Week 41 Day 3: The Efficient Frontier: Finding Your Optimal Mix

Imagine a graph: x-axis is risk (standard deviation), y-axis is return. You plot every possible combination of stocks and bonds. The upper-left boundary (highest return for each level of risk) is the efficient frontier. Any portfolio NOT on this line is suboptimal -- you could get the same return with less risk or more return with the same risk.

Building the efficient frontier with VTI and BND: 100/0 VTI/BND: return approximately 10%, risk approximately 16%. 80/20 VTI/BND: return approximately 9%, risk approximately 12.7%. 60/40 VTI/BND: return approximately 8%, risk approximately 10%. 40/60 VTI/BND: return approximately 7%, risk approximately 7.5%. 20/80 VTI/BND: return approximately 6%, risk approximately 5.5%. 0/100 VTI/BND: return approximately 5%, risk approximately 5%. Notice: the 80/20 portfolio gives up only 1% of return while reducing risk by 3.3 percentage points. That is a very efficient trade. Adding more assets improves the frontier further: adding SCHD, VXUS, VTIP, and SCHH creates more diversification opportunities, pushing the frontier further up and to the left (more return per unit of risk). Key takeaway: the 'best' portfolio is not the one with the highest return (100% stocks). It is the one on the efficient frontier that matches YOUR risk tolerance. If you can handle 16% volatility, go 100% stocks. If you can handle 10% volatility, go 60/40. If you can handle 7% volatility, go 40/60. All are on the frontier. All are optimal FOR THEIR RISK LEVEL. What is NOT on the frontier: (a) holding individual stocks (same expected return as the market, but higher risk due to concentration = below the frontier). (b) Paying 1% advisory fees (reduces return without reducing risk = moves you below the frontier). (c) Market timing (reduces return on average while not reducing risk = below the frontier).

The efficient frontier is the central concept of Modern Portfolio Theory (Markowitz, 1952). Formally, it is the set of portfolios that solve: max E[R_p] subject to sigma_p <= sigma_target, for all sigma_target in [sigma_min, sigma_max]. Equivalently, it is the solutions to: min sigma_p subject to E[R_p] >= R_target. The frontier is computed by solving a quadratic optimization problem: min w'*Sigma*w subject to w'*mu >= R_target, w'*1 = 1, w_i >= 0 (long-only constraint), where w is the vector of portfolio weights, Sigma is the covariance matrix, and mu is the vector of expected returns. The computational challenge: the inputs (expected returns and the covariance matrix) are estimated with significant error, and the optimization is highly sensitive to these inputs. Michaud (1989) showed that mean-variance optimized portfolios are 'error-maximizing' -- they concentrate in assets with the highest estimation errors. This is why naive 1/N portfolios often outperform mathematically optimized portfolios out-of-sample (DeMiguel et al., 2009). For individual investors, the practical implication is powerful: a simple two-fund or three-fund portfolio (VTI + BND, or VTI + VXUS + BND) is on or very near the efficient frontier, and adding complexity (optimization, factor tilts, alternative assets) provides minimal improvement -- and potentially negative value due to estimation error, higher costs, and behavioral complexity. The actionable takeaway: choose your risk level (by selecting a stock/bond ratio), implement it with broad index funds, and you are approximately on the efficient frontier.