The irresistible rise of financial data scientists

Big data has changed the landscape of capital markets, with an increasing need for financial analysts

Featured faculty: Cen Ling and Jo Chanik
Written by Putro Harnowo

“Data is the new oil,” according to British mathematician Clive Humby in 2006. Many have rushed to get their hands on data since then, and none are more eager to turn numbers into profits than financial institutions. Just like oil, crude data needs to be extracted and refined, which brings the dawn of new alchemists called data scientists.

A data scientist uses a blend of skills in statistics, computer science and domain knowledge to draw meaningful insights from abundant data available. The fast-growing demand for this profession outpaces the growth of talent supply, and financial institutions found themselves competing not only among themselves but also against tech companies to attract the best data savants.

“Financial institutions have significantly increased their recruitment of data scientists in the past two decades,” says Jo Chanik, Assistant Professor in the Department of Finance at the Chinese University of Hong Kong (CUHK) Business School. “Institutional investors strategically adjust portfolio allocation and recruitment decisions to maximise the benefits generated by their data scientists.”

Along with Associate Professor Cen Ling from the same department, Professor Jo observes a clear pattern of the surge of data scientists on Wall Street. The number rose from 11,799 in 2008 and slightly dipped after the subprime crisis, but since then, it has quadrupled to reach 57,050 data scientists in 2021, with the top three employers by numbers being big names like Morgan Stanley, Credit Suisse and Goldman Sachs. Some firms recruit more than others, leading to variations in the concentration of data scientists.

However, the study finds that if a small group of investors has significantly more data and analysis on certain stocks, powered by employing more data scientists, they can uncover additional information unknown to others. Having exclusive information will make these investors trade less aggressively and frequently, which can slow down the rate at which useful information is reflected in the stock price.

“Competition among data scientists speeds up the production and trade of private information, but the concentration of data scientists covering a stock reduces its price informativeness in the capital markets,” Professor Jo adds.

“Competition among data scientists speeds up the production and trade of private information, but the concentration of data scientists covering a stock reduces its price informativeness in the capital markets.”

Professor Jo Chanik

Are data scientists the modern Midas?

As explained in a paper titled, Data scientists on Wall Street, Professor Jo and Professor Cen, as well as Han Bing of the University of Toronto and Han Yanru of Stevens Institute of Technology, collected detailed resumes from the Revelio Lab, which gathers career history data from various sources, including LinkedIn and Indeed.

They then merged the dataset with data from Refinitiv and Thomson Reuters Global Ownership to link with the actual stock holdings. Data scientist roles were further identified with a US-based occupations classification system called the ONET code, resulting in 326,627 unique data scientists employed by 3,126 institutional investors from 2008 to 2021.

Data scientists were mainly classified into three groups: data collectors for those gathering and organising data, data analytics for those evaluating data for making business decisions, and data maintenance refers to storing and protecting data with proper hardware. Data analytics has the most substantial impact on trading profitability, implying that data analysis provides the most direct and valuable insights for investment decisions.

Institutional investors with more data scientists also achieve higher profits. Each additional data scientist hired leads to a 0.004 percentage point increase per quarter in “abnormal” profits, which refer to returns earned that exceed predictions by standard asset pricing models. One extra data scientist translates to a 13 per cent increase in the average trading profitability.

“Data scientists assist financial institutions in earning abnormal profits by collecting, maintaining, and analysing large datasets, particularly alternative data, to identify mispriced assets,” says Professor Jo.

“Their insights allow institutions to make profitable trading decisions, with data analysts having the strongest impact on generating investment returns.”

Unequal access creates unequal outcomes

As multiple institutional investors can hold the same stock, the researchers then focus on the concentration of data scientists. To illustrate, stock A is held by 10 investors, with one having ten data scientists and the rest having none. The same number of investors holding stock B, each owning one data scientist. While the total number of data scientists is the same, stock A has a higher data scientist concentration than stock B.

In stock A, the investor with more data scientists has an analytical edge as it has more resources focusing on one stock. In stock B, each investor has similar resources, making the race to extract and leverage information more intense, eventually diminishing the informational advantage.

When a handful of investors with higher data scientist concentration strategically tilt their asset allocation to a smaller set of stocks, they enjoy an “information monopoly,” where the generated insights are concentrated within a few. Such a privilege is not for sharing.

“It’s not the number but the concentration of data scientists that reduces price informativeness. When data scientists are concentrated among a few institutional investors, these investors gain an information monopoly and have less incentive to trade on their information quickly,” says Professor Jo. “These investors trade cautiously to protect their advantage, delaying the incorporation of information into stock prices and limiting its spread across the market.”

Investors who hire more data scientists are found to hold more concentrated portfolios. A one standard deviation increase in data scientist concentration leads to an 11 per cent decrease in price informativeness. “This concentration creates a situation where valuable insights remain with a small group of investors rather than being reflected in stock prices, leading to less efficient price formation.”

Race to build the best data army

While institutional investors compete to be one step ahead of others, hiring more and more data scientists has become the new battleground. Investors finding themselves lagging behind their peers will react by increasing their recruitment efforts, especially for data analytics roles.

Financial institutions that frequently trade stocks, cover a broader range of industries, and have larger assets are more likely to expand their data teams, as evidenced by hedge funds employing more data scientists than pension funds and banks. These investors are strategically monitoring and responding to their competitors’ hiring activities, rather than blindly following general trends. Data scientists who worked for competitors are also sought after, leading to a tense catch-up in the talent race.

The number of data personnel recruited positively correlates with the number of data-science related undergraduate programmes at local universities, suggesting that the supply in the labour market affects financial institutions’ employment.

“Data science roles in financial markets will likely continue growing rapidly, with increasing specialisation and strategic importance,” Professor Jo adds. “As the competition intensifies among financial institutions and against tech companies, we may see wider wage premiums for data talent, more sophisticated analytical techniques, and potentially regulatory responses addressing the efficiency costs of concentrated information advantages.”

Although the study used data on American institutional investors, he believes the findings would likely be relevant to emerging markets like China. “The competition for data talent and the value of information advantages are universal market dynamics. However, regulatory differences, varying levels of market efficiency, and potential differences in data availability might affect the magnitude of these effects in emerging markets.”