Constructing a historical universe for backtesting

[NYDreamer]

For accurate backtesting, one needs to define his/her historical universe.

Suppose I want to trade small and micro cap stocks only, so that I define my universe below a certain market cap limit. For accurate backtesting, I would need either historical market caps, or historical components of a certain small cap index that would do the job for me. I suppose such data is not commonly included in historical price packages.

So even when I have data without survivorship bias (i.e. including delisted stocks), I wouldn't know how to filter the stocks that I would've actually traded historically.

I wonder how others deal with this problem? Taking current market caps (effectively introducing huge biases)? Somehow constructing a proxy for historical market cap?

I was surprised to find little discussion on this topic.

 

[rolando87]

Try to see if you can find historical balance sheets for these small companies. Once you have that, find the line that says total shares outstanding, multiply that number by the stocks historical price. That'd be one slow way of doing it.

You could get a rough avg market cap for that year by taking the avg price of the stock that year and multiplying it by the total shares outstanding that year.