IMPORTANT: industry classification is now historically unbiased

Latest Release
1

This upgrade can generate different results when a sim is re-run

Dear All,

We’ve made a major upgrade to the industry classification so that it does not suffer from historical biases. Previously we had made an assumption that a company stays in an industry forever. When the industry changed it was simply updated. This started causing problems as more and more stocks were changing industry and sector. Industry changes are due to several reasons: a merger, a correction from Reuters, a company diversifies via acquisitions or sale of portions of the business. A notable example is IBM which recently changed from Computer Hardware to Business Services. This change occured on 4/22/06 and was probably due to the sale of all their hardware divisions (for example the Thinkpad line to Lenovo).

Industry changes have a major impact on historical simulations which could explain why re-running an old sim generated different results. It affects a simulation when: Industry factors are used in the ranking, Buy/Sell rules like IndWeight and SecWeight and using the Sector universes.

We believe we’ve corrected this problem. You can test this in the screener by entering the following rule:

Industry=CMPTRS

If you run this screen on 4/21/06 or earlier, IBM will come up in the results. When you run the screen on 4/22/06 or later IBM will not come up. It will come up with the rule Industry = SVSBUS

This has major implications if you re-run your sims. It should not affect any “live” portfolios because they use the latest industry classification. It should only affect historical sims/screens and ranks.

Sorry for the additional analysis this imposes. Please let us know in earnest if you experience problems. It was very tricky to rebuild historical industry classification. I’m not sure anyone even sells this data. We were able to do it because we kept all the old downloads of data which had to be re-processed to extract the historically correct industry for each stock.

FYI: in almost 6 years we’ve come up with around 500 stocks that have changed industry, which is a considerable number.

1 Like
2

I did not expect this to be such a big change but the first sim I re-ran totally blew up - less than half the return [:((] . The only Industry rule it had was InstOwn% < InstOwn%Ind (syntax?). Another sim with Price4Wk%Ind (syntax?) > 0 also was affected quite a bit. The others were not affected that much or at all.

So it is advisable to re-run your sims.

Anyway, good work Marco for correcting the data.

KJ

3

How many positions were in the sim? The smaller the number of stocks the bigger the impact.

4

I think sims or ranking systems with this factor Pr13WRel%ChgInd (or similar ones) are also impected. I rerun my favorite sims and so far I have seem only minor changes (a couple have better results).

5

My sim with major blow up had 6 stocks.

6

Hi, this thread helped me with a question regarding Transactions > Realized > Aggregate > By Sector. I wondered why realized trades appeared that originate from sectors which I actively excluded in the simulation rules and/or universe. I suspected that it's because of sector restatement and this thread helped me to check it via historic screening, and I was right.

I conclude that in the aggregation of realized returns from a sim, the script checks industry/sector classification of the ticker as of today. This is a bit confusing, especially for mergers, acquisitions or other tickers that don't trade anymore but also for businesses that completely changed over the years.

Please (if possible) consider aggregating the realized returns according to historic classification "at buy" instead.

Thanks!