The Efficient Frontier: Best Return for Every Risk

You already know the engine room: combine assets that don’t move in lockstep and a portfolio’s volatility drops faster than its return does, bending the risk–return curve inward. Lovely. But that leaves an awkward question — which mix do you actually hold? There are infinitely many ways to split money across a handful of assets, and each one is a different bet. We need a way to throw out the bad ones wholesale and keep only the keepers.

That’s the efficient frontier: the short list of portfolios where you are not leaving anything on the table. For every level of risk you’re willing to stomach, exactly one mix delivers the highest expected return — and that’s the only one worth owning at that risk. Everything else is, in a precise sense, a mistake. This lesson draws that frontier, names its parts, and shows you how to spot a portfolio that’s quietly wasting your money.

The feasible region: every dish you could cook

Picture a graph with risk (volatility, $\sigma$) on the horizontal axis and expected return ($E[r]$) on the vertical. Drop in a few risky assets — call them stocks, bonds, gold, whatever — and start mixing. Put 70% in one and 30% in another: that blend has some expected return and some volatility, so it’s a single point on the graph. Shift to 50/50: a different point. Add a third asset and vary all the weights: more points. Allow every possible weighting of every asset, and the points stop being scattered dots — they smear together into a solid two-dimensional area.

That filled-in area is the feasible region (also called the opportunity set). It’s the set of every risk–return combination you can actually build from your assets. Because diversification bends the curve, the region has a characteristic bullet-like shape, blunt nose pointing left — which is why it’s nicknamed the Markowitz bullet.

The pantry analogy: your assets are ingredients, and a portfolio is a recipe — some amount of each. The feasible region is the menu of every dish you could possibly cook from that pantry. Some recipes are delicious (great return for the effort/risk), some are a sad gruel (lots of risk, meagre return). They all exist; the menu just lists what’s cookable, not what’s worth eating.

Dominance: when one portfolio quietly beats another

Now we sort the menu. Portfolio X dominates portfolio Y if X is at least as good on both axes and strictly better on at least one — that is, X has $\ge$ the expected return and $\le$ the risk of Y, with at least one of those a strict inequality. If X dominates Y, no rational investor holds Y: X gives you the same-or-more return for the same-or-less risk. Y is dominated — a polite word for “strictly worse, full stop.”

Here’s the punchline: any portfolio strictly inside the feasible region is dominated. If you’re in the interior, you can always slide up (same risk, more return — there’s a better mix at that volatility) or slide left (same return, less risk). Free improvement, both directions. Only points on the upper-left rim have nowhere better to go.

Worked example — three real candidates:

Compare P1 and P2: same 8% return, but P2 carries 10% risk versus P1’s 14%. P2 gives you the same return for less risk, so P2 dominates P1. Now P1 and P3: same 14% risk, but P3 returns 11% versus P1’s 8%. P3 gives more return for the same risk, so P3 dominates P1 too. Poor P1 is dominated on both fronts — nobody should hold it. P2 and P3, meanwhile, don’t dominate each other: P3 has more return but also more risk, so choosing between them is a genuine taste call. Hold that thought — it’s the whole point of the frontier.

The efficient frontier: the upper-left rim

Strip out every dominated portfolio and what’s left is a thin curve hugging the upper-left boundary of the feasible region. That curve is the efficient frontier. Crisp definition: for each level of risk, the efficient frontier is the portfolio with the maximum expected return achievable at that risk. (Equivalently: for each target return, it’s the mix with the minimum risk.) Same idea read two ways.

Drag the slider below. The strong-colored upper arc is the efficient frontier; the muted lower arc is its evil twin — same volatilities, but feeble returns. The dot at the far-left tip is special, and it’s next.

Notice the shape. The frontier is a curve, not a line — it bows up and to the left, steep near the tip and flattening as you go right. That flattening is diminishing returns made visual: near the top, you have to swallow a lot more volatility to squeeze out a little more expected return. The curve is concave because diversification’s benefit fades as you push toward the highest-return (least-diversified) corner.

The minimum-variance portfolio: the leftmost tip

Follow the bullet to its blunt nose — the single leftmost point, the one with the smallest possible volatility of any mix you can build. That’s the minimum-variance portfolio (MVP). No combination of your assets is calmer; it’s the floor on risk.

The MVP is the hinge of the whole curve. Everything on the frontier above the MVP is efficient. Everything below it — the lower arc of the bullet — is inefficient, because for any point on that lower arc there’s a point directly above it at the exact same risk with a higher return. Same volatility, more return: dominated.

Worked check using the same risk levels. Suppose the MVP sits at 9% return and 12% risk (the tip). Take a lower-arc portfolio at 12% risk but only 5% return. It shares the MVP’s risk yet earns 4 percentage points less — strictly dominated by the MVP itself. Now take an efficient-arc portfolio at, say, 18% risk and 13% return; nothing else at 18% risk beats it, so it survives. The MVP is the dividing line: at and above it, keepers; below it, junk that happens to share a risk number with something better.

Choosing your point: there is no single ‘best’

Here’s the liberating part: portfolio theory does not hand you one magic best portfolio. It hands you a menu of non-stupid options — the efficient frontier — and then your risk appetite picks the point. Two equally rational investors can land on different frontier points and both be right.

• Conservative? Sit near the minimum-variance tip. Low volatility, modest return, sleep at night.
• Aggressive? Climb up and to the right. More expected return, but you’re signing up for the bigger swings that come with it.
• Somewhere in between? Pick a point in between. The frontier is continuous; every risk level has its efficient answer.

The job, then, isn’t “compute the best portfolio.” It’s: (1) build the efficient frontier from your assets, then (2) choose the spot on it that matches how much risk you can actually live with. Step 1 is math; step 2 is about you.

How the frontier shifts

The frontier isn’t fixed; change your ingredients or your rules and it moves. Three levers worth knowing:

• Add more assets, especially low-correlation ones. New ingredients that don’t move with what you already hold push the whole frontier up and to the left — better return for every risk level. This is diversification’s free lunch showing up as a literal upgrade of the menu.
• Allow short-selling. Let weights go negative (borrow an asset to bet against it) and you can build more extreme mixes, extending the frontier further out at both ends.
• Add constraints. Ban short-selling, cap how much can go in any one asset, or require minimum holdings, and you pull the frontier inward — you’ve removed options, so the best-achievable edge can only get worse or stay the same. Real-world rules almost always cost you a little frontier.

The takeaway: a frontier is only as good as the asset list and rules behind it. More uncorrelated assets and fewer constraints = a better menu.

Limitations: garbage in, garbage out

Modern Portfolio Theory is elegant, and elegance can lull you. Be honest about where it strains:

• Estimation error, amplified. The frontier needs three inputs per asset: expected returns, volatilities, and correlations. All three are estimated from noisy historical data, and the optimizer’s job is to chase the most attractive numbers — so it systematically over-weights assets whose returns were overestimated by chance. Tiny input errors become large weighting errors. Practitioners call this “error maximization”: the optimizer maximizes return and the mistakes in your inputs.
• One period, frozen. The classic model assumes a single investment period with fixed inputs. Real markets evolve — correlations spike exactly when you need diversification most (everything crashes together), and the frontier you drew last year may not be this year’s.
• Variance as the only risk. MPT measures risk purely as volatility (variance). But volatility treats a big upside surprise as just as “risky” as a big loss, and it ignores the shape of bad outcomes — it has no special fear of catastrophe.
• Returns assumed roughly normal. The math leans on returns following a tidy bell curve. Real returns have fat tails: extreme moves happen far more often than a normal distribution predicts, so the model understates the odds of a disaster.

None of this makes MPT useless — it’s the foundation everything else builds on. It just means the efficient frontier is a map drawn from estimates, not a guarantee. Treat its precise numbers with suspicion and its core logic with respect.

Key Takeaways

Next up: add a risk-free asset and the curved frontier becomes a straight line — the Capital Market Line, the upgrade that finally gives everyone a shared best risky portfolio to aim for.