Portfolio Theory & Risk
Don't Put All Your Eggs in One Basket — How do I build a portfolio and price risk?
🗺️ The Big Picture — Five Ideas in Plain English
Before diving into formulas, here is what this chapter is really about — in plain language. Each concept builds on the last.
1. Risk = Volatility
How much your investment might go up OR down. Like weather: can you predict it? No. Can you prepare for it? Yes.
2. Correlation
Do two investments rise and fall together? If not, mixing them makes your portfolio steadier — for free.
3. Efficient Frontier
For any risk level, there's a 'best portfolio.' The frontier is the menu of all best choices.
4. CAPM
A formula that says: the riskier your investment (relative to the market), the more return you should demand.
5. Sharpe / Alpha
Tools to judge: did a fund manager actually do well, or did they just get lucky by taking more risk?
Topic & Why It Matters
In 1952, Harry Markowitz published a 14-page paper that won him the Nobel Prize 38 years later. His insight was deceptively simple: investors care not just about expected return, but about the variance of that return — and when assets are combined, their variances do not simply add up. They combine in a way governed by correlation. Two assets that move oppositely can together produce less risk than either one alone. This is the mathematical foundation of diversification, and it is the only genuine free lunch in finance.
Building on Markowitz, William Sharpe developed the Capital Asset Pricing Model (CAPM) in 1964 — another Nobel-winning idea. CAPM answered a question Markowitz left open: once you can diversify away company-specific risk, what risk remains, and how should it be priced? CAPM's answer: only systematic risk — the risk of the whole market — earns a return premium. Company-specific risk earns nothing, because rational investors diversify it away for free.
Together, these ideas explain why a concentrated bet on one company is not rewarded proportionally to its total risk, why a 60/40 stock-bond portfolio outperforms a pure stock portfolio on a risk-adjusted basis, and why the cost of equity in any DCF model must account for beta — not just historical returns.
Knowledge Points
1. Risk = Volatility (But Only Sometimes)
In finance, risk is measured as standard deviation (σ) of returns — how widely actual returns scatter around the average. Total risk = Systematic risk (market) + Idiosyncratic risk (company-specific). Diversification eliminates idiosyncratic risk, but you need 20–30 uncorrelated stocks to do it. After ~30 stocks, adding more barely reduces portfolio σ. The irreducible remainder is systematic risk — the market's risk that no diversification can escape. Howard Marks observed: 'Risk is the probability of permanent capital loss.' Both definitions matter: σ for pricing, Marks' for investing.
💡 Plain English — Everyday Example
🚗 Everyday analogy: your daily commute.
On average you arrive in 30 minutes — that's the 'expected return.' Some days it's 20 minutes, other days 50 minutes. That daily variation IS the risk. Standard deviation (σ) measures exactly that unpredictability.
Two very different risk types:
• Flat tire → Idiosyncratic risk. Affects only YOUR car. Solution: carry a spare, take an Uber, have a backup plan. This risk is avoidable.
• City-wide snowstorm → Systematic risk. Every driver is stuck. No spare tire helps. This risk cannot be escaped.
In stocks: owning 1 company = you only have one car. Owning 30 stocks across different industries = 30 different backup plans. A market crash = the snowstorm — unavoidable, but survivable if you're prepared.
📖 Key Terms Explained
Greek letter sigma. Measures how spread out returns are. A stock with average return 10% and σ=20% might realistically return anywhere from -10% to +30% in a given year. Higher σ = wilder swings = more unpredictable.
Also called 'Unsystematic Risk.' Risk affecting just ONE company — a CEO scandal, product recall, or factory fire. Can be nearly eliminated by owning many different stocks in different industries.
Also called 'Market Risk.' Risk that hits the ENTIRE market — recession, financial crisis, pandemic. Cannot be reduced by diversification. Every investor faces it.
2. Correlation and the Free Lunch
Portfolio variance = w₁²σ₁² + w₂²σ₂² + 2·w₁·w₂·σ₁·σ₂·ρ₁₂, where ρ₁₂ is the correlation between assets. When ρ = −1, assets perfectly offset each other and portfolio risk collapses to zero. When ρ = +1, there is no diversification benefit at all. Real-world stock correlations run around ρ ≈ 0.3–0.6, providing partial benefit. Adding bonds (historically ρ ≈ 0 with stocks) to an equity portfolio cuts risk without proportionally cutting return. This is Markowitz's famous 'free lunch' — the only one in finance.
💡 Plain English — Everyday Example
☂️ Imagine you run two small businesses:
• A sunscreen shop: Thrives in summer, slow in winter and rainy days
• An umbrella shop: Thrives in rainy season, quiet in sunny weather
These two businesses are negatively correlated — when one struggles, the other booms. By owning BOTH, your monthly income becomes far more stable, even though each shop alone has wild swings.
📊 Correlation (ρ) ranges from -1 to +1:
• ρ = +1.0 → Two oil companies: they always move together. Zero benefit from owning both.
• ρ = 0.0 → Gold and tech stocks: mostly unrelated. Good diversification benefit.
• ρ = -1.0 → Perfect opposites: theoretically, combined risk could reach zero!
Real stocks: ρ ≈ 0.3–0.6 (partial benefit). Stocks + bonds: ρ ≈ 0 (solid benefit). This stable income gained for FREE — without reducing average earnings — is the famous 'free lunch.'
📖 Key Terms Explained
Greek letter rho. A number from -1 to +1 measuring how two assets move together. ρ=1: always same direction. ρ=-1: always opposite. ρ=0: no relationship. The lower the correlation, the better the diversification benefit.
A measure of how 'wobbly' the combined portfolio is. The formula looks scary, but the key insight: when correlation (ρ) is low, the portfolio wobbles LESS than the individual assets would on their own.
In economics, there's almost never a free benefit — you always pay somehow. But diversification is the rare exception: mixing low-correlation assets reduces risk without reducing average return. Harry Markowitz called it the only genuine free lunch in finance.
3. The Efficient Frontier
For any level of risk (σ), there is exactly one portfolio offering the maximum expected return. The set of all such portfolios traces the efficient frontier — a curved boundary in risk-return space. Rational investors should only hold portfolios on this frontier, never below it. The minimum variance portfolio is the leftmost point: the lowest achievable risk for any given asset set. The Tangency Portfolio is where a straight line from the risk-free rate just touches the frontier — in theory this equals the market portfolio, and every investor holds some combination of it and the risk-free asset.
💡 Plain English — Everyday Example
🍽️ Think of a restaurant menu — price vs. meal quality:
• $10 → decent burger
• $20 → excellent steak ✅ efficient
• $20 → mediocre salad ❌ inefficient (same price, worse food!)
• $30 → amazing 5-course meal
The 'efficient' choices are the BEST meal at each price point. Paying $20 for a mediocre salad when an excellent steak costs the same is simply irrational.
The Efficient Frontier is the 'best value meal list' for investors:
✅ On the frontier → maximum return for your risk level. Smart choice.
❌ Below the frontier → worse return for the same risk. Never do this.
⭐ The Tangency Portfolio (marked ★ in the demo) is the single 'best value meal' — highest return-per-unit-of-risk. In theory, this is exactly the global stock market index. That's why most finance experts say: just buy an index fund.
📖 Key Terms Explained
A curve showing all optimal portfolios — each gives the HIGHEST possible return for its level of risk. Think of it as the 'A-team' of possible portfolios. Any portfolio not on this curve is suboptimal: you're getting less return than you could for the same risk.
The safest possible portfolio you can build with the available assets. The leftmost point on the frontier curve. Like the gentlest hiking trail — you still make it to the destination, but with the least effort (risk).
The portfolio with the BEST reward-to-risk ratio (highest Sharpe ratio). Where the 'value line' from the risk-free rate just barely touches the frontier. In theory, this equals the global stock market index — which is why passive index investing is so powerful.
A straight line from the risk-free rate through the Tangency Portfolio. Every point on this line is achievable by mixing the Tangency Portfolio with cash (Treasury bills). Any point on the CML beats all portfolios below it at the same risk level.
4. CAPM — Capital Asset Pricing Model
E(Rᵢ) = Rf + βᵢ × (E(Rm) − Rf). Rf is the risk-free rate (US 10-year Treasury). E(Rm) − Rf is the equity risk premium (ERP): the extra return the market must offer above risk-free to compensate investors for bearing market risk — historically 4–6%. Beta (β) measures a stock's sensitivity to the market: β=1 moves with the market; β=2 is twice as volatile. Crucially, only systematic risk (β) is rewarded. Idiosyncratic risk earns no extra return because it can be diversified away for free. CAPM is used to estimate the cost of equity in WACC calculations.
💡 Plain English — Everyday Example
💼 CAPM is like a salary negotiation formula for different job types:
Three job offers:
🏛️ Government desk job → guaranteed salary, zero risk = Rf (3%)
🏢 Average corporate job → better pay, some risk = Market (Rm = 8%)
🚀 Your startup job → unpredictable income, high risk = ???
Equity Risk Premium (ERP) = Corporate pay − Government pay = 8% − 3% = 5%.
This is the 'danger pay' for leaving the safe government job.
Beta (β) = how risky YOUR startup is vs. average corporate:
• β = 0.5 (half as risky) → You should earn: 3% + 0.5×5% = 5.5%
• β = 1.0 (same risk) → You should earn: 3% + 1.0×5% = 8%
• β = 2.0 (twice as risky) → You should earn: 3% + 2.0×5% = 13%
⚠️ Critical rule: You only get extra pay for SYSTEMATIC risk (the snowstorm). If your startup is risky because of internal problems (idiosyncratic), no extra pay — investors can diversify that away for free, so they won't reward you for it.
📖 Key Terms Explained
A formula: Expected Return = Safe Rate + Beta × Market Premium. It tells you what return a stock SHOULD earn based on its systematic risk. If a stock earns less than CAPM predicts → it is overpriced. If it earns more → potentially underpriced.
Greek letter beta. Measures how volatile a stock is COMPARED to the overall market. β=1: moves exactly with market. β=2: moves twice as much (great when market rises, terrible when it falls). β=0.5: half the market volatility. β<0: moves opposite to market (very rare).
The return on the 'safest' investment available — usually US Treasury bills or 10-year Treasury bonds. Currently around 3–5%. It's your baseline: any investment must beat this to be worth taking any additional risk.
The extra return stocks offer above risk-free bonds — like 'danger pay' for investing in stocks. Historically 4–6% per year. If bonds return 3% and stocks return 8%, ERP = 5%.
5. Sharpe Ratio and Alpha
Sharpe = (Return − Rf) / σ measures risk-adjusted return per unit of total risk. Treynor = (Return − Rf) / β measures per unit of systematic risk. Jensen's Alpha = actual return − CAPM expected return: positive alpha suggests skill (or luck). The Efficient Market Hypothesis holds that alpha should average zero — few managers consistently beat. Over long periods, the S&P 500 achieves Sharpe ≈ 0.5. Great fund managers sustain 0.8+. The hard problem: distinguishing 3 years of alpha from luck requires 10+ years of data to reach 95% statistical confidence.
💡 Plain English — Everyday Example
⛽ Sharpe Ratio = investment fuel efficiency (like km per liter):
Two cars:
• Car A: 400km on 10L → 40 km/L efficiency
• Car B: 600km on 20L → 30 km/L efficiency
Car B travels farther (higher return), but Car A is MORE EFFICIENT per liter of fuel (per unit of risk). Sharpe Ratio captures this:
📊 Example:
• Fund X: 10% return, σ=15%, Rf=3% → Sharpe = (10−3)/15 = 0.47
• Fund Y: 14% return, σ=25%, Rf=3% → Sharpe = (14−3)/25 = 0.44
Fund X wins on efficiency even with lower return! Fund Y takes so much extra risk that you're not fairly compensated. A smart investor could borrow money, lever up Fund X, and end up with MORE return than Fund Y at the SAME risk level.
🎰 Alpha — the million-dollar question:
After adjusting for HOW MUCH market risk the manager took (beta), did they actually add value beyond what luck would explain?
• Positive alpha = outperformance (but could be skill OR luck)
• Need ~10 years of consistent data to distinguish skill from luck at 95% confidence
• Only ~10% of active fund managers beat their index consistently over 15 years, net of fees
📖 Key Terms Explained
Return per unit of total risk. Formula: (Return − Safe Rate) / σ. Like miles-per-gallon for investments. S&P 500 long-run average ≈ 0.5. Excellent managers achieve 0.8+. Higher = better efficiency.
Like the Sharpe Ratio, but divides by Beta (systematic risk) instead of σ (total risk). Used when comparing managers within a large diversified fund, where only systematic risk matters.
The 'extra' return earned above what CAPM predicts given the risk taken. Positive α = outperformance. But 3 years of positive alpha could be pure luck. You statistically need ~10 years of data to confirm it reflects genuine skill.
The theory that market prices already reflect ALL public information, making it impossible to consistently 'beat the market.' If true, alpha averages zero. Strong empirical support — which is why index funds beat most active managers over long periods.
Formula Reference
Every formula below has a "Plain English" column. Read that first if the formula looks scary.
| Concept | Formula | Note | Plain English |
|---|---|---|---|
| Portfolio Return | E(Rp) = Σ wᵢ · E(Rᵢ) | Weighted average of asset expected returns | 60% in stocks (10% return) + 40% in bonds (4%) = 0.6×10% + 0.4×4% = 7.6% total |
| Portfolio Variance (2 assets) | σp² = w₁²σ₁² + w₂²σ₂² + 2w₁w₂σ₁σ₂ρ₁₂ | ρ₁₂ = correlation; range −1 to +1 | How wobbly the portfolio is. Key: when ρ is low, portfolio wobbles LESS than each part alone — that's the free lunch |
| Beta | βᵢ = Cov(Rᵢ, Rm) / Var(Rm) = ρᵢm · σᵢ / σm | β=1 moves with market; β=2 is twice as volatile | Market rises 10%, β=2 stock rises ~20%. Market falls 10%, it falls ~20%. β=0.5 = half that swing |
| CAPM | E(Rᵢ) = Rf + βᵢ × (E(Rm) − Rf) | ERP = E(Rm) − Rf ≈ 4–6% historically | Expected return = Safe rate + (Your risk level) × (Market danger pay). E.g., β=1.5, Rf=3%, ERP=5% → 3%+1.5×5% = 10.5% |
| Sharpe Ratio | S = (Rp − Rf) / σp | Risk-adjusted return per unit of total risk | Return per unit of risk — like km/liter. Portfolio 10%, Rf=3%, σ=15% → Sharpe = 0.47. S&P 500 ≈ 0.5 historically |
| Treynor Ratio | T = (Rp − Rf) / βp | Risk-adjusted return per unit of systematic risk | Like Sharpe, but uses Beta instead of σ. Better for comparing funds inside a diversified portfolio |
| Jensen's Alpha | α = Rp − [Rf + β(Rm − Rf)] | Excess return above CAPM prediction; positive = outperformance | Did the manager beat what CAPM predicts? α = (actual return) − (what CAPM says you should've earned for that beta) |
| Information Ratio | IR = α / σ(α) | Active return per unit of active (tracking) risk | How CONSISTENTLY does the manager produce alpha? High IR = reliably beating benchmark, not just a one-time lucky streak |
Interactive Demo — Two-Asset Portfolio Builder
Set expected returns, standard deviations, and correlation (ρ) for two assets. Adjust the portfolio weight slider to explore how risk and return shift along the efficient frontier. The tangency portfolio (★) and minimum variance portfolio (◆) are computed in real time. Change ρ toward −1 to see the diversification benefit grow; drag it to +1 to watch it vanish.
For beginners: Think of Asset A as "stocks" (high return, high risk) and Asset B as "bonds" (lower return, lower risk). The demo shows you what happens when you mix them in different proportions — and how correlation between them changes your portfolio's overall wobbliness.
How much Asset A is expected to earn per year on average
How volatile Asset A is — bigger = wilder price swings
How much Asset B is expected to earn per year on average
How volatile Asset B is
Do A and B move together? Drag toward -1 to see risk fall. Drag toward +1 and diversification disappears.
How much of your portfolio is in Asset A vs. Asset B
The return you'd get from a perfectly safe investment like a US Treasury bond. Used to compute Sharpe ratio.
Portfolio Return
10.0%
A 60% / B 40%
Portfolio Std Dev (σ)
13.0%
Total portfolio risk — lower is safer
Sharpe Ratio
0.54
Return per unit of risk — higher is better
Diversification Benefit
2.2%
Risk saved vs. holding assets separately
| Wt A | Wt B | Return | σ (Risk) | Sharpe |
|---|---|---|---|---|
| 0% | 100% | 7.0% | 8.0% | 0.50 |
| 5% | 95% | 7.3% | 7.9% | 0.54 |
| 10% ◆ | 90% | 7.5% | 7.8% | 0.57 |
| 15% | 85% | 7.8% | 8.0% | 0.60 |
| 20% | 80% | 8.0% | 8.2% | 0.61 |
| 25% ★ | 75% | 8.3% | 8.5% | 0.61 |
| 30% | 70% | 8.5% | 9.0% | 0.61 |
| 35% | 65% | 8.8% | 9.5% | 0.60 |
| 40% | 60% | 9.0% | 10.1% | 0.59 |
| 45% | 55% | 9.3% | 10.8% | 0.58 |
| 50% | 50% | 9.5% | 11.5% | 0.57 |
| 55% | 45% | 9.8% | 12.2% | 0.55 |
| 60% ← | 40% | 10.0% | 13.0% | 0.54 |
| 65% | 35% | 10.3% | 13.8% | 0.52 |
| 70% | 30% | 10.5% | 14.7% | 0.51 |
| 75% | 25% | 10.8% | 15.5% | 0.50 |
| 80% | 20% | 11.0% | 16.4% | 0.49 |
| 85% | 15% | 11.2% | 17.3% | 0.48 |
| 90% | 10% | 11.5% | 18.2% | 0.47 |
| 95% | 5% | 11.8% | 19.1% | 0.46 |
| 100% | 0% | 12.0% | 20.0% | 0.45 |
★ = Tangency Portfolio (max Sharpe) — the market portfolio in theory; the CML passes through it. ◆ = Minimum Variance Portfolio (leftmost point on the frontier). Efficient portion: from ◆ upward. Portfolios below ◆ are dominated — same or more risk, less return. Diversification benefit = (weighted-average σ) − (actual portfolio σ).
Step-by-Step Method — Building and Evaluating a Portfolio
- Define objectives and constraints. State your investment horizon, target return, risk tolerance, liquidity requirements, and tax situation. These determine which assets are eligible and constrain the optimization. A pension fund with 30-year liabilities should optimize differently than an individual with a 5-year horizon.
- Estimate expected returns for each asset. Use historical returns with caution — past returns can mean-revert. Supplement with forward-looking estimates: CAPM-implied returns, analyst consensus, or fundamental models. Be explicit about your equity risk premium assumption (ERP = 4–6% is common; Damodaran publishes monthly updates).
- Estimate standard deviations (σ) for each asset. Use at least 10 years of historical data to capture full market cycles including crises. Recognize that volatility itself is volatile — option-implied volatility (the VIX) provides a forward-looking alternative for equity portfolios.
- Estimate the pairwise correlation matrix. Capture correlations between all asset classes using long historical windows. Stress-test by replacing calm-period correlations with crisis-period correlations (2008, 2020) — correlations spike toward +1 in crises, collapsing your diversification exactly when you need it most.
- Compute the efficient frontier. For two assets, sweep portfolio weights from 0% to 100% in steps. For many assets, use mean-variance optimization (quadratic programming). Identify the minimum variance portfolio (ρmin-σ) as the leftmost feasible point.
- Identify the tangency portfolio. Among all efficient portfolios, find the one with the maximum Sharpe ratio given your risk-free rate. This is the optimal risky portfolio. In theory, it equals the market-cap-weighted global market portfolio (justifying passive index investing).
- Blend the tangency portfolio with the risk-free asset. Your risk tolerance determines the split. Conservative: 30% tangency + 70% risk-free. Aggressive: 100% tangency. Ultra-aggressive: lever up by borrowing at Rf to invest more than 100% in the tangency portfolio — the Capital Market Line extends past the tangency point.
- Rebalance and reassess periodically. Prices drift, correlations shift, and your own risk tolerance changes over time. Rebalance at least annually to restore target weights. Update expected returns and correlations annually. Re-run CAPM beta estimates using current data — beta is not stable across time periods or market regimes.
Real-World Case — Ray Dalio's All Weather Portfolio
Ray Dalio founded Bridgewater Associates, the world's largest hedge fund. His All Weather portfolio — designed to perform in any economic environment — is one of the most cited examples of Markowitz diversification applied at the asset-class level. The core insight: stocks and bonds are negatively correlated in most regimes (when growth disappoints, bonds rally). Adding gold and commodities hedges inflation scenarios that crush both stocks and bonds simultaneously.
| Asset Class | Allocation | Ann. Return | Ann. Volatility | Role in Portfolio |
|---|---|---|---|---|
| US Equities | 30% | ~10% | ~15% | Growth engine — high return, high risk |
| Long-Term Bonds (20–25 yr) | 40% | ~5% | ~12% | Deflation hedge; historically negative correlation with equities |
| Intermediate Bonds (7–10 yr) | 15% | ~4% | ~6% | Stability buffer; lower duration risk than long bonds |
| Gold | 7.5% | ~7% | ~15% | Inflation hedge; tends to rise when real interest rates fall |
| Commodities | 7.5% | ~4% | ~20% | Inflation hedge; diversifies against supply-side shocks |
| All Weather Portfolio | 100% | ~9.7% | ~3.9% | 1984–2013 backtest · Sharpe ≈ 1.5+ · Max drawdown ~11% |
Common Pitfalls
| Mistake | Corrective Rule |
|---|---|
| Using historical σ as forward-looking risk | Correlations are not stable. In the 2008 crisis, correlations between stocks, corporate bonds, and REITs all jumped toward +1 simultaneously — diversification failed exactly when investors needed it most. Build in margin of safety; never assume the historical correlation matrix holds in a tail-risk scenario. |
| Over-diversifying into correlated assets | A portfolio of 100 bank stocks has far less diversification than 20 stocks across different industries. True diversification requires low correlation between holdings, not just a high count. Concentration within a sector — even disguised as broad diversification — retains all of that sector's systematic risk. |
| Using beta from the wrong time period | 5-year monthly beta vs. 1-year daily beta often produce very different numbers for the same stock. Growth stocks had betas near 1 in 2018; their betas exceeded 1.5 by 2020. Use a consistent methodology and update regularly. In DCF models, many practitioners use the industry median beta, de-levered then re-levered for the target capital structure. |
| Ignoring the equity risk premium estimate | ERP of 4% vs. 6% in CAPM produces dramatically different costs of equity — and therefore different valuations. ERP is not a fixed constant: Damodaran estimates it monthly. In 2022, rising risk-free rates combined with elevated ERP compressed equity valuations even before earnings fell. Always be explicit about your ERP assumption. |
| Declaring alpha after 3 years of outperformance | Statistical significance requires roughly 10+ years of consistent outperformance to distinguish skill from luck at 95% confidence. Most managers who outperform for 3–5 years underperform their benchmark over 10–15 years. Fewer than 10% of active equity funds beat their index net of fees over 15 years (SPIVA Scorecard, 2023). |
Self-Check
- Stock A has β = 0.5, Rf = 3%, and equity risk premium (ERP) = 5%. What is the CAPM expected return for Stock A?
- You hold a portfolio of 50 stocks, all in the banking sector. Is this portfolio well-diversified? Explain why or why not.
- Portfolio X returns 10% with σ = 15%. Portfolio Y returns 14% with σ = 25%. Rf = 3%. Which portfolio has a better Sharpe ratio, and what does that mean for a rational investor?
References & Further Learning
Click any card to jump directly to the course or resource and continue learning.
Yale Financial Markets (Shiller) — Open Courses
Yale OCW · Free
Lectures 4–6 cover portfolio theory, CAPM, efficient markets, and behavioural finance. Shiller's teaching is exceptionally accessible.
Financial Markets (Yale / Shiller on Coursera)
Coursera · Audit Free
Updated Coursera version of the Yale course with quizzes, peer assignments, and a certificate option.
Damodaran: Risk & Return Sessions
NYU Stern · Free
Sessions on beta estimation, equity risk premium, CAPM, and multi-factor models with current-year market data and Excel templates.
MIT OpenCourseWare: Finance Theory I
MIT OCW · Free
Portfolio theory and CAPM modules. Problem sets include mean-variance optimization and real asset-allocation exercises.