Mean-Variance Optimization, Black-Litterman, and the Total Portfolio Approach

In 1952, Harry Markowitz published "Portfolio Selection," showing that imperfectly correlated assets create diversification opportunities and that well-chosen weights reach the most return available for each level of risk. That insight later earned a Nobel Prize and became the basis of mean-variance optimization (MVO).

Where allocation fits. Institutional investors have three levers for influencing portfolio returns: (1) asset allocation, (2) market timing, and (3) fund or security selection. For many institutions, asset allocation is the most important of the three. Strategic asset allocation (SAA) is the investor's long-run target mix. It comes from the investor's goals and from long-horizon expectations of return and risk. For a passive indexer, it is the only major decision. Tactical asset allocation (TAA) changes the portfolio's systematic risk exposures through time to earn superior risk-adjusted returns.

Why MVO exists. Markowitz showed that a portfolio's risk depends on how its holdings move together. Holding assets that are less than perfectly correlated therefore creates diversification opportunities, and MVO selects the weights that reach the best attainable risk-return combinations.

Common mistakes

• Minimizing variance with no regard for risk aversion. The curriculum's MVO maximizes $\mu - (\lambda/2)\sigma^2$, and the optimal weights scale with $1/\lambda$. When $\lambda$ rises, risky exposures shrink toward zero: positive weights fall and negative weights become less negative. Trap: treating MVO as pure variance minimization with no $\lambda$ term.
• Assuming the highest excess return earns the highest weight. The inverse covariance matrix also sets the weights. In Example 1, fixed income has the lowest excess return (1 percent) yet the largest weight (42.8 percent) because of its low volatility and negative equity correlation. Trap: ranking closed-form weights by expected return alone.
• Forgetting that a negative-beta asset can have a negative hurdle rate. The hurdle is $E[R_{New}] - R_f > (E[R_p] - R_f)\beta_{New}$. With $\beta = -0.3$, an optimal portfolio at 16 percent and a 1 percent riskless rate yields a hurdle of $-3.5$ percent, so an asset can lose money and still improve the portfolio.

Bottom line

• MVO maximizes mean return minus a variance penalty scaled by $\lambda$; concave utility means risk aversion, and $\lambda$ is reliably backed out from an unconstrained optimal portfolio with a riskless asset, not guaranteed for constrained or arbitrary portfolios.
• Unconstrained closed-form weights scale risky exposures toward zero as $\lambda$ rises; liabilities add a hedging term, and a new asset helps when it clears its beta-based hurdle.
• Unconstrained MVO is an error maximizer; objective error-risk signals, resampling, shrinkage, and weight constraints temper estimation error, while liquidity penalties and factor caps are separate adjustments that express illiquidity preferences and exposure limits.
• Black-Litterman reverse-optimizes equilibrium returns, blends confidence-weighted views, and optimizes on posterior returns and covariance; its closed-form weights hold only without binding constraints.

Exam shortcut

Formula review, in one place: $\lambda = (E[R_p]-R_f)/\sigma_p^2$, guaranteed for an unconstrained mean-variance optimum with a riskless asset. $w = (1/\lambda)(E[R-R_0]/\sigma^2) + L(\delta/\sigma^2)$ for one risky asset, with $\lambda$ on a return basis and $L$ as a liability-to-asset ratio; drop the second term when there are no liabilities. Hurdle: $E[R_{New}] - R_f > (E[R_p]-R_f)\beta_{New}$. Liquidity penalty: subtract $\phi L_p$; factor cap: $b_P \le \bar{b}$. $\Pi = \delta\Sigma w_{eq}$ gives BL's prior; $QW_t = (\bar{\mu} - \Pi)/(Q - \Pi)$ measures view influence.

The full lesson (about 13,352 words, 89 min read) adds 2 worked examples, all 8 common mistakes, a self-check, free in the app.

Learning objectives

• mvo process
• mvo implementation
• mvo multiple risky
• mvo issues
• mvo adjustments
• mvo estimation error
• black-litterman model
• bl model private markets
• constraining bl model
• tpa overview
• tpa defining
• tpa governance
• tpa factor lens
• tpa competition capital
• tpa culture
• transitioning tpa
• tpa implementing
• maintaining tpa
• tpa gp-lp relationship
• core satellite
• top down bottom up
• risk budgeting
• factor risk budgeting
• risk parity
• other quant strategies
• taa
• taa process
• cash commitments illiquidity
• overcommitment model
• takahashi-alexander model
• historical simulation cash flow
• monte carlo-based cash flow
• managing unfunded commitments
• liquid alternatives
• lp direct investment
• co investments
• co investment returns
• secondary market pe
• gp led secondaries
• funds funds
• fof historical returns
• rebal buy hold constant mix
• rebal directional
• rebal cppi
• rebal obpi
• rebal dynamic illiquid
• rebal costs

Browse all free CAIA Level II lessons or jump into free CAIA Level II practice questions.