Some questions about this change in data providers …
There is a lot of intellectual capital in our existing work, hope this change is done in a way that will preserve it. Thanks.
The only way to truly preserve exactly what has been done up till now would be to stay with the existing S&P relationship. Any time a database changes, we have to expect changes to output lists, backtested results, etc. We saw this back in 2012 when we transitioned from Reuters to S&P and we expect to see it again when we transition from S&P to FactSet.
Here’s the problem: The only solution that would completely preserve all that’s been done is off the table, not by our choosing but by S&P’s choosing.
To put it bluntly, S&P does not want you to be generating this sort of intellectual property unless you pay them at least $24,000 per year and they have been relentless in working to obstruct portfolio123 from serving you, so much so that they have already served notice that our data relationship will terminate in mid 2020 and that if p123 is to have any data, it will have to be on new S&P-dictated terms that most of you will find unacceptable. So if you want complete continuity, your only option is to contact S&P and negotiate to purchase a data license, assuming (as we think and hope) they will still be willing to serve their data via p123 to license holders.
For those who are not going to be paying S&P for licenses, the question is the value of the intellectual property you have already created
versus the value of the intellectual property you will have in the new FactSet era.
It is my opinion that potential changes in the value of your intellectual property will be insignificant; in some case, it might be marginally better with S&P, in other cases, it might be marginally better with FactSet, but I do not expect a meaningful change one way or the other in the value of any intellectual property that was worth anything in the first place.
If you find that a model that was good under S&P collapses under FactSet, that is a red flag that you need to change your approach and that you would have needed to do so even if we stayed with S&P. We are not a nano-second 24-7 trading platform that seeks to profit from bits of noise traveling at warp speeds. We are a fundamental/technical platform that serves “investors” and given the nature of the datasets with which we deal, any model that so so tightly optimized for a particular data standardization protocol that it collapses under a conceptually similar but different protocol would necessarily be a severely non-robust model that is nothing more than a ticking time bomb. In fact, if you do license S&P data via p123, it would probably be a really good idea for you to also own an entitlement to use FactSet on p123 in order to do an extra check on the validity of your model. That would be the ultimate best quant practice!
So I believe you will be fine with FactSet; actually, better than fine.
Thank you Marc.
Can you expand on this, please?
Appreciated.
-Jim
BTW, am I correctly assuming that we will have a period with an overlap of data sources i.e. S&P and Factset in Q2 2020?
It will be very interesting to compare, contrast and learn…
Jerome
Hi Marc,
Thanks for the thoughtful answer. Very much appreciated.
Clarifying one of my prior questions - Can we assume that all of the existing data points will be available with FactSet?
FWIW, I am very much a long-term, value-oriented investor. That’s where the historical data has taken me, so to speak. Things like EV/Operating Income, EV/Tangible Book Value, CapEx (i.e., it’s efficiency), etc., are central to what I do. Not a comprehensive list, but a sampling of my favored data. Also, I’ve found that certain industries are more favorable to investors than others. I would hope that FactSet data will allow us to continue to find good investment nuggets using these search criteria, and that an approach that relies on long-term (i.e. 1-year) holding periods is unlikely to see a material change in backtests/results.
Thanks,
Ed
Under S&P, you’d have five years of fundamentals and ten years of estimates. With FactSet you’ll have twenty years of both.
Yes, that is absolutely our plan.
There may be a few minor differences, but it seems to me that almost all of the present data points will be available with FactSet, plus, perhaps, a few additional ones.
Correction to the SP backtest proposal: It would be 10 years of fundamentals and five years for estimates; for any model that uses both, you’d have to assume a 5-year test period.
As to the transferability of ratios and factors, I assume most or all would be reproduced; the FactSet fundamental database is, actually, a copy of the Reuters database that Reuters has to license to them as part of an antitrust settlment needed to allow the 2008 Thomson and Reuters merger to go through; that’s relevant to p123 because we started with Reuters and then, in 2012, we busted our you-know-whats to get everything translated to Compustat. The main differences between Compustat and FactSet will be in the as-ofs and timing, and in standardization algorithms.
The latter are remarkably imprecise in the world of finance. One example I can recall off the top of my head is how cost of sales is computed fur purposes of getting a gross margin. Some add income-statement to cost of sales, others do not. Who is right? Everyone says they’re right and the other guy is wrong and each has an equally persuasive argument. Lots and lots of things like that. This, by the way, is what is meant by “standardization.” No two companies format their financial statements with the exact same sets of line items, so none of the work would be do-able unless we could standardize the data; that’s the hear of the data provider’s intellectual property – it had better be because they have no rights to the raw numbers which, by law, are in the public domain.
So again, there will be differences. The real issue is whether FactSet is a credible provider. In this regard, the company (ticker: FDS) has market cap of $9.6 bl. and TTM rev of $1.435 bil. ROE is 53.38% (5Y) and 58.89% (TTM). ROA is 27.38% and 23.77%. S&P (SPGI) also has good numbers. The key is that with FactSet, whatever quibble one may have with it (or S&P), its not like we’ll be getting data from Slick Sam whose office is on the left bank under the bridge. FactSet is a real company and like S&P, its data is used by many, many, many serious professionals, just like with S&P, and just like Reuters now known as Refinitiv and soon to be known as that company the London Stock Exchange is buying. There are others out there, but these are the three major players.
Marco and I had occasion, a few years ago when S&P started acting a bit nuts, to get sample FactSet data and do a deep dive. S&P calmed down a bit so we didn’t see a pressing need to continue on with FactSet, but we had seen enough back then to recognize that FactSet would be a sensible alternative if we needed it. Now, we need it.
Congrats for the bold move. I was expecting/hoping it sooner, but I guess you have already been working on it for a while. Having the best data is no use if it makes us hostages of a provider.
A bit is an understatement. I don’t understand their strategy… So far they only seem to be pushing business away with their antics. Congrats on the move, Factset is the right choice.
One thing that will need to be added for global stocks is the ability to filter stocks based on exchange. Our brokerage accounts will not have access to every single exchange, especially in Europe. Ie. there is no point in designing a model that includes stocks we aren’t able to buy.
Very welcome news.
One question comes to mind - Will you have time to work with the FactSet data pre-launch? In other words, are you aiming for a transition from S&P to FactSet with as little disruption to our models as possible?
I think that’s most likely going to be the case. When P123 moved from Reuters to S&P Global in 2013 (I think?), they conciliated most line items before releasing it. A couple were deprecated but it wasn’t a big deal. I would expect this to be the same, but with less deprecation as most line items of S&P Global got an equivalent with Factset. However, I would expect the rank of each company to change about 10-15% between both data providers. This will most likely trigger a big rebalancing whenever your models are due to rebalance but that’s normal. If you are running a fund tracking a strategy using P123, you most likely have to warn your clients about it.
P123,
When you say we will have access to foreign markets via subscription, does that include Canada? Currently Canadian data is included in our subscription. So would Canadian data be grandfathered if it becomes a paid subscription? Also will this include Canadian ETF’s?
I would also like to request that a list of factor changes be made at the appropriate time and be sticky’d in the forum for a while. I would imagine this won’t be much more work considering you have already said you’ll be doing a data comparison before the switch. While doing comparisons does that mean you’ll be editing any factor descriptions if there are changes noted? I’m not sure how you’ll keep the two data sets separate during the overlap period if there are factor changes but I’m sure you have a plan.
It sounds like factset will actually give us PIT updates faster than S&P so I’m looking forward to testing the effects of that.
Nothing will change regarding Canada. It will still be included in our basic North American subscription.
The plan is to have a beta site with FactSet data at the same time as our normal site with S&P data for a while before the switch. Thus users can compare both and get used to using the new data. I think it’s a good idea that if any changes need to be made in how factors are calculated, we’ll devote a special space on the forum for that, as well as putting a reference document in our resource library.
These 2 points are important for many users, thanks. Even without investing in it, the Canadian market is a must for out-of-sample tests.
I will be interested in seeing what this looks like plotted in Factset versus what we have been using:
sma(5,0,#spepscny)>sma(21,0,#spepscny)
wow!
Great News! Looking Forward for international stocks, it is going to be a blast.
In Terms of data Quality (PIT P123 like Quality like in the past, etc.) I trust you guys from p123 100%.
Go for it!!!
And thank you a ton!!!
Best Regards
Andreas
Will the FactSet data include information on foreign-domiciled large-cap OTC stocks? For example, there is no data currently for Tencent, among others. It would be helpful to have access to data for such stocks. Thanks.