How can I avoid the survivor affect when doing backtesting with stocks?
January 30th, 2010 | by admin |I am trying to write a program to backtest different strategies against historical data, but the problem I am running into is factoring in the affect of turnover, in particular the effect of delisting. For example, nasdaq composite appears to have a delisting rate, with over 400 delistings for at least one year (1999). I am using data imported from yahoo finance but the lists provided there do not include delisted stocks. Does anyone know where I can get similar historical price data and stock lists for delisted stocks? Or alternately, what is the best index with the lowest turnover?
If you use a larger cap index, you will reduce the number (but not eliminate) of delisting.
Fewer mid and large caps get delisted than small caps.
You could go S&P 500 and that would give you smaller sample size than NASDAQ, but a lower amount of delistings.
3 Responses to “How can I avoid the survivor affect when doing backtesting with stocks?”
By Taranto on Jan 30, 2010 | Reply
you can get this information from the CRSP database. Most academic studies are done using these data.
WRDS (Wharton Research Data Services) at U of PA has these data. It costs money to get it — but if you are at a university — it is likely that your finance department subscribes. If you are out od school, but have an MBA, then you may be able to access it through the school’s library.
References :
By tsbr1963 on Jan 30, 2010 | Reply
If you use a larger cap index, you will reduce the number (but not eliminate) of delisting.
Fewer mid and large caps get delisted than small caps.
You could go S&P 500 and that would give you smaller sample size than NASDAQ, but a lower amount of delistings.
References :
By NC on Jan 30, 2010 | Reply
If you are looking for professional-quality data, you have to be prepared to pay professional-level price. Typically, researchers use either CRSP (Center for Research of Security Prices) or Compustat data.
As to indexes with low turnover, check out S&P 500 and Russell 1000.
References :