Keystone Strategy

Keystone Applies Data Shapley Model in Prediction of Credit Defaults

During their presentation at NABE TEC2021 on November 8, 2021, Economists from Keystone presented how to apply a "Data Shapley" to value individual data points in the context of a critically important prediction problem: who is most likely to default on a loan?
November 8, 2021   /   2 Minute Read
Data Shapley Model

SEATTLE, WA – Keystone, an innovative economics, technology, and strategy consulting firm, presented their latest research on data valuations in credit default predictions leveraging a Data Shapley application at the fifth annual National Association of Business Economists – Tech Economics Conference (NABE TEC2021). The Keystone team of authors included Junsu Choi (an Engagement Manager at the firm), and Jill Furzer, PhD, and Vitoria Rabello De Castro, both economists at Keystone.

The conference is one of the marquee events for economists across sectors, bringing together hundreds of economists, data scientists, and business leaders and this year, the event’s theme centered around “Economics in the Age of Algorithms, Experiments, and A.I.” Attendees joined the virtual TEC2021 conference to discuss and learn from experts how artificial intelligence (AI), behavioral economics, and other tools are impacting businesses. Keystone Expert and the Economics of Technology Professor at Stanford, Susan Athey, spoke on the topic of economists in the tech sector and joined other leading economists and academics in the space, including Pat Bajari (VP Core AI and Chief Economist, Amazon), David Card (Nobel Laureate; Class of 1950 Professor of Economics, UC Berkeley), Hal Varian (Chief Economist, Google). Other session topics included the future of remote work, market design and behavior, as well as the evolving landscape of competition policy.

Keystone’s own economists Jill Furzer and Vitoria Rabello De Castro presented their research assessing the value of data for a bank’s decision of who to lend to. Their application of a Data Shapley method shows how this data valuation technique can inform firm data acquisition and discarding decisions. The empirical results also reveal a tradeoff between short run costs associated with lending to individuals who are more likely to default and long-term savings from algorithmic learning from underrepresented data points.

“This research brings Keystone’s economic expertise and technical knowledge about machine learning methodologies to solve of challenging problems affecting businesses,” notes Vitoria Rabello De Castro.


If you’d like to learn more about the Economics team and how economics plays a role in the work we do at Keystone, please reach out to Fraser Thompson at [email protected].

About Keystone

Keystone Strategy is a leading innovative strategy and economics consulting firm dedicated to delivering transformative ideas and cutting-edge solutions to Fortune Global 500 companies, top law firms, and government agencies. Our unique expertise in technology-led strategy, AI-driven digital transformation, product development, healthcare, finance, antitrust, and litigation enables us to create bold strategies that have far-reaching implications on business, consumers, and public policy. We leverage a unique combination of strategic insights from the world’s leading industry and academic experts with the practical expertise of our accomplished professionals to deliver extraordinary impact for our clients. Learn more about Keystone at or


Arranged around the theme “Economics in the Age of Algorithms, Experiments, and A.I.,” TEC2021 brought together leading thinkers in economics and data science to discuss how interdisciplinary developments in these fields are impacting today’s business. For five years, NABE-TEC has served as a venue for world-class business economists elaborate on the emerging trends and future data science needs in economic research and practice.

Find out why top tech firms, Fortune 500 companies, and global law firms partner with Keystone.