I am trying to test some asset pricing models on 10 portfolios for the period of 2010-2020. The problem is that three of these portfolios included stocks that are newly listed in 2017 and 2018, so I only have three years of historical data for these stocks. Can anyone help how to solve this problem? I don't want to exclude these stocks; does imputation makes sense for such a long period?

  • $\begingroup$ The same on Cross Validated: here. $\endgroup$ May 5 at 9:43

I have found some posts to help answer your question. In a post on ResearchGate.net,

One of the responses that I think has the best answer for your is to carry forward the previous data . However, as the person stated, there is a problem with that method. “--- if you are developing a trading system, for example, and assume that prices are "tradable", then your analysis will be flawed since you can't trade at a "carried forward price." My solution to this problem is this: create a 'locf' price and as a separate variable, a "tradable price", which when data is missing, is (approximately) the next price you could have realized. “

However, Mazin A. M. Al Janabi posted a link on that page which has a better answer for you. https://www.researchgate.net/profile/Mazin-Al-Janabi/post/How-to-deal-with-missing-value-in-a-time-series-stock-market-data/attachment/5b73a9633843b0024405fcb1/AS%3A659744100450305%401534306659149/download/Dealing+with+the+Problem+of+Missing+Data-+A+short+Note.pdf He states that you either not include the missing data or you can calculate the price from a geometric mean of the previous and other price data. However using average prices will underestimate the volatility of the stock price data.

Here is another response to a similar question on Stackexchange: https://quant.stackexchange.com/questions/41171/how-to-deal-with-missing-value-in-a-time-series-stock-market-data

I have also posted a similar answer on missing data but not missing stock price data How do I fill gaps in my data?

Either case I think it is VERY important that you document that you have missing data, report which observations are missing, and what method you used to remedy it.

  • 1
    $\begingroup$ Thank you so much for your reply. I will go through all your reading suggestions. Much appreciated. $\endgroup$
    – Sima
    May 3 at 23:32
  • $\begingroup$ Awesome. Good luck with your project. $\endgroup$
    – Mike J
    May 4 at 4:14

Your Answer

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

Not the answer you're looking for? Browse other questions tagged or ask your own question.