit from option prices and serves as a global risk indicator, as in De Santis (2014). The dollar–euro exchange rate, defined as the amount of dollars for one euro, serves as an indicator of regional risk. An increase in risk in the Eurozone reduces the attractiveness of Eurozone assets. The US 10-year government yield controls for global market conditions and the German 10-year government yield for Eurozone market conditions. The ECB main refinancing rate reflects the ECB’s perspective of the economic outlook for the Euro area. Finally, is a mean-zero error term.

IV. Results

Table 2 reports the coefficient estimates for a panel including Ireland, Italy, Portugal and Spain. We do not report the estimates of the coefficients of the lags of the dependent variable. They are significant in most cases, and their magnitude shrinks with the length of the lag. Although we also have news for Cyprus and Greece, we do not include these countries as cross-sectional units in the panel. There is only a limited number of news events for Cyprus, while preliminary estimation shows that Greece responds fundamentally differently to news than the other countries. However, the news events from these countries are included as foreign news. SEs are clustered. Regression (1) includes only domestic news, and shows that before Draghi’s speech one additional domestic news event increases the change in the sovereign yield by about 2 basis points, while after the speech the effect is still significant, but only about one-third as large. The difference between the effects before and after Draghi’s speech is highly significant, as shown by the p-value of the test of in the next column.⁴ Regression (2) adds foreign news. Now, the size of the response of the absolute yield change to domestic news falls by more than half after Draghi’s speech. Foreign news is highly significant before the speech: one additional foreign news event has an effect of about 0.6 basis points. However, after the speech the effect becomes insignificant and very close to zero. This result suggests that Draghi’s speech and the ensuing OMT actions have eliminated the potentially harmful cross-border spillovers of news.

Table 2. Coefficient estimates.

	Dependent variable:
	‘Whatever it takes’ speech, 26 July 2012
Break date	(1)	p-Value of test equal coefficients before and after break	(2)	p-Value of test equal coefficients before and after break
	0.023***		0.022***
Coefficient estimates before break date 26 July 2012
	0.022***		0.021***
			0.0058***
	0.28*		0.27*
	−0.13*		−0.13*
	0.0018		0.0020
	0.20		0.27
	−0.16		−0.14
Coefficient estimates after break date 26 July 2012
	0.0073**	0.000	0.0089***	0.001
			−0.0013	0.015
	−0.014	0.074	−0.019	0.076
	−0.0089	0.078	−0.0065	0.078
	0.0017*	0.75	0.0017*	0.70
	−0.24**	0.29	−0.24**	0.21
	0.15**	0.14	0.14*	0.19
N	5584		5584
- adjusted	0.195		0.198

Notes: (i) Model is estimated over the period 1 July 2009–1 July 2016, (ii) clustered SEs, (iii) fixed effect panel regression, (iv) countries included: Ireland, Italy, Portugal and Spain, (v) N = number of observations. (vi) * = 10% significance, ** = 5 % significance, *** = 1% significance

We also observe that in both regressions the change in the ECB main refinancing rate and the VIX become significant and positive after Draghi’s speech, although the coefficient of the VIX becomes slightly smaller. The change in the foreign exchange rate becomes significant and negative. By contrast, the changes in the German and U.S. ten-year yields lose their significance.

If we set the break date at one of the other two relevant dates mentioned above, i.e. at the official announcement date of the OMT on August 2 and the formal starting date of the OMT on September 6, then the results are qualitatively and quantitatively unaltered.⁵ This is not surprising, given how close these dates are to Draghi’s speech date. Finally, we have repeated our regressions replacing the changes in German and US yields with their absolute values. The point estimate of the coefficient on foreign news in the period before Draghi’s speech becomes marginally smaller. However, as before, the coefficient is highly significant before and insignificant after the speech. ⁶

V. Conclusion

In this article we have explored how news affects daily absolute changes in sovereign yields before and after Draghi’s ‘whatever it takes’ speech and the ensuing installation of the OMT. Our results are consistent with the broad consensus in the media that his speech has had a calming effect on the markets. While both domestic and foreign news are (highly) significant in the period before, the effect of domestic news more than halves, while the effect of foreign news is almost completely eliminated and insignificant in the period after, indicating a strong reduction in potentially harmful cross-border news spillover effects.

Disclosure statement

No potential conflict of interest was reported by the authors.

Notes

¹ Related works using news data based on the EuroIntelligence news briefing are Beetsma et al. (2013; 2017).

² For example, a statement by a German politician about Greece would be considered a news event for Greece.

³ Bahaj (2014) defines these as ‘technical market news directly related to the event country sovereign bond market. This includes the results from important bond auctions (either from a liquidity perspective or due to their signaling value), pronouncements by credit rating agencies and decisions from the ISDA over whether certain policy actions (such as the bond buy-back programme) constitute technical default’.

⁴ Here, and in the sequel, significance is understood to be significance at the 10% level, or higher.

⁵ The estimates are available from the authors upon request.

⁶ The estimates are available from the authors upon request.