.jpg)
I wanted to examine more closely the timeline of the public reaction and the reaction in the stock market to the swine flu. This time series plot compares a number of different data series. The Yahoo News count, the global confirmed swine flu cases, my own swine flu stock index (see the end of the post for what that index is and how I came up with it), Google Trends search volume for terms relating to flu "stocks" and then the Google Trends search volume for swine flu.If you click on the plot and view it bigger you can see the events play out. The Yahoo News count was the first to show signs of anything on Thursday April 23. Stories of swine flu had been circulating since the first of April but on the 23rd they seemed to edge above where they had been. Then on Friday, in addition to a much clearer increase in the Yahoo News count, a number of things popped up including a 26% gain in my swine stock index, the first release of confirmed swine flu cases, and "swine flu" first showing up on the radar at Google Trends. Over the weekend things escalated dramatically with Yahoo News skyrocketing, search volumes going way up and then on Monday, my swine stock index opened 60% up.
So when I look at all of this, it appears as though the Yahoo News count metric had a slight 1 day lead in pointing to this coming. But it was a very small lead. Googlers first started searching for "swine flu" and then the following day started thinking about making money on it and started searching for "swine stocks". For the stock trader, it all happened so fast and over the weekend so it was really not playable. The only opportunity would have been Friday had one noticed the first signs, clearly some did as my swine stock index was up 26% that day.
My Swine Stock Index
To come up with my swine stock index (which was all after the fact of course since I would not have thought to have a swine stock index before all this) I started with a list of stocks from various sources listing swine flu stocks such as this. Then I did a multivariate technique called principal component analysis to figure out which subgroup of stocks within this list was most correlated to each other because I want to have a concentrated swine flu index.
This shows a loading plot which I won't try to explain today but it is a cool graphical technique that allows you to see easily which variables are most correlated to each other and to identify natural groupings in the data. If you look at it you will see a grouping to the right consisting of the stocks NVAX, PURE, BCRX, AVII, GNBT, and SVA. Those 6 stocks are highly correlated. So I combined them and made my own index by averaging the daily % movements of the stocks. That produced the index you see in the above time series plot. Going forward, if I were to try to do any stock trading on the swine flu, those 6 stocks would be the ones I would consider. I am not sure exactly what my strategy would be but I think I will watch the Yahoo News count especially to see if it goes higher (which would be bullish for these stocks from a short term trader/investor behavioral standpoint) or goes lower (which would be bearish).
...
2 comments: