Robo-Advising Day

The Center for Financial Markets and Policy (CFMP) has been hosting a Global Virtual Seminar Series on Fintech every Friday since the beginning of the coronavirus pandemic. On Friday, September 25 CFMP, working in conjunction with the Ripple University Blockchain Research Initiative, was delighted to bring together scholars to present their robo-advising-related research for Robo-Advising Day.

Robo-Advising Day recording.

Alberto Rossi, associate director of the Center for Financial Markets and Policy (CFMP) at Georgetown University’s McDonough School of Business, presented, Who Benefits from Robo-advising? Evidence from Machine Learning, with his co-author, Stephen Utkus, principal (Vanguard) on Robo-Advising Day in September.

The co-authors analyzed Vanguard’s Personal Advisor Services (PAS), the largest hybrid robo-advisor in the world with $115 billion in assets under management. A machine learning algorithm known as “boosted regression trees” was used to explain the variation in the effects of advice on portfolio allocations and performance. They sampled 350,000 investors that interacted with PAS and organized them by their demographic characteristics (i.e. age, gender, and number of years with Vanguard). 

In their paper Rossi and Vanguard found that investors who benefited the most from the investment advice are “those individuals that had high cash holdings that were trading a lot and had little experience are going to be benefiting a lot and instead the ones that had a high share in mutual funds and high indexation.” 

Investors with little cash holdings and little investment experience benefited the least from the advice. When employing PAS, the participants saw an increase in their portfolio by percentage of bonds, Indexed Funds, and international funds and they saw a decrease in individual funds, ETFs, percentage of equities, and in advising fees. Rossi and Vanguard believe that robo-advising has the potential to exponentially benefit personal investors and has the potential to disrupt the entire financial advisory industry. 

Michael Reher, assistant professor of finance, University of California San Diego, and Stanislav Sokolinski, assistant professor of Rutgers University, co-wrote Robo-Advising Day’s second paper, Does Fintech Democratize Investing? Their paper studied the reduction in the minimum account size requirement implemented by a major U.S. robo-advisor. The paper showed that such reduction increases participation in both asset management and the stock market by households from the middle quintiles of the U.S. wealth distribution. 

Reher and Sokolinski found that robo-advisors have almost zero effect on lower or upper classes. They tested their theory through Wealthfront, which is considered a “pure” robo-advisor, compared to Vanguard’s PAS. In their analysis, they examined the change in wealth before versus after implementing Wealthfront and the users’ demographic information ( i.e. liquid wealth, age, and income). They found that democratization is partial within the middle class but incomplete, and thus no change, in the lower economic classes, which suggests that automated asset management has ambiguous effects on overall inequality in returns to wealth.

Agostino Capponi, associate professor of industrial engineering and operations research, Columbia University, gave the final presentation of the day on, Personalized Robo-Advising: Enhancing Investment through Client Interactions, with co-authors, Sveinn Olafsson, associate research scientist (Columbia University) and Thaleia Zariphopoulou, professor (The University of Texas at Austin, Oxford-Man Institute, University of Oxford). 


Their paper incorporates important features of modern robo-advising systems. Automated investment managers, or robo-advisors, have emerged as an alternative to traditional financial advisors. The value of assets under management by robo-advisors is the highest in the United States, exceeding $600 billion in 2019. The framework provided in the paper is in the form of a human-machine interaction system, where the robo-advisor (i.e., the machine) has the task of optimally investing the client’s (i.e., the human’s) wealth. The client has a risk profile that varies with time and to which the robo-advisor’s investment performance criterion dynamically adapts. The framework derives and analyzes optimal investment strategies, with the client’s risk profile changing in response to market returns, economic conditions, and idiosyncratic events. 

Their study showed an optimal interaction frequency, which maximizes portfolio personalization by striking a balance between receiving information from the client in a timely manner, and mitigating the effect of the client’s behavioral biases.

Capponi, Olafsson, and Zariphopoulou’s key takeaway is that by simply rebalancing the portfolio to maintain constant weights throughout the business cycle, the portfolio’s Sharpe ratio is near optimal and the portfolio return distribution is significantly improved.

Presented By:

Alberto Rossi

Associate Professor of Finance

Associate Director, Center for Financial Markets and Policy

Georgetown University McDonough School of Business

Michael Reher

Assistant Professor of Finance

UC San Diego Rady School of Management

Agostino Capponi

Associate Professor of Industrial Engineering and Operations Research

Columbia University
Ripple University Blockchain Research Initiative logo