The diversification and welfare effects of robo-advising
GAICIL Insights
The quiet revolution in household finance may be a dull one: fewer stocks, more index funds, less time spent worrying.1 The diversification and welfare effects of robo-advising, by Alberto G. Rossi and Stephen Utkus, studies previously self-directed investors who adopt a large US hybrid robo-advisor. Five facts emerge, and together they offer one of the clearest pictures of how algorithmic advice changes household finance.
First, robo-advice reshapes portfolios by increasing indexing and reducing home bias, the number of assets held and fees. Second, these changes contribute to higher Sharpe ratios. Third, the biggest beneficiaries are investors who lacked high exposure to equities or indexing and had poorer diversification. Fourth, robo-advice cuts the time investors dedicate to managing investments. Fifth, investors who benefit most are more likely to join the service and less likely to quit.
The details read like a checklist of classic investor mistakes, corrected. Across all investors, robo-advising reduces money market mutual fund holdings and increases bond holdings. It reduces idiosyncratic risk by lowering holdings of individual stocks and US active mutual funds and raising exposure to low-cost indexed funds. It eliminates home bias by significantly increasing international equity and fixed income diversification. Over the relatively short sample, it raises overall risk-adjusted performance.
A machine learning twist explains who gains. Using Boosted Regression Trees, the authors explain cross-sectional variation in the effects of advice. Beneficiaries have little investment experience, high prior cash holdings and high trading volume before adopting advice. Algorithmic advice corrects common mistakes, and those making them gain most. An algorithm studies the algorithm's impact, a fitting method for the era.
This is artificial intelligence democratizing expertise. Diversified, low-cost portfolios were once the domain of investors who could afford professional advice or knew how to build them. A robo-advisor delivers similar discipline at scale. Higher risk-adjusted returns, lower fees and saved time are tangible welfare gains. What used to be a private-banking privilege becomes a default setting.
Self-selection is the market working as it should. Investors who benefit most join and stay. People who need help find it and keep using it when it helps. That supports a market-based approach to robo-advice, with competition among providers improving quality and cost. A product whose most loyal users are those it helps most passes a demanding test.
The hybrid model deserves a closer look. The service combines algorithmic portfolio management with human advice, pairing the consistency of algorithms with the reassurance of human contact. As AI capabilities grow, the balance between algorithmic and human elements will evolve, and studies like this one can track the consequences. The human may matter most on the days markets frighten people.
Which points to the open question. The sample period is relatively short, and long-run effects across market cycles remain to be studied. How investors respond to algorithmic advice during downturns matters especially. Evidence that robo-advice helps investors stay the course when markets fall would strengthen the case for wider adoption.
Time savings count in their own right. Investors spend less effort managing portfolios after adopting advice, freeing attention for other uses. Financial performance measures alone understate the benefits of tools that reduce the burden of everyday decisions. Hours not spent staring at tickers are part of the return.
For policy, the research supports frameworks that enable algorithmic advice while protecting investors. Rules designed for human advisors may fit robo-advisors poorly. Regulators can focus on transparency, fee disclosure and suitability while allowing innovation that lowers costs and improves diversification.
Read as a whole, AI can help households make better financial decisions. By correcting under-diversification, home bias and high fees, robo-advice improved outcomes for those who needed it most. As it spreads, it could narrow the gap between sophisticated and inexperienced investors, a meaningful contribution to financial well-being and to the efficient allocation of savings.
A closing challenge: the most consequential financial AI of the decade may never pick a winning stock. Its achievement could be stopping millions of people from trying.
Introduction
We study the diversification and welfare effects of a large US robo-advisor on the portfolios of previously self-directed investors and document five facts. First, robo-advice reshapes portfolios by increasing indexing and reducing home bias, number of assets held, and fees. Second, these portfolio changes contribute to higher Sharpe ratios. Third, those who benefit most from robo-advice are investors who did not have high exposure to equities or indexing and had poorer diversification levels. Fourth, robo-advice decreases the time investors dedicate to managing their investments. Fifth, those investors who benefit most are more likely to join the service and not quit it.
Takeaway
We study the effects of a large U.S. hybrid robo-advisor on the portfolios of previously selfdirected investors. Across all investors, robo-advising reduces idiosyncratic risk by lowering the holdings of individual stocks and active mutual funds and raising exposure to low-cost indexed mutual funds. It further eliminates investors’ home bias and increases investors’ overall risk-adjusted performance, mainly by lowering investors’ portfolio risk. We use a machine learning algorithm, known as Boosted Regression Trees (BRT), to explain the cross-sectional variation in the effects of advice on portfolio allocations and performance. Finally, we study the determinants of investors’ sign-up and attrition.1
The paper has benefited from the comments made at presentations at the NBER Big Data and Securities Markets meeting, NBER Summer Institute on Aging, AFA Annual Meeting, EFA Annual Meeting, New York Fed FinTech Conference, GFLEC Seminar, Virtual Finance Seminar Series in Corporate Finance and Investments, INFORMS Annual Meeting, ABFER 7th Annual Conference, Luohan Academy, Georgetown Global Virtual Seminar on FinTech, Workshop on FinTech and Machine Learning (Singapore), Vanguard Research Symposium, Cornell University, Georgetown University Vanderbilt University, University of Notre Dame, Boston University, Lehigh University, Baruch College, Michigan State University, Temple University, Queen Mary University, University of Technology Sydney, Vienna Graduate School of Finance, American University. We are grateful to Reena Aggarwal, Linda Allen, John Ameriks, Turan Bali, Gurdip Bakshi, Brad Barber, Agostino Capponi, James Choi, Youngmin Choi, Francesco D’Acunto, Xi Dong, Paul Gao, Xiaohui Gao, Christopher Hansman (the EFA discussant), Kathleen Hanley, Steven Kou, Craig Lewis, Claire Liang, Vladimir Mukharlyamov, Nandu Nayar, Lalitha Naveen, Andreas Neuhierl, Maureen O’Hara, Cindy Pagliaro, Lin Peng, Alessandro Previtero, Tarun Ramadorai (the NBER discussant), Michael Reher (the AFA discussant), Jonathan Reuter, Samuel Rosen, Jesus Salas, Gustavo Schwenkler, Berk Sensoy, Yao Shen, Vicky Tang, Sterling Yan, Bart Zhou Yueshen (the ABFER discussant), YinYin Yu, Kate Waldock, Yajun Wang, Steve Zeldes, Haibei Zhao, Dexin Zhou, Jonathan Zinman, for comments and suggestions. Stephen Utkus was employed at Vanguard in a research capacity and retired in August 2020.
Rossi is an unpaid consultant at Vanguard in order to access the anonymized data. ∗ McDonough School of Business, Georgetown University, Washington, DC, USA.
Robo-advisors have surged in popularity in recent years as investors seek low-cost, automated investment advice. They allow investors to set up customized, diversified portfolios and can give access to other wealth management services previously limited to affluent investors, such as portfolio tax efficiency, cash flow forecasting, and retirement income planning. Many such services emphasize investment in low-cost index funds, minimal trading, tax efficiency and global diversification—arguably the benchmark for optimal portfolio advice from the empirical finance literature. In addition to being comparably inexpensive, robo-advisors have the potential to be superior to human financial advisors, as the latter have been shown to display behavioral biases and cognitive limitations (see Linnainmaa, Melzer, and Previtero, forthcoming). As a result, robo-advisors are quickly attracting attention from policymakers and investors at all levels. This paper provides a comprehensive analysis of the largest robo-advisor in the US—Personal Advisor Service (PAS) from Vanguard.
Notes
- Alberto G. Rossi & Stephen Utkus, The diversification and welfare effects of robo-advising, 157 J. Fin. Econ. 103869 (2024), https://doi.org/10.1016/j.jfineco.2024.103869. Takeaway source: Conclusions, reproduced from a free author or repository copy (working-paper version). ↩
Cite as: Alberto G. Rossi & Stephen Utkus, The diversification and welfare effects of robo-advising, 157 J. Fin. Econ. 103869 (2024), https://doi.org/10.1016/j.jfineco.2024.103869
Header image: Campus of Iowa State University 01.jpg by Dirk DBQ, CC BY 2.0, via Wikimedia Commons.
