La forma de experimentar el recorrido escolar según el nivel

There are two sources of financial market information for this chapter; Bloomberg professional services and Twitter. Information events sourced via the former facilitate the inclusion of times during which scheduled and unscheduled public information arrives. For scheduled events, a unity value is included as a dummy variable for the event window commencing at the minute the information is released via Bloomberg. The inclusion of said event windows allow for the testing of the papers’ main hypothesis while controlling for any volatility jumps attributable to scheduled public information events. From Bloomberg professional services, I collect timestamped data for 20 categories of scheduled public information releases, totalling in 429 events of this type for the period September 29, 2013 to May 08, 2015. Further, 250,000 unscheduled public information arrivals (news headlines) are collected via Bloomberg for the same period. The times of such news headlines are cross-referenced with the sample of timestamped rumour information. Any rumour observations which either repeat such Bloomberg information arrivals or occur concurrently are eliminated from the sample.3

Details of public information data is provided in Table 2.2. For the purpose of this chapter, I collate scheduled public information in twenty categories, FOMC rate decisions, ECB rate decisions, FOMC meeting minutes release, speeches given by prominent ECB and FOMC Committee members, US employment reports, Category 1 economic data (US GDP, US CPI, US ISM manufacturing data, US consumer confidence, German ZEW economic confidence data, German IFO economic confidence data and Eurozone CPI) and Category 2 economic data (US retail sales data, US Durable Goods, US Manufacturing PMI, German Employment Report, European PMI manufacturing, German Industrial Production and German Factory Orders). A relevance indicator provided by Bloomberg determines the constituent economic data events included in the latter two categories.4 In further research there is scope to include additional control variables, particularly those relating to the microstructure of the EUR/USD market.

Preliminary analysis of the EUR-USD market shows that of the largest 25 absolute returns for the sample, 10 occur concurrently within the arrival of news associated with the ECB (see Table 2.3). It is therefore appropriate, that I focus on highly relevant market rumours relating to forthcoming ECB actions or changes in remit. Such highly relevant market rumours are appropriate examples of actionable information discerned by market actors but not yet

3 _{The inclusion of further control variables such as sentiment indicator could potentially augment this research. However, for simplicity and}

to avoid possible endogeneity problems, I have chosen to exclude such indicators.

4 _{It is worth noting that for Eastern Standard Time (New York) the change to daylight savings time occurs sooner and ends later than in Western}

Europe. This can cause some disparity when observing US related information events. I control for these disparate periods when constructing public information dummy variables.

24

investigated by economists. Rumours of this type are quoted as ‘ECB sources’ stories. These rumours are regularly reported by ‘in the know’ financial market commentators via Twitter. ECB sources stories are particularly prevalent within a one-week window of the ECB’s Governing Council meeting that takes place on a monthly basis. I can gauge the popularity of the ECB sources story by the number of times the quoted rumour is repeated. It is relatively simple to search Twitter archives for the phrase ‘ECB sources’. I select ECB rumour events where the quoted story is repeated by more than 50 ‘in the know’ financial market commentators. I then perform an advanced search for the full quoted story i.e. “ECB Sources:

ECB is working on a discussion chapter to execute government bond buying. 3 different options”, and pinpoint the time of the first broadcast of the quote. In total, I collect times for

the first broadcast of 63 ECB rumour events. Some ‘sources’ stories gain so much traction among financial commentators that they are then reported via Bloomberg professional services. Details of the 63 ‘ECB sources’ events are given in Appendix A1.

The applied threshold of 50 repeats of a given rumour is determined arbitrarily, but nonetheless is reasonable method of preventing the selection of stale rumours and those which are largely

Table 2.2

Scheduled public information arrival for period September 29, 2013 to May 08, 2015.

Announcement Regular Time

(GMT)

Bloomberg

Relevance Indicator

Number of Observations

FOMC Rate Decision 1900/1800 97.6 13

ECB Rate Decision 1245 97.7 18

FOMC Minutes 1930/1830 97.6 12

ECB Speakers Various N/A 46

FOMC Speakers Various N/A 52

US Employment Report 1330 99.2 20

US CPI (Cat 1) 1330/1230 94.4 19

US GDP (Cat 1) 1330/1230 96.8 19

US ISM (Cat 1) 1500/1400 96.0 19

German ZEW (Cat 1) 1000 98.3 19

German IFO (Cat 1) 0900 96.6 19

Eurozone CPI YoY (Cat 1) 1000 95.3 19

US Consumer Confidence (Cat1) 1330/1230 95.2 19

US Durable Goods (Cat 2) 1330/1230 92.1 19

US Retail Sales (Cat 2) 1330/1230 91.3 19

US Manufacturing PMI (Cat 2) 1445/1345 90.0 19

German Employment Report (Cat 2) 0855 90.0 19

Eurozone Manufacturing PMI (Cat 2) 0800 to 0900 90.0 19

German Industrial Production (Cat 2) 0700/1100 93.2 20

25

ignored by financial market commentators. Since the rumours are collected manually, the threshold also serves as a means of eliminating selection bias. In effect reducing problems associated with hand selected information samples by automating selection. There is scope in future research to perform sensitivity analysis to gauge a more precise threshold for a market rumour to be considered ‘actionable’.

26