Dear Wilmoters,I'm writing a paper on the topic of assessing the likelihood of having news in the financial market based solely on trade data. Here's the abstract:Abstract: In this paper we look into the problem of identifying the strength of the incoming of news in the financial market. With the support of a microstructure model we are able to derive a simple formula that can estimate the likelihood of having news for any given tradable asset in a particular time period, based only on trade data. The formula can be easily implemented and takes just one input, the probability of a zero trade price differences conditional on the incoming of concessutive same sign trades. In the empirical part of the paper we investigate the properties of this proposed estimator of news intensity for twenty stocks from the Brazilian equity market, convering two full years from 2010 to 2012. The results are very encouraging and consistent across assets. First we find that the strength of news have a common component across all assets. We attribute that to the fact that the incoming of new information regarding the Brazilian economy will affect all stocks. We also see that the likelihood of news is strongly and positively related to volatility of price differences and negatively related to trade volume. The first can be explained by the fact that volatility is a bi-product of news and the second by the presence of traders avoiding the disclosure of private information by trading smaller volumes.The paper can be downloaded Here.I would really apreciate for some feedback on it. Negative or positive reviews are very welcome.PS: I has been a couple of years since I don't post anything here. I used to be a lot more active while I was a PhD student (today I am a full time academic). I'm glad to see that the guys from my time are still around.best,Marcelo.

Hi msperlin,It is an interesting paper. I would include VALE3, PETR4, USIM3 also to see whether the results are the same for different classes of stocks issued by the same company.The following step would be to group companies by sector (I would add BRAP4 to VALE3 and VALE5 and would avoid putting PETR4 and OGXP3 together though) and see whether they present a common reaction (e.g. today large co-movement of USIM5 and CSNA3 based on import taxes announced yesterday).I guess gaps from close to open would not be measured by the methodology, but they could indicate news as well.Best,MCarreira

QuoteOriginally posted by: MCarreiraHi msperlin,It is an interesting paper. I would include VALE3, PETR4, USIM3 also to see whether the results are the same for different classes of stocks issued by the same company.The following step would be to group companies by sector (I would add BRAP4 to VALE3 and VALE5 and would avoid putting PETR4 and OGXP3 together though) and see whether they present a common reaction (e.g. today large co-movement of USIM5 and CSNA3 based on import taxes announced yesterday).I guess gaps from close to open would not be measured by the methodology, but they could indicate news as well.Best,MCarreiraThanks for the reply MCarreira.Yes, this is the next step. I'm going to separate the data into stocks from the same industry. Ideally, there should be a higher correlation within those since industry related news can affect a particular group only. Bovespa already provides a formal segmentation of the stocks by industry, so no problem there as I can just follow it.The close-open gaps are not there explicitally, but you could think of it as the volatility of the innovations in the efficient price process. In the paper I only look at the probability of having or not news (and not how intense (high price change) it is).

Tell me if this is what you are saying. If a stock goes up 1% with only 10 consecutive ticks at the offer, then a news search will almost surely show some news for that ticker symbol, or some general market-moving news. But if the same stock goes up 1% after ticking at the offer 98 out of 100 ticks, probably not, probably just a big buyer.

QuoteOriginally posted by: farmerTell me if this is what you are saying. If a stock goes up 1% with only 10 consecutive ticks at the offer, then a news search will almost surely show some news for that ticker symbol, or some general market-moving news. But if the same stock goes up 1% after ticking at the offer 98 out of 100 ticks, probably not, probably just a big buyer.Not necessarilly. It is all about what happens to the trade prices in a sequence of buys and sells. For the first case, it depends on how the trade price changed in between the trades (ticks). If you have, lets say for a day, 10 trades at the offer (all buys) and each of them moved the trade price by 1%, then for this day the models says that there is news 100% of the time. pnew=1 - 0/10For the second example, if you have 98 consecutive buys and 2 consecutive sells, and the trade price has changed for 50 of the consecutive (same sign) 99 trades, then the probability of having news in this particular day is:pnews=1- 50/99 = 49%