If I collect market data from some third party (such as brokers or some service API) and redistribute the data with others (could be for profits or non-profit), do I need to get licensing from exchanges? I am thinking about a project to build something like wikipedia for market data where people can contribute and maintain together but not sure whether there are legal concerns like licensing. I remember databento has posted some discussion about this on and it seems the conclusion is that in the US, only CME requires licensing but I couldn't find the post.
www.quandl.com was bought by Nasdaq. They have some free stocks and futures wiki databases as well as the paid.
Actually — usually, no. 1. The data belongs to the trading venue(s). It doesn't belong to the third party or broker(s). They themselves have to license the data for redistribution from the trading venue(s). 2. Most trading venues take the position that their data becomes free to redistribute once the data becomes historical. At this moment, two major exceptions to this conventional practice are CME and ICE. CME has a fee for historical distribution by API, whereas ICE do not but will often want their licensees to fill out a form. 3. With that said, we've seen some third parties add on a stipulation in their user agreement or master service agreement that you must not redistribute the data that you've received from that particular third party. If your third party vendor claims exchange data received from them cannot be redistributed, there's a good chance it's only codified in your service agreement with them, and not an exchange policy. (In our case we don't have such a restriction, but require attribution.) 4. We're not offering qualified legal advice here, but as a matter of observation, we don't know if this type of clause in (3) has ever been tested in civil court. The data doesn't belong to the third party, so it's less certain how it will damage the third party. As a practical matter, it's more likely that they have recourse to terminate the relationship if they don't like you redistributing their data. 5. I feel the main reason very few bother to re-redistribute data is also a practical one rather than a legal one: Is [re-][re-]redistributed data really useful? Usually the more layers of redistribution that the data goes through, the worse its quality and the shadier its provenance. I thought it's interesting to comment on this topic because we had to make a similar commercial decision: Should we allow our users to redistribute bootleg copies of data from us and potentially compete with our business? We decided in our case that we're absolutely fine with users redistributing our data because the data itself belongs to the trading venue and is NOT our value-add, but rather it's our API/frontend, infrastructure, and data provenance that's providing the value-add. So yes, if you happen to acquire historical data through Databento, by all means go create a wikipedia for market data! We'd love to support it.
Thanks for Tessa's explanation. This is great news for me. Would you mind sharing an example of where we can find a venue's policy on redistributing historical data? I did find an academic paper, "Understanding the market for US equity market data" by Charles M. Jones that makes the same claim as Tessa (on page 44): "Data more than 15 minutes old is considered “historical data” and subscribers, including third-party vendors, can use that data as they wish, including redistributing or reselling the data without any payments to the CTA and UTP Plans." However, he does not cite a source for this claim. I have tried to look into this issue by browsing exchange websites such as NYSE's, but have found nothing relevant. Could someone please share some thoughts on this? Thank you.
The easiest way to know is to talk to their sales reps. We just happen to know by recent experience because we take all of the prop feeds. But you can search for terms like "data policy" or "data license agreement" and find each venue's market data licensing policy. For example, NYSE's is here: https://www.nyse.com/publicdocs/nys..._Market_Data_Comprehensive_Policy_Package.pdf Usually it's free by omission, meaning they define what is "delayed", which is usually between 15 minutes and end-of-day, and then "real-time" is defined as anything earlier than "delayed". Certain fees apply to "real-time" and "delayed", and hence everything that is omitted (which we call "historical data") is free for redistribution. The 3 major exchange groups and CTA/UTP SIPs tend to harmonize their definitions, with minor differences, making these practices commonplace. The exceptions to this are venues like IEX, who essentially treat anything over 15 milliseconds delayed as "historical", and allow free redistribution of such.