Not logged in.
Quick Search - Contribution
|Title||Explaining co-movement with co-mentions in financial media|
|Institution||University of Zurich|
|Faculty||Faculty of Economics, Business Administration and Information Technology|
|Number of Pages||44|
|Abstract Text||The aim of this work was to study the relation between nancial news and co-movement of stocks. Until now research on nancial media focused mostly on relationship between textual data and characteristics of single companies, e.g. trying to determine, if news sentiment has predictive power for stocks' returns. My goal was to explore a dierent eld - I wanted to analyze the co-existence of companies in nancial news and check if it has any connection to co-movement between their returns. Co-movement of asset prices has already been widely covered in literature, going back to the works of Shiller (1989), who analyzed co-movement in stock prices and dividends or Rotemberg (1990), who analyzed co-movement of commodity prices. Many theories have evolved so far, but to my knowledge, nobody has analyzed the relation between co-movement and nancial news. I wanted to nd out whether one can identify any aliation between these two, seemingly unrelated types of data. Analysis of co-occurrence in nancial news - with all its implications - can later be used in practical investment decisions, e.g. concerning portfolio diversication. If this is in fact the case, that one can predict changes in return correlations using news about companies, one could early identify any shifts and adjust portfolio positions to prot from better diversied collection of stocks. I approached this problem from two dierent directions. First, using a well-known phenomenon of index inclusions I tested, if this exogenous shock causes increased co-movement between stocks. The explanation behind co-movement of stock returns, according to the traditional theory of nance, is straightforward - as long as there are no changes to common factors determining present values of two companies, there should be no excess co-movement detectable. This means, there should be no increase in co-movement between the company added to the S&P 500 index and its other members, since addition to the index does not change fundamentals (the committee deciding about changes to the index uses several criteria, most importantly nancial soundness of the company to be added, but this assessment is made based on public information available before the actual inclusion, therefore it should already be re ected in the price and the decision of the S&P committee should have no eect). It was already shown by Barberis et al. (2005), that companies added to the S&P 500 experience increased co-movement with other S&P 500 members, which stands against the traditional view. However, to draw any conclusions in my work, regarding the relation between co-movement and news, I had to check, if I can nd similar results in my data sample. My results in this area were in line with what was reported previously - I showed that companies added to the S&P 500 between 2003 and 2013 experienced increased co-movement with other members of the index, as well as decreased co-movement with companies not included in the index. After analyzing stocks' returns around these events, I moved to analysis of news data over the same periods. I checked news coverage, co-mentions of "to-be-included" companies with other S&P 500 members as well as news sentiment. The second direction, from which I tackled the relation between news and co-movement was by analyzing a sample of companies without the exogenous shock (of S&P 500 inclusion). I focused on relations between companies that were S&P 500 members between 2003 and 2013. Using all possible pairs within my sample I checked if there's a positive eect between changes in news similarity and changes in return correlation. Within the scope of my work I present the following results: companies included in the S&P 500 index (between 2003 and 2013) experience statistically signicant increase in loading on S&P 500 index returns and decrease in loading on non-S&P 500 index returns, which conrms results of Barberis et al. (2005), using an independent data sample companies joining S&P 500 experience increase in news coverage around the inclusion companies joining S&P 500 experience increase in news coverage during the period before the inclusion, although not as substantial companies joining S&P 500 experience increase in number of co-mentions with other S&P 500 members around the inclusion subsample of companies with most substantial increase in news coverage around inclusion into S&P 500 experiences the most substantial increase in with S&P 500 (co-movement) calculated using univariate regression model subsample of companies with most substantial increase in co-mentions around inclusion into S&P 500 experiences the most substantial increase in with S&P 500 (co-movement) calculated using bivariate regression model companies joining S&P 500 experience increase in negative sentiment of news, which supports previous results from literature, saying companies added to the S&P 500 have worse operating performance after inclusion, compared to period before inclusion pair-wise analysis of companies shows there's a positive eect of increased news similarity on return correlation|