Abstract: We investigate the role of financial advisors in shaping clients' asset allocation during retirement. Using data on more than 37,000 advised Canadian retirees, we document that advised retirees maintain high equity shares of 60-70 percent well into old age. This share of risky assets is roughly twice the level prescribed by common rules of thumb, target-date funds, or life-cycle models calibrated to non-advised portfolios. Conflicts of interest are unlikely to explain this risk-taking, as retired advisors hold similarly high equity shares in their own portfolios. We show that the observed portfolios are consistent with a standard life-cycle model featuring moderate risk aversion and modest financial wealth levels. Under a "money doctor" interpretation, advisor-induced beliefs about market returns can rationalize advisor fees of up to 140 basis points per year.
Taha Choukhmane, Massachusetts Institute of Technology
Tim de Silva, Stanford University
Weidong Lin, Massachusetts Institute of Technology
Matthew Akuzawa, Massachusetts Institute of Technology
Abstract: We develop and implement a novel method to study personal financial advice from Large Language Models (LLMs). Studying this advice is challenging because it depends on the model used (i.e., supply), the questions individuals ask (i.e., demand), and their evolving circumstances. We address these challenges by surveying a representative sample of adults and asking them to write prompts seeking spending and investing advice from an LLM. We then simulate the lifetime paths that result from following this advice under realistic asset and labor market conditions. Applying our method to GPT-5.2 and Gemini 3.0 Flash, we document three facts about AI-generated financial advice. First, following LLM advice would move most survey respondents closer to the prescriptions of life cycle theory relative to their current behavior, including broader participation in diversified equity funds, equity shares that decline with age, and sizeable saving buffers. Second, replacing individual-written prompts with academic prompts moves LLM advice even closer to life cycle theory, with better consumption smoothing and less reliance on simple heuristics. Third, LLM advice varies systematically with individual characteristics, such as gender and financial literacy. These differences accumulate over the life cycle into wealth differences at retirement of 4-5% between groups and reflect both demand (i.e., systematic variation in the prompts written by different individuals) and supply (i.e., differences in advice for a given prompt). These facts highlight the potential of generative AI to improve financial decision-making, but suggest that its impact is likely heterogeneous across households and depends on how the technology is used.
Abstract: We study how investors update their beliefs and how these updates shape their decisions. Using a randomized information experiment with 2,800 investors across seven regions, we exploit the variation in the horizon of historical index returns shown to each investor to examine how the scope of information affects expectations. Investors revise their beliefs in the direction of the returns they see, with larger adjustments among those who are initially more uncertain. The sensitivity to information rises with the length of the return history and is highest when ten-year average returns are disclosed. These patterns are inconsistent with full-information rational expectations or simple extrapolation. Instead, they support a fading-memory learning model in which investors use longer histories to infer the long-run average return. This mechanism provides a structural explanation for the extrapolative expectations commonly observed in financial markets.
Discussant: Markus Ibert, Copenhagen Business School
