Robo-advisers do not use a prescribed norm to evaluate investors’ risk preferences and budget. Their methodology and output can be simplistic and inefficient. Nonetheless, because of the low threshold necessary to test and use digital advisory services, traditional wealth managers should not be complacent.
How Is This Research Useful to Practitioners?
Robo-advice varies from platform to platform. A tension exists between time efficiency and an appropriately in-depth analysis of investor risk tolerance and allocation. On average, robo-advisers tend to ask fewer questions in their risk assessments than their human counterparts. At times, some of the questions may not be germane to risk assessment. The authors explore these issues against a backdrop of scarce information through their rigorous analysis of risk assessment techniques of robo-advisers in three developed economies—Germany, the United Kingdom, and the United States—over a nine-week period in 2016.
Robo-advice, as a burgeoning subset of the advisory business, suffers from a number of flaws. The assessment process lacks uniformity. Questions vary among the robo-advisers under review, and some advisers fail to adequately capture the depth of investor risk tolerance or aversion. Additionally, the algorithms can produce somewhat simplistic results. The procedures need to be more efficient and should incorporate big data, artificial intelligence, and social media into the psychometric process. The user experience should also be more entertaining. Excessive rigidity in the risk assessment process leads to suboptimal outcomes, as does the lack of a multidimensional consideration of individual investors’ risk perceptions.
Traditional client-facing financial advisers may take comfort in the knowledge that risk prescriptions from robo-advisory platforms are inconsistent. Yet because these tools are so easy to use, advisers who deliver traditional services need to be vigilant. Relationship and sales managers, as well as compliance officers, will find the authors’ conclusions both concerning and timely.
How Did the Authors Conduct This Research?
The authors begin by surveying the literature on robo-advice, which to date has received little scientific attention. The quality of robo-advice will determine whether it becomes disruptive to the investment advisory landscape. Opinions and findings diverge on a meaningful uniform application of psychometric standards to the risk assessment process. Indeed, specialists endeavor to produce appropriate questionnaires that consider the complexity of risk appetite—a category that may provide insufficient information on its own.
This study is an exercise in the analysis of sparse information. The authors use data from seven German, three US, and three UK robo-advisers from 10 April 2016 to 26 June 2016. The US market is a point of reference because of its depth relative to the other two, and the UK market is relevant because of its position as the seat of the European banking industry. For a robo-adviser to be included in the sample, no login credentials can be required to access it, including the questionnaire and resulting assessment, and the robo-adviser must be a market leader in terms of assets under management and/or media coverage.
The authors determine which information from the questionnaires is processed for the robo-advisor to arrive at a risk profile and the degree to which it correlates with recommended portfolios. Each evaluation requires information about the investor to produce a bespoke portfolio. The authors’ analysis considers the detail level of the input process and answer combinations that result from the questionnaires. Most evince inadequate reliability, scoring below the necessary Cronbach alpha coefficient, a psychometric evaluation tool, thus confirming that robo-advisors need to economize end users’ time.
Because robo-advisers use different questions to determine risk tolerance, the authors attempt to create a process that makes the questions comparable. They group questions with similar intentions into 35 subcategories that, in turn, condense to the categories of general information, risk tolerance, and risk capacity. The advice varies in detail. Some involves product recommendations, whereas other guidance allocates by sector or allocation to equities and fixed income. Interestingly, some robo-advisers pose questions that are not used for risk assessment and are consequently irrelevant to the advice offered—a result that can be disconcerting to investors, who might thus forgo the advice. To be cost effective, robo-advice must offer standardized, rather than individually chosen, portfolios—the number of which allows the robo-advisor to match the investor’s risk profile with an adequate portfolio.
Opinions remain divided on the efficacy and sustainability of robo-advice. The process is subjective and wants for a uniform psychometric standard of comparison. The authors’ consideration of advice across three distinct investment cultures presents a challenge, especially given the survey’s short period and limited data, but the results show commonality nonetheless. The algorithms tend to draw more perfunctory, rather than in-depth, conclusions and err on the side of conservatism. Robo-advice needs to assess risk with less rigidity and greater depth than it currently does, as well as be more engaging to the end user—all in a time-efficient and user-friendly manner. Although it is potentially cost effective, robo-advice appears to have no advantage over human nuance and analytic acumen. Yet advisers should not be overconfident: Robo-advice is a burgeoning space in the advisory profession, and these are early days.