Portfolio managers often face the challenge of building long-term investment strategies using return data observed over much shorter horizons. This creates a “horizon mismatch” between a portfolio’s design and how it is ultimately held. While traditional mean–variance optimization is often dismissed as unrealistic for long horizons due to non-normal returns, we show that the horizon mismatch actually rescues the mean–variance framework. In fact, efficient mean–variance portfolios based on short-horizon data, with a buy-and-hold strategy for the long run, are actually optimal in the multi-period case, and they are located on the multi-period mean–variance efficient frontier. Thus, by employing the mean–variance rule, investors can still achieve near-optimal outcomes even over long investment horizons. This article offers a practitioner-focused perspective on why mean–variance optimization remains highly relevant—and how it can be safely applied across different investment horizons, even in the face of returns with skewed distributions.
Haim Levy (Tue,) studied this question.