Prediction Markets, Twitter and Bigotgate

Size: px
Start display at page:

Download "Prediction Markets, Twitter and Bigotgate"

Transcription

1 Department of Economics gareth.jones Section name Economic Analysis Research Group (EARG) Prediction Markets, Twitter and Bigotgate by Leighton Vaughan Williams and J. James Reade Discussion Paper No. 114 November 2014 Department of Economics University of Reading Whiteknights Reading RG6 6AA United Kingdom Department of Economics, University of Reading 2014

2

3 Prediction Markets, Twitter and Bigotgate Leighton Vaughan Williams Nottingham Trent University J. James Reade Department of Economics, University of Reading and Programme for Economic Modelling (EMoD), Institute for New Economic Thinking at the Oxford Martin School October 11, 2014 Abstract We consider the impact of breaking news on market prices by looking at activity on the micro-blogging platform Twitter surrounding the #bigotgate scandal during the 2010 UK General Election, and subsequent movements of betting prices on a prominent betting exchange, Betfair. We find that the response of market prices appears sluggish, as over a thousand tweets are sent before any price movement is registered (despite trading taking place). However, this slow movement appears to be explained by the need for corroborating evidence via more traditional forms of media; once important Tweeters begin to Tweet, once hyperlinks are added to Tweets, and once television and radio news bulletins begin, prices begin to move. JEL Classification: G14, L83, D72. Keywords: Information and market efficiency, Gambling, Political Elections. 1 Introduction A central concern of economists is the information efficiency of markets; how quickly is information reflected in market prices? In this paper we investigate the breaking of a prominent news story and its subsequent impact on market prices. On 28 April 2010, unaware that he was still linked to a microphone, then UK Prime Minister Gordon Brown expressed his disgust with a voter he had just met on the campaign trail, who had mentioned immigration negatively, referring to her as a bigoted woman. Immediately this event was publicised by the television company which picked up the feed, and became known as the bigotgate scandal. It is widely regarded as having had an impact on the eventual election outcome, significantly harming Brown s re-election chances. Bigotgate is an interesting case study in that it coined a term that hitherto had never been used, and as such is very easy to trace. From that day until We would like to thank the Open Society Foundations (OSF) and Oxford Martin School (OMS) for their financial support in the writing of this paper. 1

4 Frequency of Google Searches Relative Search Frequency Bigot Bigotgate Date Figure 1: Frequency of searches on Google for bigot and bigotgate since 2004; scale is relative to week with highest search volume over sample period. Source: Google Trends. election day the term bigotgate was mentioned 15,560 times on Twitter, the micro-blogging website where tweeters are limited to 140 characters in their public messages, but before that, the term was never used. Figure 1 shows the frequency of searches on Google for the terms bigot and bigotgate over a much wider time interval, showing that bigotgate itself was not searched for at any point before the week containing April , and after the election finished was not searched for again. Even the more general term bigot was searched for 20 times more than usual in the week containing 28 April Thus we have a clearly identifiable news event, and seek to trace its impact on market prices. Prediction markets allow participants to trade contracts contingent on the verifiable outcome of events such as political elections. In this note we consider the impact on the prediction market prices for bets on the outcome of the 2010 UK General Election, the context into which Bigotgate fell, of the event itself. We use prices agreed on the commercial prediction market Betfair, and information from Twitter in order to trace the impact of the event on market prices. We find that the impact of the event on market prices is quite sluggish, and propose a number of potential explanations for this. In Section 2 we consider the theoretical context of information efficiency, and introduce both Twitter and Betfair, in Section 3 we introduce our datasets, in Section 4 we analyse the Bigotgate event in terms of Twitter and Betfair activity, and in Section 5 we conclude. 2

5 2 Theory, Information, Betfair and Twitter At its simplest, the economic theory of perfect competition implies that all relevant information about a good is reflected in its market price. With a large number of buyers and sellers, none of whom is able to influence the market price, information regarding the relative scarcity or abundance of a good s supply or demand is reflected in the price and any movements that occur. Building on that basic concept is the idea of market efficiency, or the efficient markets hypothesis. Weak-form market efficiency states that any market price reflects all past publicly available information. Semi-strong form efficiency states that markets respond instantaneously to all publicly available information, and strong-form efficiency says markets respond to even privately held information (Bachelier, 1900; Fama, 1965, 1970, 1998). If markets are inefficient it suggests there exist some barriers to participants trading on available information, such as transactions costs. A central difficulty in appraising the information efficiency of markets is measuring the point at which news broke ; while on the stock market a firm may make a profits warning and see a subsequent share price fall, it could be that via informal networks this news had broken earlier (see, for example, Fama et al., 1969; Ball and Brown, 1968; Asquith, 1983). Croxson and Reade (2014) circumvent this problem by using sports bets traded on Betfair; Betfair is a prediction market that allows tweeters to trade contracts whose value is contingent on some future event occurring. Sports bets have a number of attractive properties that Croxson and Reade (2014) exploited; they are traded over a short time horizon (no time discounting), the events are repeated numerous times (allowing appraisal of the true value of contracts), and news breaks at very distinct points in time and via television is spread very quickly. Furthermore in football, the subject Croxson and Reade (2014) study, goals are big news events, commonly moving markets by about 20 percentage points, and their exact timing cannot be predicted with any reasonable degree of accuracy. Croxson and Reade (2014) find that Betfair markets are semi-strong form market efficient. That said, also in the sporting context, Brown (2014) finds that mispricing exists as agents have information processing constraints which bind when multiple contracts exists for single matches, and when sporting matches are played head-to-head with other matches. Our study of Twitter in this paper enables us to add to this body of evidence regarding the efficiency of markets. Twitter is a micro-blogging website which enables any tweeter to broadcast 140-character messages globally by simply typing the message and clicking tweet. All twitter users (tweeters or twitterers) have their own twitter feed, where all the tweets sent by tweeters they follow are posted. Hence any Twitter tweeter who writes and sends a tweet ensures that at least momentarily that message appears on the twitter feeds of all the tweeters that follow them. As such, Twitter enables private information to be made public to a potentially global audience. How widely that message is viewed depends to a large extent on how important the tweeter is, in terms of their number of followers. While all tweets can be searched hence anybody can see the tweet, the seemingly more likely route for wider exposure is via retweets; other twitter tweeters can retweet a tweet, which is then seen by their followers, widening the immediate audience of the message. Already academic research using Twitter is flourishing; a Google Scholar 3

6 search for the term Twitter in the title of an article produces 12,500 hits. 1 One theme of research has been the predictive power of Twitter. For example, researchers have found that Twitter sentiment can help predicting the stock market indices (Bollen et al., 2011; Porshnev et al., 2013; Rao and Srivastava, 2012; Sprenger et al., 2013), the Euro/US dollar exchange (Janetzko, 2014), the success of box-office movies (Treme and VanDerPloeg, 2014), the outcome of elections (Tumasjan et al., 2010) and even NFL scores (Sinha et al., 2013). Another theme has been Twitter and politics, upon which our paper touches. Feng and Yang (2011) consider the adoption of Twitter by politicians, while the use of Twitter has been investigated during political election campaigns in the UK (Anstead and O Loughlin, 2014), Australia (Burgess and Bruns, 2012) and Sweden (Larsson and Moe, 2012). Anstead and O Loughlin (2014) discuss the extent to which extracting sentiment from tweets rivals political polling in attempting to reflect public opinion. To the extent that political polls are samples selected by pollsters whereas Twitter is populated by selfselected individuals, the two are likely to offer contrasting reflections on public opinion. 2 Twitter also affords an opportunity to identify and measure the breaking and spread of new information, and hence the speed at which market prices respond to such news; how soon after tweets on a subject begin to appear do relevant market prices begin to adjust? What is required in order to conduct such research is a suitable market price to observe. We focus on Betfair as it provides market prices on a wide variety of everyday events and hence its tweeters trade contracts whose value is contingent on such events. It can be argued that the price on Betfair for a particular event reflects all known information regarding that event; clearly agents trading financially on future events such as a political election will seek to avail themselves of all relevant sources of news related to that event, and evidence suggests prices are efficient (Croxson and Reade, 2014). Semi-strong form efficiency suggests that the Betfair price ought to react instantaneously to any information that becomes public. Hence we can use Twitter to detect the speed with which information is reflected in prices on Betfair. If the market is strong-form efficient, it ought to react to the events before any tweets are sent, since the moment the gaffe occurred it was private knowledge for a period of time. If the market is semi-strong-form efficient, it ought to react instantly the moment that news regarding bigotgate broke, immediately reflecting updated beliefs about event outcomes without any associated drift. If the market is not efficient then we would expect sluggish updating. Hence it is important to specify when the news broke and, indeed, what qualifies as news. It is well regarded that Twitter is a source of breaking news, both from official and unofficial sources. 3 The first Tweet sent relating to the events of #bigotgate could be defined to be the point that the news broke. Looser measures may take into account that only a small subset of Tweeters 1 See count correct as of 9 October Duffy (2014) noted that the weight of opinion expressed on Twitter for the Scottish independence referendum (#indyref) in September 2014 was dramatically different to the votes expressed at the polls. On Twitter, 73% of tweets were pro-independence, but in the referendum only 45% of voters voted for independence. 3 Elliott (2013) notes nine events broken on Twitter from the Royal Wedding to the Hudson river plane crash in New York, two of which occurred before #bigotgate. 4

7 would have seen the original Tweet, and that the impact of the information may not have been initially clear. As information breaks, the nature of the news may not be immediately clear; in our context, this would be whether the news is good or bad for the Labour Party s likelihood of winning the forthcoming election. In considering what constitutes news, we can think in terms of signal and noise. We define a signal to be relevant information for some specific event, but noise to be irrelevant information. In the context of an election campaign and the likely outcomes thereof, there will be events that occur which are irrelevant to the outcome, which we denote noise, and events that affect the outcome, which we consider to be the signal. Noise will not affect the likelihood of the event occurring, but signals will, and hence crucial to the efficiency of markets is the timely recognition of signals rather than noise. We compare a vehicle for breaking news, the micro-blogging website Twitter, to a prediction market upon which contracts contingent on outcomes of news events can be traded, Betfair. We chose a specific event during the election campaign that has great salience and hence is easily identifiable in our two datasets; outside of the election debates it is viewed as a defining moment. 4 Using the uniqueness of the #bigotgate term, and timestamped data we can trace Betfair price movements to activity on Twitter and thus learn about the nature of the market response to information. 3 Data We introduce our two data sources separately; first Betfair (Section 3.1), then Twitter (Section 3.2). 3.1 Betfair Bigotgate occurred in the final days before the General Election. We firstly look at Betfair activity in the run-up to this day. We have prices for four Betfair markets related to the 2010 general election. The first was the market for a Labour majority, with contracts stipulating the buyer pay the seller if Labour won a majority in the election; a similar market existed for the Conservatives (Tories), paying out if the Tories won the most seats in the election. A further market existed for any other party to win a majority (although realistically, the only party this market covers is the Liberal Democrats), and a final market for no overall majority if no party won enough seats to command a majority in the House of Commons. Over all four markets, just over 6m was traded, with 94% of that in the markets for a Conservative and No Overall Majority. 5 Figure 2 shows that election day, May 6, was the day with the heaviest trading, followed by the days either side of it. 6 The day of the bigotgate 4 The Wikipedia page on the 2010 General Election (http://en.wikipedia.org/wiki/united Kingdom general election, 2010 as of 9 October 2014) devotes an entire section to the episode, suggesting its gravity. 5 Of that 6m, 4.75m was traded during the statutory campaigning period between April 6 and May 6, 3.66m between the day of #bigotgate and election day, and 101, on the day of #bigotgate itself. 6 As polling stations close at 10pm and counting begins immediately, the result is not known until early morning of the day after election day. 5

8 Number of trades per day for all markets Number of Trades Apr 05 Apr 12 Apr 19 Apr 26 May 03 Date Figure 2: Number of trades per day, broken down by Betfair market. scandal, April 28, is not particularly outstanding; more trading occurs than on the previous day, but less than the following day, which was the day of the final televised debate between the party leaders. The two most heavily traded markets are for a Tory majority and no overall majority, reflected by their status, shown in Figure 3 which plots the implied probability throughout, as the most likely two events. From mid-2008 onwards, these two events take up more than 80% of the probability space for the election. Zooming in more on the final few weeks of the campaign, Figure 4 shows that the likelihood of a Labour majority was viewed as highly unlikely according to the market. There is no necessary implication that #bigotgate had minimal impact simply because Brown and Labour were unlikely to win; while a Labour majority was unlikely, it remained plausible that Labour would win enough seats to ensure there would be No Overall Majority. 7 Indeed, it is widely regarded that the improved polling performance of Labour in the weeks up to April 28 is what had contributed to the likelihood of no party winning an overall majority, alongside the performance of the Liberal Democrats. If #bigotgate reduced the number of seats Labour would win, this increased the likelihood of a Conservative majority and hence remains a significant event. In order to determine whether or not it was a significant event, it is indicative to study the time series of the Betfair market prices during the course of the election campaign. The first election debate is shown as the leftmost blue shaded area, and is notable for apparently setting in motion a downward trend in the Conservative majority probability, and increasing the likelihood of no overall majority (green). The new level that prices settled at, a few days after the first debate, remained in place until April 28, the day of Bigotgate, which is the dark red shaded area. There is a large immediate fall in the implied probability of no majority, and an equally large increase in the implied probability of a Conservative majority, appearing to occur instantaneously. This then appears to be followed by a period of drift which continues through the final debate the 7 In that situation Labour might have made a coalition with the third party, the Liberal Democrats, thus keeping the Conservatives from office. 6

9 Betfair Prices for 2010 UK Election Outcomes Implied Probability Date Figure 3: Implied probability of particular election outcomes throughout existence of markets. following day, albeit with a blip in both prices around that time. Hence based on an ocular analysis, we might conclude that Bigotgate was a significant event, as it appears to have prompted price movements; however, as the response is a seemingly immediate jump followed by drift, whether or not the market is semi-strong form efficient is yet to be determined. Zooming in on the day further, in Figure 3.1 we plot Betfair transactions on the day of the Bigotgate scandal, starting on the left at 8am that morning, and ending on the right hand side at 10pm. The green points are those bets agreed on no majority, the blue points for a Conservative majority, red for Labour and yellow for other majority. Vertical dashed lines are drawn for the hours of 11am, 12 noon and 1pm, and dotted lines for 11:30am, 12:30pm and 1:30pm. These lines suggest that the scandal broke at around 11:30am, as it is shortly after this time that the matched prices for a Conservative majority increase from 33% at 11:45am to 36% by 12 noon, and up to 39% shortly afterwards. 3.2 Twitter and Social Media Turning to Twitter, we have all tweets from April 28th until May 7th that contain the word bigot ; this constitutes 78,147 tweets. 8 While this doesn t give a particularly long baseline from which to observe the post-bigotgate effect, Figure 6 shows that there is a clear difference in tweet volume before and after the event. In using the search term bigot, it is clear we will pick up, after 11:05, a tiny amount of false positives; on average before 11:05 there is about a tweet every other minute, and after 7am (when the likelihood of tweets from people located in the USA is lowest), there are just 45 tweets before 11:05am, equivalent to one every five minutes. As such we can be confident that in our text-based metrics and analysis, we are detecting tweets that are related to our event of interest, #bigotgate, to an acceptable degree of error. 8 This dataset was purchased from Texsifter via Gnip. 7

10 Betfair Prices for 2010 UK Election Outcomes Implied Probability Labour Majority Conservative Majority Any other party No Majority Apr 16 Apr 21 Apr 26 May 01 May 06 Date Figure 4: Betfair prices for various events around the debates in the final weeks of the election campaign. Election Day Betfair Prices for 2010 UK Election Outcomes Implied Probability Labour Majority Conservative Majority Any other party No Majority 08:00 13:00 18:00 Date Figure 5: Betfair prices for Labour majority, Conservative majority, other majority and no majority on April

11 At 11:05:49 Darren McCaffrey of Sky News tweeted More Brown... Just ridiclous [sic]...there will be... Aagh, everything, she was a sort of bigoted woman... said she used to be a.... By 11:11:06 there were eight tweets, six of which were by Tweeters identifying themselves with Sky News. Between 11:05 and 11:20 there were 329 tweets mentioning the term bigot. The initial peak volume occurs between 11:37 and 11:38 when there are 240 tweets, and between then and 15:30 there was, on average, 92 tweets per minute mentioning the term bigot. There is then a slow decline in the volume of tweets, albeit with a surge between 9:30pm and 10pm, likely coinciding with topical news programmes on the BBC and late night news flashes. In our dataset there are 38,710 individual tweeters posting tweets, hence on average two tweets per tweeter. However, only about a third of these tweeters (12,592) actually tweet more than once in our entire dataset, and only 2% tweet more than ten times, with one particular tweeter tweeting over a thousand times. There are many ways in which we may wish to refine the tweets we consider; aside from featuring the term bigot, characteristics of a tweet or tweeter may indicate the quality of information provided, specifically whether it is signal or noise. For example, we might focus only on more influential tweeters as conveying information more likely to be relevant for the outcome of the election, and all other tweeters as providing noise (Greetham et al., 2013; Palacios-Huerta and Prat, 2010). Similarly we might categorise types of tweet; for example, into broadcast and conversational tweets, with the former constituting simple stand-alone messages, and the latter being messages responded to by other tweeters. Just less than half of our tweets (37,497) contain symbol, which indicates that a tweeter has tweeted a messaged specifically mentioning another tweeter. There are two categories of tweets mentioning another tweeter; one is messages directed specifically at another tweeter, and the other is retweets. Considering the latter, just under a third of tweets (23,639) contain the abbreviation RT, which is short for retweet, and indicates that a tweeter saw fit to re-post a tweet sent by another tweeter. While the distinct interpretation of a retweet remains a matter for discussion, at the very least this represents a spreading of information, as that tweet then appears on the news feed of many more tweeters. In our context it may represent the spreading of information regarding bigotgate, or alternatively of an opinion regarding it. Furthermore, a retweet will contain the username of the original tweeter, and as such, these tweets do not necessarily represent conversations. Of the former category of tweets using symbol, there are 14,303 that are not retweets and hence tweets directed at one or more other tweeters; such messages may form part of conversations, rather than simple broadcast messages. Before determining whether they form part of genuine conversations, we refer to these tweets as directed tweets. It may be that many of these 14,303 messages are only attempted conversations, rather than actual conversations; it may be that the tweeter mentioned in a tweet never responds to the initial tweet(s). Greetham and Ward (2012) studies the nature of conversations on Twitter, finding that two-person conversations are relatively balanced, whereas those involving more than two tweeters tend to be dominated by one or two participants. Considering conversations in our dataset, we need to first define a conversation. The fact that just 2% of tweeters tweet more than ten times in our entire 9

12 Tweets per minute mentioning 'bigot' on April Number of Tweets Original Tweets Retweets 03:00 08:00 13:00 18:00 23:00 Time Figure 6: Number of tweets per minute mentioning the word bigot on April Source: Twitter dataset is indicative of a very small number of conversations. If we define a conversation to involve the same tweeters, then we find there are just 2004 tweets that form part of a conversation, and only 1240 groups of tweeters producing those tweets. This, however, could simply be one tweeter tweeting another repeatedly; requiring that in pairwise comparisons the tweeter tweeting be different reduces the number of conversations to 569, with only 416 tweeters taking part in these conversations. Figure 7 shows that after April 29, the day of the final debate (in which Brown makes reference to his gaffe), the frequency of tweets using the term bigot settles down to a lower level, albeit one still twice as high as that before 11:05 on the morning of April 28, with a few spikes at various points. 4 Analysis Figure 8 shows that price movement on Betfair lags behind tweets; the event itself happened at about 10:45, the story broke at 11:05, by 11:20 there had been over 300 tweets, and by the time Betfair prices start moving at about 11:35, there had been around 1000 tweets. Between 10:45 and 11:35 there are 90 trades conducted on Betfair, spread reasonably evenly between the four markets. 9 Zooming in further, in Figure 9, it seems that is it at around 11:35 36 that the prices for the Conservative and No Overall majority begin to move. 9 There are 28 trades on the other majority market, 25 for no overall majority, 22 for Labour majority and 15 for Conservative majority. 10

13 Tweets per minute mentioning 'bigot' April 28 May Number of Tweets Apr 29 May 01 May 03 May 05 May 07 Date Figure 7: Number of tweets per minute mentioning the word bigot between April 28 and May Source: Twitter Although there is an increase in the number of tweets in these two minutes relative to the two before that, these are behind 11:27 and 11:28 in terms of denseness of tweeting, suggesting it cannot be volume of tweets per minute alone which drives the beginning of the price adjustment. At this point Brown was twenty minutes into his interview with Jeremy Vine at the BBC Manchester studios, but before Vine confronted Brown on the recording of the incident. The movement is strongest at about 11:45, which coincides with when Brown was confronted (timings are GMT/UTC for both, whereas on March 28 British Summer Time (one hour ahead of GMT) began, and hence BBC News footage (e.g. shows the time at about 12:44). Furthermore, after this initial movement has plateaued at 36% for Tory majority, 60% for no majority, there appears to be a pause from 11:50 until 12:01 where only a small number of trades occur, followed by another, sharper price movement. This closely coincides with many TV and radio news bulletins beginning on the hour (and perhaps BBC Radio 4 s The World At One). By 12:07 the Tory price had risen to 38.6% and the no majority price was down to 58%. It seems hard to escape the fact that the response to the news seems slow, since the first tweet was at 11:05 and by the time the heaviest movement occurs in prices, more than a thousand tweets had been sent. 4.1 Types of Tweets The information contained in a tweet need not provoke a market response. If the information is not new, then it ought to have already been processed and 11

14 Implied Probability (dots) Bigotgate Day Betfair Prices for Conservative/No Majority 08:00 13:00 18:00 Date Number of tweets involving word 'bigot per minute (bars) Figure 8: Price for no majority and Tory majority alongside number of tweets per minute. Implied Probability (dots) Bigotgate Day Betfair Prices for Conservative/No Majority 11:00 11:30 12:00 12:30 13:00 Date Number of tweets involving word 'bigot per minute (bars) Figure 9: Price for no majority and Tory majority alongside number of tweets per minute between 11am and 1pm. 12

15 reflected in prices. In this section we provide a number of alternative ways of viewing the tweets that occurred during #bigotgate, to seek an explanation for the seemingly sluggish market response. First, we might consider a very basic measure of the importance of a tweeter in order to shed some light on the type of tweeter tweeting at particular points on the day of #bigotgate; we simply consider the frequency of tweeters; those who have tweeted more we consider to be more important. Just 2% of all tweeters send more than ten tweets in our entire dataset, and hence we categorise this group of tweeters as a select group of important tweeters. In Figure 4.1 we break down tweets by how often the tweeter tweets in our dataset. The thickest and most grey vertical bars are tweeters who tweet the least, and the thinnest and blackest line is tweeters that tweet the most; here, those that tweet more than ten times. Note the plot is cumulative, hence the section of the thinnest line represents the number of tweeters that tweet more than ten times; for example, at 11:38, the peak of activity, there are 36 tweets by tweeters who send more than ten tweets, rather than 137. Also plotted in Figure 4.1 is the Betfair price for the Conservative (blue) and No Majority (green). Before 11:05, all tweets are by tweeters who barely tweet, but by by 11:38 a quarter of all tweets are sent by tweeters who sent more than ten tweets in the entire dataset (who we define as important ). Perhaps importantly, this period of tweets by important tweeters coincides with the first movement in prices on Betfair, between about 11:38 and 11:50. This suggests that it may be that until important and influential tweeters begin to tweet about the news it is not regarded as being important news. By 12 noon, when the second large price shift occurs, the proportion of important tweeters has fallen to only around 1 in 5 tweeters; in the first three minutes after 12 noon, the proportion of less important tweeters is relatively large. This may reflect tweeters reacting to hourly news programmes started at 12, before more important tweeters pick up on events covered in the news bulletins and propagate the information. Of course, our measure of important tweeters is basic since, for example, highly prominent news figures such as Adam Boulton only tweet once in our dataset; it may also be that such news figures tweet only around their news bulletins and this prompts the significant price shift after the hour. Regardless, it appears likely that traditional news media bulletins drive the post noon price fall. In terms of categorising tweets, we can further consider retweets, and we might propose that retweets are not new information. Nonetheless, they reflect the propagation of information; an initial tweet will only likely be seen by a small number of followers, since for anyone else to view it they would need to have searched for it. As we have no information on the number of followers of each Tweeter we cannot directly infer the impact of retweets and the spread of information. Nonetheless, we can consider the proportion of all tweets that are retweets. Secondly, we can think about directed tweets; tweets that involve a reference to another tweeter that are not retweets; i.e. tweets that elicit no reply by the tweeter(s) mentioned. Thirdly, we can consider actual conversational tweets; tweets mentioning other tweeters that received responses by some of those tweeters referred to. Finally, we might consider tweets with hyperlinks embedded in them. These are tweets that point the reader to another webpage, often a news page regarding a story of relevance. Figure 4.1 provides a rich breakdown of tweets, with retweets in red, tweets 13

16 Tweets by Frequency of Tweeter Total Tweets tweet 2 tweets 3 tweets 4 tweets 5 tweets 6 tweets 7 tweets 8 tweets 9 tweets >10 tweets 11:00 11:30 12:00 12:30 13:00 Time Figure 10: Tweets broken down by the number of tweets each tweeter sends, plotted against the Betfair market price for a Conservative (blue) and No Majority (green). 14

17 Bigotgate Day Betfair Prices for Conservative/No Majority and Types of Tweets Implied probability (dots) Remainder Retweets Directed Tweets Http Tweets Conversation Number of Tweets 11:00 11:30 12:00 12:30 13:00 Time Figure 11: Different types of tweets alongside movements in Betfair prices for Conservative/No Majority between 11am and 1pm on April with hyperlinks in blue, directed tweets in green, and conversation tweets in light blue, alongside the Betfair prices for Conservative and No Majority. The numbers of each tweet type are stacked on top of each other. The latter are clearly the smallest group, and only appear to occur after the initial peak of tweets after 11:45. They certainly appear to have no impact on price movements. In the early stages after 11:05, the proportion of all tweets that are retweets is very high, up to 80%, suggesting the propagation of information is taking place. Despite this, however, market prices are not moving. The blue bar refers to the number of tweets that contain hyperlinks; it is only after 11:30 that such tweets start increasing in prevalence, and indeed at the time that the Conservative majority price is moving the most, at about 11:45, more than 40% of tweets contain hyperlinks. This suggests that although information referring to the event was propagated, it required links to news stories, video/audio clips and potentially live feeds of the BBC Jeremy Vine interview, actual evidence, for market prices to begin moving. This would appear to suggest that while news may break on Twitter, simply the text contained in each tweet is insufficient to move markets; links to other media, and likely old forms of media such as radio and television, are what is necessary in order to move market prices. Corroborating, non-twitter, evidence would appear to be necessary before markets decide information is important enough to be reflected in prices. The directed tweets, those that mention tweeters but are not reciprocated, increase with time, but too slowly to really be viewed as contributing to price movements; it is only after around 12 noon that these begin to constitute 15 20% of all tweets. 4.2 Sentiment Analysis The simple volume of tweets may not be the appropriate indicator with which to determine the weight of new information entering the public sphere. News, 15

18 by and large, is non-neutral, being positive or negative for parties involved. As such, the early tweets, such as the first one at 11:05, may simply have been factual rather than attaching any judgement regarding the nature of the news to any of the parties involved in the election. It may thus be that this determines the delay in any Betfair response, as the market determined the nature of the news. We construct a sentiment index for our tweets in order to determine whether the balance of words used in the tweet is positive or negative. Our sentiment index is very simple; we use Hu and Lui s opinion lexicon index to assign words a positive or negative score, depending on the nature of that word. We then sum up the scores of the words and get a score for a given tweet. There are many criticisms of sentiment indices, as Anstead and O Loughlin (2014) point out; by considering words in isolation contextual information is lost, so for example sarcasm will be missed by such measures. We nonetheless use a simple index in order to determine whether even in such a basic form it can help understand the sluggish response to news breaking on Twitter. Figure 12 displays the average scores for tweets each minute over the first two hours of the Bigotgate scandal, from 11am to 1pm. The sentiment of the tweets in this period is overwhelmingly negative; just 4 of 120 minutes have a positive score, fewer than would be expected at a 5% significance level. The plot also shows that tweets become no more negative in the time surrounding when the market price begins to change. To get a sense of wider trends in terms of the sentiment of tweets as measured by this simple index, Figure 13 shows the sentiment per minute measured over the entire dataset, and shows that prior to 11:05 on April 28 the index was essentially random, but after 11:05 it is almost exclusively negative every minute until at least the next day. Also plotted on Figure 13 are a number of moving averages; the data each minute can be very noisy and as such a moving average can help smooth out some of this noise and give a better sense of the broader patterns. In the context of the sentiment of tweets, we might propose that a moving average gives some sense of recent mood. The 5- and 10-minute averages are sometimes positive prior to 11:05am, but after that are negative until well into Thursday, reflecting the changing time-series properties of sentiment. Prior to 11:05am it is essentially random; with no significant autoregressive terms, but after 11:05am it becomes autoregressive, albeit with a small (but significant) autoregressive parameter. 10 The sentiment index, however, is unable to provide any insight onto the seemingly sluggish market reaction; from Figure 14 which superimposes the 20-minute moving average on to the Tory and no majority prices, we see that sentiment, and even mood, has turned negative long before any price reaction occurs. 4.3 Information Content of Tweets Sentiment indices give some idea about the nature of news; it may be helpful to consider the textual context of tweets in a slightly different manner: What new information is being provided in each new tweet? As already discussed, retweets 10 We do not report the autoregression results, but the coefficient was 0.11, suggesting that the series, and hence mood, was persistent but not particularly strongly persistent. 16

19 Implied Probability (dots) Bigotgate Day Betfair Prices for Conservative/No Majority 11:00 11:30 12:00 12:30 13:00 Date Average Sentiment of tweets involving word 'bigot per minute (bars) Figure 12: Sentiment Analysis on Bigotgate Day. Sentiment of tweets containing word 'bigot', April Sentiment Score Sent MA 5 MA 10 MA 30 Wed Thu Fri Date/Time Figure 13: Wider Sentiment Analysis 17

20 Implied Probability (dots) Bigotgate Day Betfair Prices for Conservative/No Majority 11:00 13:00 15:00 17:00 19:00 Date Sentiment of tweets involving word 'bigot' per minute (red line) Figure 14: Using a Moving Average to determine movements in prices. have a role in propagating information, but in and of themselves they are not the breaking of news. New information on Twitter is a collection of words, the collection of which constitutes important information, but the constituent parts of which may contain words not previously used, or not used in a long time. The premise of this investigation is the salience of the term bigot, but equally the naming of names (Brown, Duffy, etc) marks the breaking of new information the individuals involved in a news break. Our idea regarding the information content of tweets is to think about the uniqueness of the words used in tweets. We start with the first tweet in our dataset and construct a set of words that have been used in tweets, in order to identify tweets introducing new terms and hence new information. The idea is that whilst the initial news regarding Brown s referring to Duffy as a bigot, further news stories are his apology on live radio, his subsequent journey back to talk with Duffy, the length of that meeting, the reference to himself as a penitent sinner on exiting Duffy s house, and so on. All these will likely add new words into the set of words used in #bigotgate tweets. It may be that as prices do not immediately move, there may have been subsequent analyses, or subsequent events like those listed which actually led to prices starting to move. Very simply, we count the number of new words used. By a new word, we mean a word not previously used in any tweet in our dataset. We start at the beginning of our dataset in order to get a baseline of commonly used words, from which to evaluate each tweet from 11:05am onwards. On average, a tweet in our dataset has words (mean 16.5, median 17, mode 20); of those, on average about 0.8 per tweet are original, within the context of tweets mentioning the term bigot % of tweets contain less than two original words, and 92% contain less than three. More new words in tweets reflect new information, and hence may lead to price movements, and thus we check for this. Figure 6 shows the number of new words per tweet on April 28; the black 11 On average tweets in our dataset are just over 100 characters (mean 99.6, median 108, mode 126), and the average word length is around 6 10 characters (mean 9.88, median 9, mode 7). 18

General Election Snapshot & Final Report

General Election Snapshot & Final Report General Election Snapshot & Final Report The votes have now been counted Throughout the 2015 General Election campaigns, the polls suggested that the out come was too close to call. With Labour and the

More information

Predicting Elections with Twitter What 140 Characters Reveal about Political Sentiment

Predicting Elections with Twitter What 140 Characters Reveal about Political Sentiment Predicting Elections with Twitter What 140 Characters Reveal about Political Sentiment Andranik Tumasjan, Timm O. Sprenger, Philipp G. Sandner, Isabell M. Welpe Workshop Election Forecasting 15 July 2013

More information

Katy Young s Guide to... Twitter

Katy Young s Guide to... Twitter 21/08/13 Step by step guide followed by advanced techniques guide INTRODUCTION Twitter is a social media platform where users tweet content. It s culture is open and encourages users to tweet without needing

More information

the beginner s guide to SOCIAL MEDIA METRICS

the beginner s guide to SOCIAL MEDIA METRICS the beginner s guide to SOCIAL MEDIA METRICS INTRO Social media can be an incredibly important business tool. Tracking the right social metrics around your industry, company, products, competition and

More information

Social Market Analytics, Inc.

Social Market Analytics, Inc. S-Factors : Definition, Use, and Significance Social Market Analytics, Inc. Harness the Power of Social Media Intelligence January 2014 P a g e 2 Introduction Social Market Analytics, Inc., (SMA) produces

More information

Predicting Stock Market Fluctuations. from Twitter

Predicting Stock Market Fluctuations. from Twitter Predicting Stock Market Fluctuations from Twitter An analysis of the predictive powers of real-time social media Sang Chung & Sandy Liu Stat 157 Professor ALdous Dec 12, 2011 Chung & Liu 2 1. Introduction

More information

Social Journalism Study 2013 United Kingdom. Report by Cision & Canterbury Christ Church University (UK)

Social Journalism Study 2013 United Kingdom. Report by Cision & Canterbury Christ Church University (UK) Social Journalism Study 2013 United Kingdom Report by Cision & Canterbury Christ Church University (UK) 1. Executive Summary Social media is now an everyday professional tool with 96% of UK journalists

More information

Do Tweets Matter for Shareholders? An Empirical Analysis

Do Tweets Matter for Shareholders? An Empirical Analysis Do Tweets Matter for Shareholders? An Empirical Analysis Brittany Cole University of Mississippi Jonathan Daigle University of Mississippi Bonnie F. Van Ness University of Mississippi We identify the 215

More information

A Description of Consumer Activity in Twitter

A Description of Consumer Activity in Twitter Justin Stewart A Description of Consumer Activity in Twitter At least for the astute economist, the introduction of techniques from computational science into economics has and is continuing to change

More information

Finance and Economics Discussion Series Divisions of Research & Statistics and Monetary Affairs Federal Reserve Board, Washington, D.C.

Finance and Economics Discussion Series Divisions of Research & Statistics and Monetary Affairs Federal Reserve Board, Washington, D.C. Finance and Economics Discussion Series Divisions of Research & Statistics and Monetary Affairs Federal Reserve Board, Washington, D.C. Business to Business Credit to Small Firms Traci Mach 2014-55 NOTE:

More information

Influencing the influencers

Influencing the influencers Influencing the influencers As corporate communicators, we think we know a lot about which media social and traditional are well read and highly trusted by opinion formers. But do we? We often talk about

More information

TWITTER 201 STRATEGIES AND BEST PRACTICES FOR USING TWITTER. Twitter 201 for Child Welfare Professionals:

TWITTER 201 STRATEGIES AND BEST PRACTICES FOR USING TWITTER. Twitter 201 for Child Welfare Professionals: TWITTER 201 Twitter 201 for Child Welfare Professionals: STRATEGIES AND BEST PRACTICES FOR USING TWITTER to Reach Foster, Adoptive, and Kinship Families Twitter 201 for Child Welfare Professionals - Page

More information

3. Understanding the part-time RAB charge

3. Understanding the part-time RAB charge 3. Understanding the part-time RAB charge Gavan Conlon, Partner, London Economics and Maike Halterbeck, Economic Consultant, London Economics Since 2001/02, there has been a marked decline in part-time

More information

Sentiment Analysis. D. Skrepetos 1. University of Waterloo. NLP Presenation, 06/17/2015

Sentiment Analysis. D. Skrepetos 1. University of Waterloo. NLP Presenation, 06/17/2015 Sentiment Analysis D. Skrepetos 1 1 Department of Computer Science University of Waterloo NLP Presenation, 06/17/2015 D. Skrepetos (University of Waterloo) Sentiment Analysis NLP Presenation, 06/17/2015

More information

Applying Machine Learning to Stock Market Trading Bryce Taylor

Applying Machine Learning to Stock Market Trading Bryce Taylor Applying Machine Learning to Stock Market Trading Bryce Taylor Abstract: In an effort to emulate human investors who read publicly available materials in order to make decisions about their investments,

More information

Jobs Online Background and Methodology

Jobs Online Background and Methodology DEPARTMENT OF LABOUR LABOUR MARKET INFORMATION Jobs Online Background and Methodology DECEMBER 2009 Acknowledgements The Department of Labour gratefully acknowledges the support of our partners in Jobs

More information

Big Data, Socio- Psychological Theory, Algorithmic Text Analysis, and Predicting the Michigan Consumer Sentiment Index

Big Data, Socio- Psychological Theory, Algorithmic Text Analysis, and Predicting the Michigan Consumer Sentiment Index Big Data, Socio- Psychological Theory, Algorithmic Text Analysis, and Predicting the Michigan Consumer Sentiment Index Rickard Nyman *, Paul Ormerod Centre for the Study of Decision Making Under Uncertainty,

More information

INTRODUCTION TO THE WEB

INTRODUCTION TO THE WEB INTRODUCTION TO THE WEB A beginner s guide to understanding and using the web 3 September 2013 Version 1.2 Contents Contents 2 Introduction 3 Skill Level 3 Terminology 3 Video Tutorials 3 How Does the

More information

How effective is the use of social media by the top ten Scottish Universities?

How effective is the use of social media by the top ten Scottish Universities? How effective is the use of social media by the top ten Scottish Universities? Contents Effective use of social media by the top ten Scottish universities 2015 I. Introduction... 3 II. Methodology... 4

More information

HOW TO BET ON GREYHOUNDS. Gambling can be addictive. Please play responsibly. http://responsiblegambling.betfair.com

HOW TO BET ON GREYHOUNDS. Gambling can be addictive. Please play responsibly. http://responsiblegambling.betfair.com HOW TO BET ON GREYHOUNDS Gambling can be addictive. Please play responsibly. http://responsiblegambling.betfair.com Greyhound racing in the UK takes place every single day (except Christmas Day). The first

More information

Sentiment Analysis of Twitter Feeds for the Prediction of Stock Market Movement

Sentiment Analysis of Twitter Feeds for the Prediction of Stock Market Movement Sentiment Analysis of Twitter Feeds for the Prediction of Stock Market Movement Ray Chen, Marius Lazer Abstract In this paper, we investigate the relationship between Twitter feed content and stock market

More information

BRAND ENGAGEMENT. Introduction to Brand Engagement ATTACHMENT ENGAGE FOCUSED MESSAGE ATTRACTION DEVELOP ADVERTISING ADVERTISING

BRAND ENGAGEMENT. Introduction to Brand Engagement ATTACHMENT ENGAGE FOCUSED MESSAGE ATTRACTION DEVELOP ADVERTISING ADVERTISING Engagement Intelligent Market Research BRAND COMMUNICATION COMMUNICATION ORGANISATION EXPERIENCE CREDIBILITY MESSAGE ATTACHMENT ATTACHMENT RELIABLE MANAGEMENT SOCIAL MEDIA SOCIAL MEDIA CREDIBILITY MEDIA

More information

Email Marketing in Ireland 2011 Email Usage by Irish Consumers and Marketers. April 2011

Email Marketing in Ireland 2011 Email Usage by Irish Consumers and Marketers. April 2011 Email Marketing in Ireland 2011 Email Usage by Irish Consumers and Marketers April 2011 89 Harcourt Street Dublin 2 Tel: + 353 1 475 9286 Email: info@circulator.com Web: www.circulator.com Table of contents

More information

Applying Behavioural Insights to Organ Donation: preliminary results from a randomised controlled trial

Applying Behavioural Insights to Organ Donation: preliminary results from a randomised controlled trial Applying Behavioural Insights to Organ Donation: preliminary results from a randomised controlled trial Introduction Last year over one million people registered to join the NHS Organ Donor Register, bringing

More information

Sentiment analysis on tweets in a financial domain

Sentiment analysis on tweets in a financial domain Sentiment analysis on tweets in a financial domain Jasmina Smailović 1,2, Miha Grčar 1, Martin Žnidaršič 1 1 Dept of Knowledge Technologies, Jožef Stefan Institute, Ljubljana, Slovenia 2 Jožef Stefan International

More information

ANNEX 2 PUBLIC PERCEPTIONS

ANNEX 2 PUBLIC PERCEPTIONS ANNEX 2 PUBLIC PERCEPTIONS Annex 2 A2-1 Annex 2 A2-2 ANNEX 2 PUBLIC PERCEPTIONS This review identified the current social media presences of Defence in order to understand the organisation s official penetration

More information

The Scientific Literature page 1. The Scientific Literature In Psychology

The Scientific Literature page 1. The Scientific Literature In Psychology The Scientific Literature page 1 The Scientific Literature In Psychology As we saw in the reading on science and the scientific method, falsifiability and replication are aspects of the scientific method

More information

CORRALLING THE WILD, WILD WEST OF SOCIAL MEDIA INTELLIGENCE

CORRALLING THE WILD, WILD WEST OF SOCIAL MEDIA INTELLIGENCE CORRALLING THE WILD, WILD WEST OF SOCIAL MEDIA INTELLIGENCE Michael Diederich, Microsoft CMG Research & Insights Introduction The rise of social media platforms like Facebook and Twitter has created new

More information

Research paper London property market snapshot JULY 2015

Research paper London property market snapshot JULY 2015 Research paper London property market snapshot JULY 2015 UK economy The average asking price increased by 3pc between May and June as buyers and sellers reacted to the vote. There was a major surprise

More information

Multiple Kernel Learning on the Limit Order Book

Multiple Kernel Learning on the Limit Order Book JMLR: Workshop and Conference Proceedings 11 (2010) 167 174 Workshop on Applications of Pattern Analysis Multiple Kernel Learning on the Limit Order Book Tristan Fletcher Zakria Hussain John Shawe-Taylor

More information

SOCIAL JOURNALISM STUDY 2012

SOCIAL JOURNALISM STUDY 2012 SOCIAL JOURNALISM STUDY 2012 2012 Social Journalism Study - United Kingdom Report by Cision & Canterbury Christ Church University (UK) www.cision.com 1. EXECUTIVE SUMMARY Key findings: 28.1% of UK journalists

More information

ARLA Members Survey of the Private Rented Sector

ARLA Members Survey of the Private Rented Sector Prepared for The Association of Residential Letting Agents ARLA Members Survey of the Private Rented Sector Fourth Quarter 2013 Prepared by: O M Carey Jones 5 Henshaw Lane Yeadon Leeds LS19 7RW December,

More information

The Viability of StockTwits and Google Trends to Predict the Stock Market. By Chris Loughlin and Erik Harnisch

The Viability of StockTwits and Google Trends to Predict the Stock Market. By Chris Loughlin and Erik Harnisch The Viability of StockTwits and Google Trends to Predict the Stock Market By Chris Loughlin and Erik Harnisch Spring 2013 Introduction Investors are always looking to gain an edge on the rest of the market.

More information

SOCIAL MEDIA PROFILE ANALYSIS. Media Organisations and Journalists

SOCIAL MEDIA PROFILE ANALYSIS. Media Organisations and Journalists SOCIAL MEDIA PROFILE ANALYSIS Media Organisations and Journalists Report developed by Dialogue Consulting. Copyright to the Mindframe National Media Initiative. February 2013. Background to the study The

More information

BETTING SCHOOL INSIDERS NEWS REPORT Honest and Independent Reviews of Betting Products

BETTING SCHOOL INSIDERS NEWS REPORT Honest and Independent Reviews of Betting Products Special Edition Part I 2008 BETTING SCHOOL INSIDERS NEWS REPORT Honest and Independent Reviews of Betting Products Welcome to this the third Special Edition of the Betting School Insiders News report.

More information

Strategies for Effective Wall Posts: A Timeline Analysis

Strategies for Effective Wall Posts: A Timeline Analysis Strategies for Effective Wall Posts: A Timeline Analysis DATA REPORT Introduction 3 Methodology 4 When to Post 5 Best Days to Post By Industry 6 Facebook Interaction is Nocturnal, Post During Non-Busy

More information

school resource CREATE DEBATE THE

school resource CREATE DEBATE THE school resource CREATE THE DEBATE welcome The Create the Debate pack has been produced by Parliament s Education Service with help from the UK Youth Parliament and BBC Three s Free Speech. This easy-to-use,

More information

Credit Card Market Study Interim Report: Annex 4 Switching Analysis

Credit Card Market Study Interim Report: Annex 4 Switching Analysis MS14/6.2: Annex 4 Market Study Interim Report: Annex 4 November 2015 This annex describes data analysis we carried out to improve our understanding of switching and shopping around behaviour in the UK

More information

Social Media Strategy:

Social Media Strategy: Social Media Strategy: The Beginners Guide to Strategy Success in 7 Steps A Simple Guide In association with: Chapter Title Contents Page Page 2 What is Social Media...3 What is a Social Media Strategy

More information

CSE 598 Project Report: Comparison of Sentiment Aggregation Techniques

CSE 598 Project Report: Comparison of Sentiment Aggregation Techniques CSE 598 Project Report: Comparison of Sentiment Aggregation Techniques Chris MacLellan cjmaclel@asu.edu May 3, 2012 Abstract Different methods for aggregating twitter sentiment data are proposed and three

More information

MEASURING GLOBAL ATTENTION: HOW THE APPINIONS PATENTED ALGORITHMS ARE REVOLUTIONIZING INFLUENCE ANALYTICS

MEASURING GLOBAL ATTENTION: HOW THE APPINIONS PATENTED ALGORITHMS ARE REVOLUTIONIZING INFLUENCE ANALYTICS WHITE PAPER MEASURING GLOBAL ATTENTION: HOW THE APPINIONS PATENTED ALGORITHMS ARE REVOLUTIONIZING INFLUENCE ANALYTICS Overview There are many associations that come to mind when people hear the word, influence.

More information

{ { Calculating Your Social Media Marketing Return on Investment. A How-To Guide for New Social Media Marketers. Peter Ghali - Senior Product Manager

{ { Calculating Your Social Media Marketing Return on Investment. A How-To Guide for New Social Media Marketers. Peter Ghali - Senior Product Manager { { Calculating Your Social Media Marketing Return on Investment A How-To Guide for New Social Media Marketers Peter Ghali - Senior Product Manager This guide provides practical advice for developing a

More information

Social Media Analytics

Social Media Analytics Social Media Analytics Ahmed Abbasi University of Virginia 1 Outline Social Media Overview Social Media for Communication and Collaboration Social Media Analytics Application areas Challenges Social Media

More information

INSIGHTS WHITEPAPER What Motivates People to Apply for an MBA? netnatives.com twitter.com/netnatives

INSIGHTS WHITEPAPER What Motivates People to Apply for an MBA? netnatives.com twitter.com/netnatives INSIGHTS WHITEPAPER What Motivates People to Apply for an MBA? netnatives.com twitter.com/netnatives NET NATIVES HISTORY & SERVICES Welcome to our report on using data to analyse the behaviour of people

More information

Santorum leads GOP Twitter race despite Romney poll wins

Santorum leads GOP Twitter race despite Romney poll wins Press Release March 01 2012 FOR IMMEDIATE RELEASE Santorum leads GOP Twitter race despite Romney poll wins Social media analytics firm Brandwatch has been tracking Republican candidates on Twitter since

More information

Strategies for Effective Tweeting: A Statistical Review

Strategies for Effective Tweeting: A Statistical Review Strategies for Effective Tweeting: A Statistical Review DATA REPORT Introduction 3 Methodology 4 Weekends Are Good for Relaxing and Tweeting 5 Best Days to Tweet By Industry 6 When Followers Are Busy Give

More information

An Evaluation of Cinema Advertising Effectiveness

An Evaluation of Cinema Advertising Effectiveness An Evaluation of Cinema Advertising Effectiveness Jason Dunnett and Janet Hoek The purpose of this study was to explore the effectiveness of cinema advertising. Specifically: To quantify the proportion

More information

PPC - Pay Per Click. What it is and how to build a successful campaign. A Publication of Digital Marketing Experts theeword

PPC - Pay Per Click. What it is and how to build a successful campaign. A Publication of Digital Marketing Experts theeword PPC - Pay Per Click What it is and how to build a successful campaign A Publication of Digital Marketing Experts theeword Introduction This guide provides a brief introduction to PPC, helping businesses

More information

Will the NHS be there for me?

Will the NHS be there for me? Will the NHS be there for me? A report by Benenden Healthcare Society Launched at the Palace of Westminster London 11 th July 2012 www.benenden.org.uk Foreword By Ken Hesketh Chief Executive Benenden

More information

What TV Ratings Really Mean

What TV Ratings Really Mean What TV Ratings Really Mean You just heard that a TV Show was ranked #1 in the Nielsen Media Research ratings. What does that really mean? How does Nielsen Media Research "rate" TV shows? Why do shows

More information

Economic background of the Microsoft/Yahoo! case

Economic background of the Microsoft/Yahoo! case Economic background of the Microsoft/Yahoo! case Andrea Amelio and Dimitrios Magos ( 1 ) Introduction ( 1 ) This paper offers an economic background for the analysis conducted by the Commission during

More information

Amp Up Your Email Marketing with Social Media

Amp Up Your Email Marketing with Social Media Amp Up Your Email Marketing with Social Media Social Media www.constantcontact.com 1-866-876-8464 INSIGHT PROVIDED BY 2011 Constant Contact, Inc. 10-1722 Great Together Social media networks such as Facebook,

More information

Scalping Trade Strategies April 14, 2009

Scalping Trade Strategies April 14, 2009 www.fxcm.com sales@fxcm.com Scalping Trade Strategies April 14, 2009 DailyFX Research Team John Kicklighter David Rodriguez Currency Strategists 1-212-897-7660 1-888-50-FOREX (36739) Just a decade ago,

More information

Prediction of Stock Performance Using Analytical Techniques

Prediction of Stock Performance Using Analytical Techniques 136 JOURNAL OF EMERGING TECHNOLOGIES IN WEB INTELLIGENCE, VOL. 5, NO. 2, MAY 2013 Prediction of Stock Performance Using Analytical Techniques Carol Hargreaves Institute of Systems Science National University

More information

SMEs and the communications market: 2006

SMEs and the communications market: 2006 SMEs and the communications market: 2006 Publication date: June 2006 Contents Section Page 1 Background, research objectives and approach 1 2 Executive summary small businesses and sole traders 3 Small

More information

Pivot Trading the FOREX Markets

Pivot Trading the FOREX Markets Pivot Trading the FOREX Markets A report by Jim White April, 2004 Pivot Research & Trading Co. 3203 Provence Place Thousand Oaks, Ca. 91362 Phone: 805-493-4221 FAX: 805-493-4349 Email: Jwhite43@adelphia.net

More information

Network Rail October 2007 Strategic Business Plan. Supporting Document. Demand Forecasting in the SBP

Network Rail October 2007 Strategic Business Plan. Supporting Document. Demand Forecasting in the SBP Network Rail October 2007 Strategic Business Plan Supporting Document 2 Executive Summary The capacity strategy in the SBP is drawn from a range of sources. The HLOSs specify a small number of key schemes

More information

Measure Social Media like a Pro: Social Media Analytics Uncovered SOCIAL MEDIA LIKE SHARE. Powered by

Measure Social Media like a Pro: Social Media Analytics Uncovered SOCIAL MEDIA LIKE SHARE. Powered by 1 Measure Social Media like a Pro: Social Media Analytics Uncovered # SOCIAL MEDIA LIKE # SHARE Powered by 2 Social media analytics were a big deal in 2013, but this year they are set to be even more crucial.

More information

smart. uncommon. ideas.

smart. uncommon. ideas. smart. uncommon. ideas. Executive Overview Your brand needs friends with benefits. Content plus keywords equals more traffic. It s a widely recognized onsite formula to effectively boost your website s

More information

Would You Like To Earn $1000 s With The Click Of A Button?

Would You Like To Earn $1000 s With The Click Of A Button? Would You Like To Earn $1000 s With The Click Of A Button? (Follow these easy step by step instructions and you will) This Version of the ebook is for all countries other than the USA. If you need the

More information

The Educational Backgrounds of Members of the House of Commons and House of Lords

The Educational Backgrounds of Members of the House of Commons and House of Lords The Educational Backgrounds of Members of the House of Commons and House of Lords December 2005 1 Key Findings House of Commons Overall, almost one third (32%) of current MPs attended independent schools,

More information

Let s Get to Know Spread Bets

Let s Get to Know Spread Bets Let s Get to Know Spread Bets Spread betting is pretty cool. Here are three reasons why. Even if you ve never traded before, you probably know how the financial market works buy in and hope it goes up.

More information

Using Twitter to Build a Case for Social Media

Using Twitter to Build a Case for Social Media Using Twitter to Build a Case for Social Media Karin Basaraba / Insurance Corporation of British Columbia / North Vancouver, British Columbia NEED/OPPORTUNITY COMPANY OVERVIEW The Insurance Corporation

More information

M EASUREM ENT GUIDE. Googl

M EASUREM ENT GUIDE. Googl M EASUREM ENT GUIDE Ul t i ma t e Gui det o Me a s ur i ng Googl e + j effal yt i cs. com The Ultimate Guide to Measuring Google Plus 2013 Jeff Sauer www.jeffalytics.com jeff@jeffalytics.com - 1 Table of

More information

Interview: Professor Adrian Payne. Handbook of CRM: Achieving Excellence in Customer Management.

Interview: Professor Adrian Payne. Handbook of CRM: Achieving Excellence in Customer Management. Interview: Handbook of CRM: Achieving Excellence in Customer Management. Hello, this is Steve Macaulay from Cranfield School of Management. I am here today to discuss with his book Handbook of CRM: Achieving

More information

PAYMENT PROTECTION INSURANCE RESEARCH

PAYMENT PROTECTION INSURANCE RESEARCH PAYMENT PROTECTION INSURANCE RESEARCH ANALYTICAL REPORT NOVEMBER 2015 ABOUT COMRES ComRes provides specialist research and insight into reputation, public policy and communications. It is a founding member

More information

INVESTING & TRADING WITH THE JSE PRICE/EARNINGS (P/E) RATIO

INVESTING & TRADING WITH THE JSE PRICE/EARNINGS (P/E) RATIO INVESTING & TRADING WITH THE JSE PRICE/EARNINGS (P/E) RATIO "At the end of the day, all stock price movements can be traced back to earnings." Page 1 INTRODUCTION This e-book shows how to use the daily

More information

Italian Journal of Accounting and Economia Aziendale. International Area. Year CXIV - 2014 - n. 1, 2 e 3

Italian Journal of Accounting and Economia Aziendale. International Area. Year CXIV - 2014 - n. 1, 2 e 3 Italian Journal of Accounting and Economia Aziendale International Area Year CXIV - 2014 - n. 1, 2 e 3 Could we make better prediction of stock market indicators through Twitter sentiment analysis? ALEXANDER

More information

Whose trust matters most? Paper for the session Meeting expectations from users within resources IAOS Conference Da Nang, Vietnam, 8-10 October 2014.

Whose trust matters most? Paper for the session Meeting expectations from users within resources IAOS Conference Da Nang, Vietnam, 8-10 October 2014. Whose trust matters most? Paper for the session Meeting expectations from users within resources IAOS Conference Da Nang, Vietnam, 8-10 October 2014. Richard Laux and Richard Alldritt The importance of

More information

Impact of the recession

Impact of the recession Regional Trends 43 21/11 Impact of the recession By Cecilia Campos, Alistair Dent, Robert Fry and Alice Reid, Office for National Statistics Abstract This report looks at the impact that the most recent

More information

Study into the Sales of Add-on General Insurance Products

Study into the Sales of Add-on General Insurance Products Study into the Sales of Add-on General Insurance Quantitative Consumer Research Report Prepared For: Financial Conduct Authority (FCA) March, 2014 Authorised Contact Persons Frances Green Research Director

More information

Assessing the Effects of Buybacks on Investment Trust Discounts. Faculty of Actuaries Investment Research Group

Assessing the Effects of Buybacks on Investment Trust Discounts. Faculty of Actuaries Investment Research Group Assessing the Effects of Buybacks on Investment Trust Discounts Faculty of Actuaries Investment Research Group Andy Adams, Roddy Macpherson, Brian Moretta Abstract: Buybacks for investment trusts have

More information

Business Expectations Survey

Business Expectations Survey Business Expectations Survey Dun & Bradstreet Q1 2016 FINAL RESULTS RELEASED 5 JANUARY 2016 Index CapEx plans up despite low expectations The results from Dun & Bradstreet s December Business Expectations

More information

Understand how PPC can help you achieve your marketing objectives at every stage of the sales funnel.

Understand how PPC can help you achieve your marketing objectives at every stage of the sales funnel. 1 Understand how PPC can help you achieve your marketing objectives at every stage of the sales funnel. 2 WHO IS THIS GUIDE FOR? This guide is written for marketers who want to get more from their digital

More information

Rating Systems for Fixed Odds Football Match Prediction

Rating Systems for Fixed Odds Football Match Prediction Football-Data 2003 1 Rating Systems for Fixed Odds Football Match Prediction What is a Rating System? A rating system provides a quantitative measure of the superiority of one football team over their

More information

Stock market simulation with ambient variables and multiple agents

Stock market simulation with ambient variables and multiple agents Stock market simulation with ambient variables and multiple agents Paolo Giani Cei 0. General purposes The aim is representing a realistic scenario as a background of a complete and consistent stock market.

More information

Can Twitter Predict Royal Baby's Name?

Can Twitter Predict Royal Baby's Name? Summary Can Twitter Predict Royal Baby's Name? Bohdan Pavlyshenko Ivan Franko Lviv National University,Ukraine, b.pavlyshenko@gmail.com In this paper, we analyze the existence of possible correlation between

More information

SOCIAL MEDIA DID YOU KNOW: WHAT IS SOCIAL MEDIA? IGNORE IT AT YOUR PERIL! WWW.SPORTENGLANDCLUBLEADERS.COM ANYWHERE GETTING GREYER

SOCIAL MEDIA DID YOU KNOW: WHAT IS SOCIAL MEDIA? IGNORE IT AT YOUR PERIL! WWW.SPORTENGLANDCLUBLEADERS.COM ANYWHERE GETTING GREYER The world s top brands are using social media as a meaningful way of deepening relationships with their customers. Is it now time for Sport Clubs to join in? WHAT IS SOCIAL? According to the Chartered

More information

Marketing Mix Modelling and Big Data P. M Cain

Marketing Mix Modelling and Big Data P. M Cain 1) Introduction Marketing Mix Modelling and Big Data P. M Cain Big data is generally defined in terms of the volume and variety of structured and unstructured information. Whereas structured data is stored

More information

Gambling Commission. Trends in Gambling Behaviour 2008-2014

Gambling Commission. Trends in Gambling Behaviour 2008-2014 Gambling Commission Trends in Gambling Behaviour 2008-2014 Report Name Prepared for Authors Project team Trends in Gambling Behaviour 2008-2014 Gambling Commission Doros Georgiou Karl King Adrian Talbot

More information

CHAPTER 11: THE EFFICIENT MARKET HYPOTHESIS

CHAPTER 11: THE EFFICIENT MARKET HYPOTHESIS CHAPTER 11: THE EFFICIENT MARKET HYPOTHESIS PROBLEM SETS 1. The correlation coefficient between stock returns for two non-overlapping periods should be zero. If not, one could use returns from one period

More information

90 Digital igaming Reports: SEO Risk in igaming. Risk of being hit by a Google update has doubled in 17 months. June 2013 - Gillis van den Broeke

90 Digital igaming Reports: SEO Risk in igaming. Risk of being hit by a Google update has doubled in 17 months. June 2013 - Gillis van den Broeke 90 Digital igaming Reports: SEO Risk in igaming Risk of being hit by a Google update has doubled in 17 months June 2013 - Gillis van den Broeke About the Author Gillis van den Broeke is the Director of

More information

Stock Returns Following Profit Warnings: A Test of Models of Behavioural Finance.

Stock Returns Following Profit Warnings: A Test of Models of Behavioural Finance. Stock Returns Following Profit Warnings: A Test of Models of Behavioural Finance. G. Bulkley, R.D.F. Harris, R. Herrerias Department of Economics, University of Exeter * Abstract Models in behavioural

More information

Will 16 and 17 year olds make a difference in the referendum?

Will 16 and 17 year olds make a difference in the referendum? Will 16 and 17 year olds make a difference in the referendum? 1 Author: Jan Eichhorn Date: xx/11/2013 Comparisons between the newly enfranchised voters and the adult population 2 ScotCen Social Research

More information

Online Media Kit 2014-FCC_OnlineMediaKit 12/4/2014 8:56 AM Page 1 nline Odvertising A

Online Media Kit 2014-FCC_OnlineMediaKit 12/4/2014 8:56 AM Page 1 nline Odvertising A Online Advertising The Network Forum Communications Company online advertising network includes 46 websites across Minnesota, Wisconsin, North Dakota, and South Dakota. Our entire network reaches 2+ Million

More information

Creative Industries Economic Estimates January 2015

Creative Industries Economic Estimates January 2015 Creative Industries Economic Estimates January 2015 Statistical Release Date: 13/01/2015 The Creative Industries Economic Estimates are Official Statistics and have been produced to the standards set out

More information

Underutilization in U.S. Labor Markets

Underutilization in U.S. Labor Markets EMBARGOED UNTIL Thursday, February 6, 2014 at 5:45 PM Eastern Time OR UPON DELIVERY Underutilization in U.S. Labor Markets Eric S. Rosengren President & Chief Executive Officer Federal Reserve Bank of

More information

Carpe diem - Seizing the moment in Social Media. How to maximise website traffic using social media during major global events

Carpe diem - Seizing the moment in Social Media. How to maximise website traffic using social media during major global events Carpe diem - Seizing the moment in Social Media How to maximise website traffic using social media during major global events Contents Executive Summary Page 3 An update on Social Media and Online News

More information

EVALUATI G TV A D PRI T SALES EFFECTS

EVALUATI G TV A D PRI T SALES EFFECTS Worldwide Readership Research Symposium 2001 Session 10.8 EVALUATI G TV A D PRI T SALES EFFECTS Bruce Goerlich, Starcom MediaVest Group Acknowledgement Many thanks to Simon Broadbent for his help and insights.

More information

17 of the Internet s Best Banner Ads. Love em or Hate em They Do Work!

17 of the Internet s Best Banner Ads. Love em or Hate em They Do Work! Love em or Hate em They Do Work! Banner Ads What are they? Ever since the Internet started to take off in the mid 90 s, banner ads have been an acceptable way of advertising on the Web. Banner ads come

More information

Carbon Dating the Web

Carbon Dating the Web Carbon Dating the Web: Estimating the Age of Web Resources Hany M. SalahEldeen & Michael L. Nelson Old Dominion University Department of Computer Science Web Science and Digital Libraries Lab. Hany SalahEldeen

More information

How to Integrate Email Marketing With Your Social Media Efforts

How to Integrate Email Marketing With Your Social Media Efforts How to Integrate Email Marketing With Your Social Media Efforts Here are the 4 methods of integrating social media with your email marketing. 1. Put faces to your email subscribers and retarget them All

More information

Reading Gaps in Charts to Find Good Trades

Reading Gaps in Charts to Find Good Trades Reading Gaps in Charts to Find Good Trades One of the most rewarding and challenging things I have done in my 40-year trading career is teach elementary school students the basics of technical analysis

More information

Lecture 8: Stock market reaction to accounting data

Lecture 8: Stock market reaction to accounting data Lecture 8: Stock market reaction to accounting data In this lecture we will focus on how the market appears to evaluate accounting disclosures. For most of the time, we shall be examining the results of

More information

Stock Volatility during the Recent Financial Crisis

Stock Volatility during the Recent Financial Crisis Stock Volatility during the Recent Financial Crisis G. William Schwert William E. Simon Graduate School of Business Administration, University of Rochester and the National Bureau of Economic Research

More information

SOCIAL MEDIA PROPOSAL

SOCIAL MEDIA PROPOSAL SOCIAL MEDIA PROPOSAL Prepared for: NSVRC July 22, 2011 OBJECTIVE To create a branded online presence through the use of social media that will allow NSVRC to: Connect to a wider audience Spread its message

More information

Internet Appendix to Momentum, Reversal, and Uninformed Traders in Laboratory Markets * 1. I. Instructions for Experiment One

Internet Appendix to Momentum, Reversal, and Uninformed Traders in Laboratory Markets * 1. I. Instructions for Experiment One Internet Appendix to Momentum, Reversal, and Uninformed Traders in Laboratory Markets * 1 I. Instructions for Experiment One Overview During this session, you will trade shares in 12 different securities

More information

A Short Study on Security Indicator Interfaces

A Short Study on Security Indicator Interfaces A Short Study on Security Indicator Interfaces Technical Report UCSC-WASP-15-04 November 2015 D J Capelis mail@capelis.dj Working-group on Applied Security and Privacy Storage Systems Research Center University

More information

The Credit Card Report May 4 The Credit Card Report May 4 Contents Visa makes no representations or warranties about the accuracy or suitability of the information or advice provided. You use the information

More information

Numbers as pictures: Examples of data visualization from the Business Employment Dynamics program. October 2009

Numbers as pictures: Examples of data visualization from the Business Employment Dynamics program. October 2009 Numbers as pictures: Examples of data visualization from the Business Employment Dynamics program. October 2009 Charles M. Carson 1 1 U.S. Bureau of Labor Statistics, Washington, DC Abstract The Bureau

More information