ICE

Moving fast +

breaking things

What makes data science different at ICE

Varun Pawar,
ICE Senior Director, Fixed Income Innovation Initiatives

Published
September 2021

Intensifying competition and shrinking margins have pushed firms to invest aggressively in technology and data over the past decade. This has led to an explosion in data consumption as firms change their business models to be more data centric and data driven to efficiently scale their operation.

Historically, clients consumed most of their data on an end-of-day basis due to technology constraints. As firms invested in their technology stack, this evolved into an intraday basis. With the advent of the cloud, clients are consuming vast amounts of data in near real time. With elastic storage and compute, firms are no longer bound by onerous in-house tech infrastructure costs and are free to focus on analyzing data that can add the most alpha.

The sheer volume of data has also outpaced traditional data analysis skills - “drinking out of a firehose” is a common refrain by clients. Basic coding expertise and the ability to work with large datasets are now part of the desired skillset for those considering a career in finance. Fixed income, currency, and commodity (FICC) trading headcounts have been decreasing as banks actively invest in automation. Risk, compliance, and technology roles have proliferated, amplifying demand for data expertise. At ICE, data scientists can help with analytics, backtesting, index composition, risk management and credit allocation based on vast amounts of historical data.

team members writing on a board

Members of ICE’s data science team in our New York offices

As an exchange operator and data provider, ICE uses data and technology to make markets more efficient - across fixed income, energy, equities, and commodities. ICE’s data scientists work in small nimble groups alongside production technology teams. The ability to iterate quickly in parallel environments allows us to draft, design, prototype, build and break new concepts to help bring the precision required by our most demanding customers. It’s what we call “move fast and break things”, looking for flaws and pressure testing new concepts without slowing down.

For example, during the March 2020 market dislocation, ICE’s data scientists worked with the bond evaluators to develop machine learning algorithms capable of calibrating bid/ask spreads despite increased price uncertainty and decreasing bids/offers from market makers. The team designed new algorithms and backtested results in a few months which statistically showed meaningful improvements toward precision.

Apart from enhancing traditional financial datasets, the team is actively involved with prototyping solutions based on alternative datasets, to help bring transparency to markets. One example involves bringing workforce demographic data to the municipal market. As municipal bonds are primarily serviced by tax revenue, having timely and in-depth data describing the tax base for a given location is key. We partnered with a leading payroll provider to obtain data that not only helps explain migration, but also describes the socio-economic profile of people migrating. This data can help answers two critical questions, “how many people moved” and “what type of people moved”. Putting it bluntly, if a large number of young, educated, high-income earning individuals migrate out of a city/county or state it is probably not a good sign for the risk profile of municipal bonds issued by those jurisdictions.

More recently we collaborated with colleagues at ICE Mortgage Technology to provide the ICE Data business with mortgage loan origination data. By pairing anonymized mortgage application data with mortgage-backed securities (MBS) we’re working to provide transparency into the future pipeline of mortgages that could be securitized. Clients benefit from having additional transparency and can utilize this data to gauge supply and demand while allocating capital. Additionally, incorporating timely anonymized origination data can give us the ability to enhance proprietary prepayment models that improve prepayment accuracy, thereby improving precision of bond prices.

Across multiple industries, data science initiatives continue to gather momentum. A recent survey of 300 asset managers showed 98% of firms surveyed are already using or planning to adopt data science tools in their investment strategy in the next one to two years, according to Northern Trust Asset Management. Still, there are indications that current systems and tools in many firms lack the sophistication to fulfil the potential of what data science can offer. Partnering with a firm like ICE can provide new insight, with access to expertise, extensive data sets, and the latest technology. And as the discipline of data science evolves, we’re confident in an approach that allows us to “move fast and break things” so we can meet client demand across dynamic global markets.