Small error in S&P 500 Total Return data

Recently a long time series was needed to use in some exercises and in this case the choice was made to use the the default closing Price (PI) and the Total Return (RI) data for the S&P 500 index. The source for the data was Datastream from Thomson Reuters. The price data from Datastream for the S&P 500 index goes back to December 31st of 1963. The Total Return data from Datastream for this index goes back to January  1st of 1988.

While working with the data a strange outlier in the data was discovered for January 24th 1990. Statistically speaking, the calculated return made a huge jump there. See the example here:

The data was then compared to data from Yahoo Finance (using the Quandl add-in to do a quick download to excel). Using the Yahoo Finance data to calculate the Total Return again this appeared to give a comparable result to the manually calculated RI using the Price data for the index from Datastream.
After contacting the Thomson Reuters helpdesk it became clear that there was a small error in their spreadsheet that is used to calculate the RI: for January 24th an incorrect price was used for the RI calculation: 324,17 instead of 330,26.

Unfortunately, Thomson Reuters will not change this small error for their calculated RI. If you need to go back and have to use the data from that time you might use a corrected RI for this day which could be: 354,53. This number is more in line with statistical variance for the return over a period of 25 year or more. See also:

This goes to show that it always pays off to check for outliers and (when in doubt) to contact the vendor/source and discuss possibilities.


%d bloggers like this: