A Japanese government bond pays 0.5%. A Brazilian corporate bond pays 12%. Both are written promises of money tomorrow. The market prices the difference, and the price has a name. This module is about that name — and about diversification, the only free lunch the market actually serves.
A 5-year Japanese government bond, in early 2026, pays roughly 0.5% per year. A 5-year Brazilian government bond, at the same moment, pays around 12%. Both are written promises by sovereign governments to return your money on a specified date. One offers more than twenty times the other. Why does the market price these so differently?
The answer is a single word: risk. The Japanese bond is almost certain to be repaid, in yen of relatively stable value, by a government with one of the world's deepest and most liquid bond markets. The Brazilian bond carries multiple layers of uncertainty — will the Brazilian government repay on schedule? Will inflation eat the real value of the repayment by then? Will the real depreciate against major currencies between now and the maturity date? Each uncertainty demands compensation.
The compensation has a name: the risk premium. It is the central organizing concept of modern finance. Every interest rate, every dividend yield, every expected return on every asset can be decomposed into:
If we strip out all the premiums, every asset in the world should pay the same risk-free rate (adjusted for currency and inflation). The fact that they don't tells you how the market prices uncertainty.
This module is about identifying those somethings — what they are, how they're measured, how diversification reduces them, and how their price ties back to the formulas you already know. By the end, you'll see that every present-value calculation in Module 03 had a hidden assumption: that the discount rate matched the risk of the cash flow. Module 06 will build directly on that link.
In everyday speech, "risk" usually means "something bad might happen." In classical finance theory, the meaning is more specific and slightly counterintuitive: risk is variability of outcomes. An investment that might pay 4% or 12% with equal probability is riskier than one that always pays 8%, even though both have the same expected return — and even though both possible outcomes of the first are positive.
This definition has consequences. A coin flip that pays $0 or $200 is risky. A guaranteed payment of $100 is not. Both have an expected value of $100. The first is risky because outcomes vary; the second is not because they don't. In finance, the upside variability and the downside variability both count, even though we mostly care about the downside emotionally.
Different kinds of risk have different sources. Six are worth knowing by name:
The risk that affects the whole market — recessions, geopolitical shocks, central-bank policy shifts. Cannot be diversified away by holding more securities of the same kind.
The risk attached to a single asset or company — a CEO scandal, a failed product, a single-firm fraud. Can be reduced sharply by holding many securities.
The risk that a borrower won't repay. Higher for corporate bonds than government, higher for emerging-market governments than developed-market ones.
The risk that you can't sell quickly without accepting a steep discount. Real estate has it. Some emerging-market bonds have it. Money-market funds usually don't.
The risk that exchange-rate moves erode your foreign-asset return when converted home. A US investor's Japanese stock return is a yen-stock return plus a yen-dollar move.
The risk that purchasing power erodes faster than your nominal return — covered fully in Module 04. Hits cash and long-duration nominal bonds hardest.
This list isn't exhaustive — there's also interest-rate risk, regulatory risk, political risk, and more — but these six cover most of what shows up in personal-finance decisions. Notice that they aren't independent: a credit crisis in emerging markets can trigger currency moves, liquidity drying up, and sharp falls in market prices all at once. Risks correlate. That correlation is exactly why diversification has limits, as Section 05 will show.
If risk is variability, we need a way to put a number on variability. There are three measures worth knowing.
The most common measure. It captures how spread out an asset's returns are around their average. A higher standard deviation means a wider range of possible outcomes. By convention, financial volatility is usually expressed as an annualized figure.
Rough benchmarks for what counts as "high" or "low":
| Asset class | Typical volatility | Interpretation |
|---|---|---|
| Cash / money-market | ~ 0–1% | Effectively no variation |
| Short-term government bonds | ~ 3–5% | Mild variation, mainly from interest-rate moves |
| Long-term government bonds | ~ 7–10% | Moderate, very sensitive to rate shifts |
| Investment-grade corporate bonds | ~ 5–8% | Adds modest credit-spread variation |
| Developed-market equity index | ~ 15–18% | The standard "stocks are risky" benchmark |
| Emerging-market equity index | ~ 20–25% | Adds currency and political risk |
| Single individual stocks | ~ 25–60+% | Specific risk dwarfs market risk |
| Cryptocurrency | ~ 60–100% | An order of magnitude above equities |
A useful rule of thumb: if returns are roughly bell-shaped, then in any given year, an asset will land within one standard deviation of its average about two-thirds of the time, and within two standard deviations about 95% of the time. So a portfolio with 8% expected return and 18% volatility will, in roughly two-thirds of years, return between −10% and +26%. That sounds like a wide range. It is.
The deepest peak-to-trough decline an asset has suffered, measured from any past high. Where volatility is statistical, drawdown is visceral — it answers, "how bad did it actually get?" The S&P 500 has experienced peak-to-trough drawdowns of about −86% (1929), −57% (2008), and roughly −34% (2020). Each was followed by recovery, but the path mattered enormously to anyone who had to sell during the descent.
A relative measure: how sensitive an asset's returns are to the overall market. Beta = 1 means the asset moves roughly one-for-one with the market; beta > 1 means it amplifies market moves; beta < 1 means it dampens them. A utility stock might have a beta of 0.5; a high-growth tech stock might have a beta of 1.5; gold has historically exhibited low and unstable equity-market beta.
Beta is most useful as a partial decomposition of total risk: how much of an asset's risk comes from broad market exposure (beta-driven) versus how much is specific to the asset (the rest). It's also the foundation of the Capital Asset Pricing Model — the academic framework for pricing risk that we won't fully unpack here, but that underlies most of how institutional investors think.
The headline empirical claim of modern finance is that, in the long run, riskier asset classes have delivered higher real returns. This is not a guarantee in any specific year or decade — there are 10-year periods when bonds have beaten stocks, and even longer periods in some countries. But across the major developed markets, with a century or more of data, the pattern appears broadly in long-run historical data.
Approximate long-run real returns and volatilities, drawn from the work of Dimson, Marsh, and Staunton, who have compiled the best long-horizon dataset for major markets:
| Asset class (US data, 1900–present) | Real return p.a. | Volatility |
|---|---|---|
| Treasury bills (cash equivalents) | ~ 0.6% | ~ 4% |
| Long-term government bonds | ~ 2.0% | ~ 10% |
| Investment-grade corporate bonds | ~ 2.5% | ~ 8% |
| Equities (S&P 500 type) | ~ 6.5% | ~ 20% |
| Small-cap equities | ~ 8.0% | ~ 28% |
| Gold | ~ 0.8% | ~ 16% |
Historical estimates vary substantially depending on start date, methodology, and market. These numbers should be treated as illustrative. The tendency for equities to outperform safer assets over long horizons appears broadly across many countries and historical periods.
Two observations matter. First, the equity premium — the gap between equity returns and risk-free returns — is roughly 5–6 percentage points per year in the US data. That premium is what compensates equity holders for bearing the volatility, the drawdowns, and the occasional decade where stocks underperform bonds. Strip out a few percentage points of expected return from stocks and the case for holding them collapses. The premium is not optional.
Second, the relationship is not perfectly tight. Gold is more volatile than long-term bonds but has historically returned less. Small caps are roughly 50% more volatile than the broad market but have historically returned only about 25% more. Risk and return are correlated, but the correlation is not 1.0 — and you can find asset classes that look efficient (high return per unit of risk) and others that don't.
The same equity premium shows up across countries — though with meaningful differences in size, driven mostly by historical accidents (wars, hyperinflations, regime changes):
| National equity market (1900–present) | Real return p.a. | Volatility |
|---|---|---|
| 🇺🇸 United States | ~ 6.6% | ~ 20% |
| 🇬🇧 United Kingdom | ~ 5.4% | ~ 20% |
| 🇦🇺 Australia | ~ 6.7% | ~ 17% |
| 🇨🇭 Switzerland | ~ 4.5% | ~ 19% |
| 🇯🇵 Japan | ~ 4.1% | ~ 30% |
| 🇩🇪 Germany | ~ 3.3% | ~ 33% |
| 🌍 World ex-US | ~ 4.5% | ~ 18% |
Source: Dimson, Marsh & Staunton long-horizon equity data, illustrative figures.
A few patterns stand out. Germany and Japan show lower long-run real returns than the US — the result of catastrophic 20th-century episodes (wars, hyperinflations, lost-decade stagnations). Even Switzerland, famously stable, has underperformed because its currency has been unusually strong in real terms — a good thing for the country, but it pushes nominal-currency-denominated returns lower. The US has been the standout performer of the past century, but past performance is not a guarantee.
Historical equity-market data systematically over-represents successful markets. Russia, China, and several others had functioning stock exchanges in 1900 that were subsequently wiped out by revolution or expropriation — taking decades of investor wealth with them. Long-run "average" return figures across "major markets" are biased upward because they generally exclude the markets that ceased to exist. The lesson: no risk premium is guaranteed, anywhere, ever. The premium is what you should expect on average across many possible futures, only one of which actually unfolds.
Most things in finance are tradeoffs — give up something to get something else. Diversification is the rare exception. By combining assets that don't move perfectly together, you can reduce the volatility of a portfolio without proportionally reducing its expected return. The market gives this away for free.
The math, in its simplest two-asset form:
where w is the weight in each asset, σ is each asset's volatility, and ρ (rho) is the correlation between the two assets, ranging from −1 (perfectly opposite) to +1 (perfectly aligned).
The crucial term is the last one. When ρ = +1, the two assets move in lockstep, and portfolio volatility is just the weighted average of individual volatilities — no benefit. As ρ falls, the last term shrinks, and the portfolio becomes less volatile than the weighted average. When ρ = 0 (uncorrelated), the benefit is substantial. When ρ < 0 (negatively correlated), the portfolio can become less volatile than either asset alone — sometimes dramatically so.
A worked example. Two stocks, each with 20% volatility, in a 50/50 portfolio:
That last case never happens in practice — no two real assets are perfectly negatively correlated — but the principle holds. The lower the correlation, the bigger the free lunch. Tool 01 below lets you explore this directly.
Diversification eliminates specific risk — the risk attached to individual companies, individual countries, individual sectors. As you add more independent assets, the portfolio's specific risk falls toward zero. Most of the benefit shows up in the first 10–20 stocks; beyond that, the marginal reduction is small.
Diversification does not eliminate market risk — the risk that all assets move together because something has shaken the whole system. A global recession hits diversified portfolios too. A war hits diversified portfolios. A pandemic hits diversified portfolios. That residual systematic risk is what investors are ultimately compensated for in the form of the risk premium. The free lunch is real but bounded.
Two assets, your weights, and one correlation. Watch what happens to the portfolio's volatility as you slide correlation from +1 (lockstep) to −1 (mirror image). The chart shows the locus of all possible portfolio combinations.
Two final pieces. The first concerns how time interacts with risk; the second connects everything in this module back to the formulas of Module 03 and forward to capital budgeting in Module 06.
An asset that swings ±20% in any given year is genuinely risky over one year. Over thirty years, the same asset's annualized return swings less, because good years and bad years partially cancel out. The mathematical fact: the standard deviation of the average annual return scales with 1/√T. After 30 years, the year-by-year noise of a 20%-volatility asset has shrunk to about 3.6% per year on an annualized basis.
This is the principle behind age-based investment advice: a 25-year-old saving for retirement at 65 has 40 years for variance to wash out, and typically has greater capacity to tolerate short-run volatility because there is more time for recovery. A 60-year-old retiring at 65 has 5 years, during which a market drawdown could be permanent in real-world terms — there's not enough time to recover before the money is needed. The two should hold different portfolios, not because their preferences differ, but because their time horizons differ.
Tool 02 below makes this visceral. It shows the range of plausible final wealth values given a starting amount, an expected return, and a volatility, over various time horizons.
The math above describes the spread of annualized returns. The spread of final wealth actually widens over time, because the underlying values are multiplicative. After 30 years, you're more likely to be near the median — but the gap between the best and worst plausible outcomes, in dollar terms, is wider than after 5 years. Both statements are true. Most people should care about the annualized version (which narrows), since that's what determines whether they can afford retirement. But the dollar version (which widens) explains why even a long horizon doesn't make risk vanish.
Module 03 taught the present-value formula: PV = FV ÷ (1 + r)n. We treated the discount rate r as a given. Now we know better. The discount rate appropriate to any future cash flow has to reflect the riskiness of that cash flow.
A guaranteed cash flow gets discounted at the risk-free rate. A risky one gets discounted at a higher rate that includes a premium reflecting how risky it is. The riskier the cash flow, the higher the discount rate, and the lower its present value.
This is why a venture capitalist values a startup's projected $10M profit five years out very differently than a bank values a government's promise to pay $10M five years out. Both are projections of the same nominal sum. The startup cash flow is highly uncertain — discount it at 25% and its present value is just $3.3M. The government payment is nearly certain — discount it at 4% and its present value is $8.2M. The same future number maps to different present values because the risks behind them differ.
Module 06 will build directly on this insight. Capital budgeting — the framework for deciding which projects, investments, or major personal financial decisions are worth pursuing — is essentially the disciplined application of present-value math at the right discount rate. Risk-adjusted discount rates are the tool that makes the whole framework honest.
Given an expected return and a volatility, what's the realistic range of outcomes for a starting amount over time? The chart shows the median path with 25th–75th and 5th–95th percentile bands. Risk is the width of those bands.
Calculated assuming annual log-returns are normally distributed with the parameters above. Real markets have fatter tails — meaning extreme outcomes are more common than this model suggests, especially on the downside.
The questions test whether you can think clearly about risk — not just compute it. Choose the answer that follows from the principle.