3

I'm pursuing an academic research project aiming to look at the effects of news coverage on a company's stock price. In order to do so, I hope to index coverage of an event into a single time-series variable tracking coverage across the time horizon of the event. Then, I'd like to use that time-series variable to help predict stock price, which of course is also a time-series variable in itself, but considering that the effects of news coverage on the stock will likely die out before the financial effects are priced in. Therefore, I am running into a couple of issues:

  • First of all, there is simultaneity between stock price and news, in that news can affect the stock price but drastic stock price movements can trigger news events as well.
  • The scales of the two time-series variables I am planning on working with are different, and I have yet to learn in the classroom how to handle such situations.

I am really just struggling in terms of figuring out what type of analytical framework I am going to end up pursuing, and exactly what types of models I want to run. I bit off a lot with this project, and once I am pointed in the right direction I think I can pull it off, but need that initial push.

Shayan Shafiq
  • 633
  • 6
  • 17
Jason
  • 31
  • 1

1 Answers1

1

You could take a look at vector autoregression (VAR; note that "VaR" can also stand for "value at risk" in similar contexts, which is not what you are looking for). VAR is precisely used for modeling time series that influence each other, typically in a macroeconomic context.

The remaining problem is that it really presumes normally distributed innovations, which may not describe your news coverage well (though it should be a better approximation for the stock prices).

Stephan Kolassa
  • 95,027
  • 13
  • 197
  • 357