My current test is to take monthly proportional price changes for stock XYZ and subtract out the proportional changes of the S&P500. Then compare the mean of a sample of XYZ-S&P (e.g. trailing 12 months) to the mean of the population (48 months preceding the sample). The test is too stringent, as I can not show significant difference even when cherry-picking stocks that are going gangbusters. The method described is derived from a book where the author uses a 6-month interval and data spanning decades.

I want my test to identify significant changes in stock price on a shorter time scale. Can you give me some guidance? Your insights are greatly appreciated.

  • $\begingroup$ In the research community they solve this problem by aggregating lots of stocks together. For example to study what happens when a company announces good earnings they will combine data for thousands of stocks together after aligning the earnings release dates. So you need a lot of data. $\endgroup$ – Alex C Mar 17 '16 at 23:15
  • $\begingroup$ Thanks for your comment Alex. Is there a statistical method I can use on individual stocks, though? I am thinking along the lines of a screening tool that would tell me "average monthly change in XYZ over the last 12 months is significantly higher than in the preceding 48 months." $\endgroup$ – jklaus Mar 18 '16 at 1:11
  • $\begingroup$ You could calculate the beta of the stock vs the index, or more precisely a beta range using a confidence level of x% (say 95%) - you can then check if the return of the stock is within the range predicted by the betas and if not conclude that the stock under/out-performed the index at the x% confidence level. $\endgroup$ – assylias Mar 18 '16 at 13:09

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.