You will find below the results of my investigation on the quality of data between S&P Global and Factset for the US Large Caps Equity Universe.
Universe: PRussell 1000
Note: I could have chosen the S&P 500 but I wanted to include some of the “smaller” large caps in the US for the purpose of this exercise.
I decided to look at the main financial statement line items to compare both database. I made a one node rank with each line item, cut the universe into deciles (100 stocks) and looked the first date where all the data is available for all stocks. In other words, when the ranking is able to compute the line item for each decile without returning “0”. You will find below the results.
Although Factset clearly lost the battle against S&P Global for the Canadian Equities database, the results are surprisingly mixed for US Large Cap Equities.
Both data providers are fairly close when it comes to balance sheet and earnings estimates. S&P Global is better for income statement inputs:
S&P GLOBAL - EBITDA (WINNER)
FACTSET - EBITDA
However for some reason Factset is better for cash flow statement inputs:
S&P GLOBAL - FCFPSTTM
FACTSET - FCFPSTTM (WINNER)
Conclusion:
Factset is a viable alternative to S&P Global for US Large Cap Equities. If you are an investor looking to backtest only these kind of stocks, then feel confident that the impact between both providers, when it comes to data availability, is going to be minimal.
We’re aiming for October. FactSet tries to cover all the world’s stock exchanges. We’ll be offering the major ones at least. We haven’t yet finalized a schedule or pricing.
There are a lot of hurdles. We haven’t started looking into them, but just off the top of my head, you have lots of companies that report only annually, and then there’s all the currency conversion issues, which are complicated by currency exchange rates that change radically over time.
FactSet only offered me one subscription that includes everything, they never differentiated between US-only and Global data.
The only issue I’ve seen like that in ex-US developed markets is that most Japanese companies report semi-annual cash flow statements.
If converting into the USD is necessary (its probably not in most cases since almost every variable used is currency free, ie. ratios, growth rates, etc), its just an index for each datapoint to the corresponding exchange rate at the right date…
Yuval, I don’t understand this point. If I decide to test models in Euro markets, I want all data, sim returns and benchmarks in the original currency, not converted to USD. Investors or funds have a portfolio in Euro either because it is their base currency, or because they made a strategical choice of Euro market exposure. Taking into account exchange rates is not only complicated: it adds a macro randomness that may dwarf the randomness of the model itself. Keep it simple.
I would absolutely want EUR data for EUR companies, GBP data for GBP companies, etc.
People can separately choose to take (or not) the FX risk via their own FX hedging.
I’d very much like to run rankings in universes with more than one currency (for example Norway, Sweden and Denmark). In cases like this it’s going to be hard to avoid dealing with currency issues.
If a ranking system is based on ratios, you don’t need currency rates to run it and get a meaningful result.
From the moment currency rates are used, they bring an additional risk in the evaluation (past) and in the execution (future) of a model.
Having thought about it some more, you’ll need currency conversion in certain situations… For example, if you want to bias towards large caps, comparing the market cap of a Japanese company (measured in Yen) to a US company (in USD) without currency conversion is useless.
My ideal setup would be to have a single model in a single universe for (a subset of) Northern Europe. This will enable me to find the “best” companies in the overall universe, without having to compare companies in different models by hand. You make a good point in that most common ranking “dimensions” will not be affected, but a few will: mktcap as you mention, momentum, volatility are examples. I’m not sure what the best solution is.
No longer being a p123 insider, I haven’t had much occasion to rant here but I think it’s time to throw one for old time’s sake.
I can’t tell you what the to-do list is for the FactSet conversion or global, bur I was involved with going from Reuters to SP and adding Canadian, and based on that, I can tell you there are countless details you’re missing and would never think about unless Marco and team plowed ahead like you’re wishing they would at which time you’d howling and screaming at what would be a bug-plagued mess.
Do yourselves a favor and stop pestering Marco and Yuval over nonsense. The team has plenty to do to get the FactSet conversion done and will have a sh**load more to do to go global. Don’t bog them down reading gibberish and thinking and/or discussing internally whether to respond or debate. Sometimes in life, you just have to trust that the insiders are busting their butts to do the best that can be done for you and that it’s being done as expeditiously as possible.
And by the way, the initial SP vs FactSet data quality study was a useless effort that completely missed the point. The proper comparison is FactSet versus the shallow shell of a platforms those of you who can’r pony up for a professional license would have been left with, (The status quo of staying with SP as is was not on the table as a choice.)
With all the brain power in these forums I think that asking questions and thinking about problems out loud probably helps more than it hurts, wouldn’t you say?
I’d imagine that some users have plenty of database experience in multiple programming languages and could maybe help…
The exercise wasn’t a futile effort. What I did proved me I can’t rely on Factset for Canadian Equities before 2012, and many of my models are Canadian related.
I did this to decide if I should pay 4 times more for S&P Global or keep the same price for Factset. There are only 2 weeks left to take the decision so it is an urgent matter.
In the end I will “pony up” for a professional S&P Global license to have an extra year to let all the bugs get sorted out on Factset, win some peace of mind temporarily.
12K is really expensive and it’s gonna hurt to pay that much out of my pocket as an “individual” but I feel it is the right choice based on my research, at least for me.
