What you will be able to do
Compute a required return with the CAPM and read it on the security market line.
Explain why a share's total volatility does not set its required return, while its beta does.
Choose a risk-free rate, a market premium and a beta for a 2026 valuation, and show how much the answer moves with each choice.
Name the main limits of the CAPM and what size, value, momentum and liquidity premia add.
Read a yield curve, bootstrap zero-coupon rates from coupon bonds and compute a forward rate.
Discount each cash flow at the rate of its own maturity and explain why this matters.
1The CAPM: only market risk earns a premium
Investors are not paid for every risk they take. They are paid for the part they cannot get rid of. The CAPM turns that sentence into a formula.
A diversified investor ignores risks that cancel out
Two owners of the same company demand different returns
Léa Moreau owns a large share of Harmony and most of her wealth depends on it. If the bike market collapses, or a battery recall hits, or a dealer goes bankrupt, she suffers. An investor holding a thousand shares across Europe barely notices a recall at Harmony, because the other nine hundred and ninety-nine companies carry on. What even she cannot escape is Harmony's tendency to fall when the whole market falls.
That tendency is called beta, and it is the only risk measure the model prices
Beta measures how much a share moves, on average, when the market moves by 1%. A beta of 1 means the share moves in step with the market. A beta of 1.34, Harmony's, means it tends to move 1.34% for each 1% in the market, in both directions. A beta of 0.6 means a calmer share, such as a food retailer or a water utility.
1
With rF = 3.0% and an equity risk premium of 5.5%, what does the CAPM require from Harmony's shares (beta 1.34)?
Right. 3.0% + 1.34 × 5.5% = 10.37%, rounded to 10.4%.
That is the premium part only; the risk-free rate is missing.
That is the market's own return, for a beta of 1.
Company-specific risk is diversified away and earns no premium.
A small biotech swings 40% a year, but its beta is 0.5. Why does it require less than Harmony?
Volatility mixes market and specific risk; only the first earns a premium.
The window affects estimates, not the principle.
The share is very volatile; its risk is simply diversifiable.
Right. Only co-movement with the market is paid for.
The equity risk premium rises from 5.5% to 6.5% while rF stays at 3.0%. What happens to the security market line?
A higher premium makes it steeper, not flatter.
That is what a one-point rise in the risk-free rate does.
Right. Harmony goes from 10.37% to 11.71%.
The slope of the line is the premium.
2Feeding the model in 2026
The CAPM is only as good as its three inputs, and none of them can be observed directly. Each one needs a choice, and each choice moves the result.
The risk-free rate must match the currency and the horizon of the cash flows
Use a government bond, in the same currency, with a maturity close to the cash flows
The closest thing to a risk-free rate is the yield on the debt of a government that issues its own currency or, in the euro area, of the highest-rated states. For euro cash flows, practitioners use German government bonds or a euro-area curve built from them. For dollar flows, US Treasuries. For sterling, UK gilts. The rule is consistency: euro flows get a euro rate, because a dollar rate carries different inflation expectations.
LIBOR has disappeared. The short-term references are now SOFR for the dollar, SONIA for sterling and, for the euro, €STR alongside the reformed EURIBOR, which most euro loans still use. They sit at the short end of the curve (Part 4).
Beta is a noisy estimate, so improve it before you trust it
A regression beta depends on the window, the frequency and the index
The usual estimate is a regression: plot the share's returns against the market's returns over past periods and take the slope. Different choices give different slopes for the same company. Five years of monthly returns may give 1.20 for Harmony, two years of weekly returns 1.48. Those two estimates imply required returns of 3.0% + 1.20 × 5.5% = 9.6% and 3.0% + 1.48 × 5.5% = 11.14%. The gap is 1.5 points, from a choice nobody would call fundamental.
Two repairs make betas more reliable: pull them towards 1, and use the peers
Extreme betas tend to drift back towards 1 over time. A common adjustment, named after the economist Marshall Blume, takes two thirds of the measured beta and one third of 1. A measured beta of 1.60 becomes 0.67 × 1.60 + 0.33 × 1 = 1.40. A better route still is the average beta of comparable firms (the bottom-up beta, CF25), which has much less noise than one regression.
A defensible rate shows where each input comes from
2.1 Two betas, one decision
A broker reports a regression beta of 1.80 for a young competitor of Harmony. Compute the Blume-adjusted beta. With rF = 3.0% and a premium of 5.5%, give the required return from the raw and from the adjusted beta. By how many points do they differ?
Adjusted beta: 0.67 × 1.80 + 0.33 = 1.536. Required return: raw 3.0% + 1.80 × 5.5% = 12.9%, adjusted 3.0% + 1.536 × 5.5% = 11.45%. The gap is about 1.45 points. An extreme measured beta probably includes chance, so pulling it towards 1 is sensible.
2
Which risk-free rate suits a long-term valuation of euro cash flows?
A short rate paired with a long-run premium is inconsistent.
Right. Same currency, similar maturity.
That includes a credit spread; it is not risk-free.
Dollar rates carry different inflation expectations.
A broker reports a regression beta of 1.80. What is the Blume-adjusted beta?
That is two thirds of the beta, without the third of 1.
The adjustment pulls towards 1 but does not reach it.
That is the raw estimate, before any adjustment.
Right. 0.67 × 1.80 + 0.33 = 1.536.
Claire Dubois tests reasonable extremes and finds Harmony's cost of equity between about 7.5% and 13%. What should she report?
Right. A decision that only works at one end is not robust.
Two decimals suggest a precision the inputs cannot deliver.
Padding the rate hides the doubt in a number nobody can check.
A different point is still a point; the range is the message.
3Where the CAPM falls short and what richer models add
The CAPM is a clean idea and a rough instrument. Knowing where it breaks decides when to add a premium, and when to leave the model alone.
The CAPM rests on assumptions that reality only half respects
Three weaknesses are worth remembering
The market portfolio is not observable. The theory needs every risky asset in the world, so a stock index stands in for it, and the choice changes betas. Richard Roll argued in 1977 that the model can therefore never be cleanly tested.
Beta is unstable. It moves as the business, the debt and the size of a company change.
The line is flatter than predicted. In many studies low-beta shares earned more than the line gives, and high-beta shares less.
None of this makes the model useless. It separates time from risk and forces you to say which risk you are paying for. The danger is false precision: quoting a cost of equity to two decimals as if the model could deliver it.
Multifactor models price several sources of risk, not just one
If there is more than one way to be hurt in bad times, there is more than one premium
The arbitrage pricing theory of Stephen Ross allows several broad risks, each with its own premium. The best-known practical version is the Fama and French model. It adds a size factor (small companies have historically earned more than large ones) and a value factor (cheap companies relative to book value have earned more than expensive ones). Practitioners also use momentum, quality and low volatility. Some add a size, liquidity or country premium by hand. Each such addition needs a reason, and the evidence for a size premium has weakened in several markets.
k = rF + β market × premium market + β size × premium size + β value × premium value
The limits call for a cross-checked range, not a better point
3
Which of these is a known weakness of the CAPM?
It pays only for market risk.
Right. With an unobservable market portfolio and unstable betas, it is one of three main flaws.
Beta is central and noisy.
The risk-free rate is its starting point.
With illustrative exposures (market 1.25, size 0.5, value −0.2) and premiums (5.5%, 2.0%, 1.5%), the three-factor model gives Harmony…
The value exposure is negative and lowers the result.
The extra factors change the figure, a little here.
Right. 3.0% + 6.875% + 1.0% − 0.3%.
That leaves out the 3.0% risk-free rate and the value term.
An analyst adds a 3-point size premium to Harmony's cost of equity “to be safe”. What is the best response?
The size premium is debated and has faded in several markets.
The model remains a useful benchmark despite its flaws.
An unjustified premium is unjustified everywhere.
Right. The answer is a cross-checked range, not a padded point.
4The yield curve: the price of time depends on the date
So far the risk-free rate was a single number. In reality, lending for one year and for ten years earn different rates. The set of those rates is the yield curve, and it matters for every discounting exercise.
A yield curve plots the rate for each maturity
Zero-coupon rates are the clean version, because one payment has one rate
A zero-coupon bond pays a single amount at maturity and nothing before. The annual return that bond earns is the zero rate for that maturity. Because there is only one payment, the rate is unambiguous. A bond that pays coupons mixes several dates, so its yield to maturity is a sort of average of the zero rates for those dates.
The curve of zero rates by maturity is the term structure of interest rates. Three shapes appear. It is normal when longer rates are higher than shorter ones. It is flat when they are equal. It is inverted when short rates are above long rates. The euro and dollar curves inverted in 2022 and 2023 as central banks raised short rates quickly while longer rates rose less.
Coupon bonds hide the zero rates, which can be dug out one maturity at a time
Most traded bonds pay coupons, so zero rates are not quoted directly. They are derived by bootstrapping: start with the shortest maturity, whose zero rate is simply its yield, then use it to strip out the part of the next bond's price that comes from the first coupon. What remains gives the next zero rate, and so on.
Each cash flow should be discounted at the rate of its own maturity
If a payment arrives in two years, the right discount rate is the two-year zero rate, not the three-year rate and not an average. Discounting each flow at its own zero rate is the consistent way to value a stream of known payments.
a curve builder where the learner drags the zero rates for one to five years. The page redraws the curve, shows the implied forward rates and the present value of a payment stream such as Harmony's contract. The learner sees how a steeper or inverted curve changes values, and where flat-rate shortcuts go wrong.
A forward rate is the rate for a future year that today's curve already implies
Two ways to invest for two years must give the same result, which pins down the future rate
Suppose you can lend for two years at z₂, or lend for one year at z₁ and then lend again for a year at a rate fixed today. If both routes earn the same, the rate for the second year is determined by the curve. It is the forward rate.
(1 + z₂)² = (1 + z₁) × (1 + f₁,₂) so f₁,₂ = (1 + z₂)² / (1 + z₁) − 1
From the curve above: f₁,₂ = 1.0301² / 1.025 − 1 = 3.52%, and f₂,₃ = 1.0342³ / 1.0301² − 1 = 4.25%. A rising curve implies rising forward rates. A forward rate is a rate you can lock in today, for example with a forward contract, which is why it is a price and not a prediction.
What the shape of the curve tells a CFO
The curve says what fixed and floating funding cost, and which risk-free rate to use
Choosing the risk-free rate. Match the maturity of the cash flows.
Fixed against floating debt. Floating debt follows the short end of the curve (CF33, CF43).
Pricing corporate debt. Borrowing cost is the risk-free curve plus a credit spread (CF17). The banker's view is in the DCM module on this site.
Zero and forward rates can be built from any two bonds
4.1 A two-year curve from one bond
The one-year zero rate is 2.0%. A two-year bond with a 3.0% coupon sells for 100. Find the two-year zero rate, then the forward rate for the second year.
100 = 3 / 1.02 + 103 / (1 + z₂)². The first term is 2.941, so 103 / (1 + z₂)² = 97.059, and z₂ = 3.02%. The forward rate is 1.0302² / 1.02 − 1 = 4.04%.
The second year costs about 2 points more than the first, which is what a rising curve implies.
Hugo, a taxi driver with no finance training, hears on the radio that the yield curve has inverted. Explain to him what that means, using the idea of a bank that pays more for a one-year deposit than for a ten-year deposit, and say in one sentence why it is not a guarantee of anything.
Normally a bank pays more if you lock money away for ten years than for one. When the curve inverts, it pays more for one year than for ten. That happens when short-term rates are high, often because the central bank is fighting inflation, and people expect them to fall.
It is a signal, not a guarantee. Inverted curves have often come before slowdowns, but the link is a habit, not a law, and the timing varies.
4.2 Why is the yield to maturity of a coupon bond a mixture of several zero rates?
A coupon bond pays at several dates, each worth its own zero rate. The yield to maturity is the single rate that reproduces the price, so it is a weighted average of those zero rates.
4.3 Harmony has one payment of 10 due in five years. Which rate discounts it, and why not an average of the one-to-five-year rates?
The five-year zero rate: one payment, one date, one rate. An average would give weight to dates with no payment.
4.4 A curve is normal, with a ten-year rate of 3.0% and a one-year rate of 2.0%. Does that prove the market expects the one-year rate to reach 3.0%? Explain.
No. The gap can come entirely from a term premium. The market may expect rates to stay near 2.0% and still price the ten-year at 3.0%.
4
The one-year zero rate is 2.5% and the two-year zero rate 3.01%. What is the forward rate for the second year?
That averages the two rates instead of linking them.
That is the gap between the two rates, not a rate.
That is the two-year rate itself, an average of both years.
Right. 1.0301² / 1.025 − 1.
Harmony has one payment of 10 due in five years. Which rate discounts it?
Future one-year rates are unknown today.
Right. One payment, one date, one rate.
A coupon bond's yield mixes several zero rates.
That would weight dates with no payment.
The curve is normal: 2.0% at one year, 3.0% at ten. Does the market expect the one-year rate to reach 3.0%?
Right. A forward rate is a price, not a forecast.
An inverted curve is the one usually read as expected falls.
The curve does not date anything with certainty.
That is the pure expectations theory, which ignores the term premium.
Summary and cheat sheet
Les formules du module
Touchez un terme pour voir ce qu’il veut dire et sa valeur chez Harmony.
Modèle d’équilibre des actifs financiers (MEDAF)
= + ×
3,0 % + 1,34 × 5,5 % = 10,37 %, arrondi à 10,4 % dans le reste du cours.
Bêta ajusté (Blume)
= 0,67 × + 0,33 × 1
Taux à terme
= (1 + )² / (1 + ) − 1
Un prix qu’on peut bloquer aujourd’hui, pas une prévision : il contient une prime de terme.
Valeur d’un flux certain sur la courbe
= Σ / (1 + )ᵗ
| Idea | What to remember |
|---|---|
| Why only market risk is paid for | Diversified investors set prices, so company-specific risk cancels out for them. Only co-movement with the market earns a premium. |
| Beta | Sensitivity of a share to the market. 1 means in step with it, above 1 amplifies it, below 1 dampens it. |
| CAPM | k = rF + β × equity risk premium. Harmony: 3.0% + 1.34 × 5.5% = 10.37%. |
| Security market line | Required return plotted against beta. A change in rates shifts it up or down. A change in the premium tilts it. |
| Risk-free rate | Government yield, same currency, similar maturity as the cash flows. Never mix a short rate with a long-run premium. |
| Equity risk premium | Estimated, never observed. Historical or implied by prices. Quote a range, about 4% to 6.5% for mature markets. |
| Multifactor models | Add premia for size, value, momentum, quality and others. Useful for very small or unusual firms. Adjustments need a reason. |
| Yield curve | Zero rate by maturity. Normal, flat or inverted. Each cash flow is discounted at the rate for its own date. |
| Forward rate | f = (1 + z₂)² / (1 + z₁) − 1. A rate you can lock in today. It contains a term premium, so it is not a forecast. |
The one-sentence version
A required return is the price of time at the right maturity plus the price of the market risk the asset carries, and honest inputs give a range, not a point.
Key terms
| Term | Plain-language meaning |
|---|---|
| Beta | How much a share moves, on average, when the market moves by 1%. |
| Equity risk premium | The extra return investors expect from the stock market over a risk-free government bond. |
| Security market line | The straight line of required return against beta. |
| Risk-free rate | The yield on a government bond in your own currency, with no default worry and a chosen maturity. |
| Multifactor model | A model that explains required returns with several risk sources, each with its own premium. |
| Zero rate | The annual return on a bond that pays one amount at one date and nothing before. |
| Yield curve | The rates for every maturity, plotted against maturity. |
| Bootstrapping | Extracting zero rates from coupon bonds, one maturity at a time. |
Going further
Vernimmen, Quiry, Le Fur — Finance d'entreprise (Dalloz), chapters on required return and market equilibrium, and on the term structure of interest rates.
Sharpe (1964) — Capital asset prices: a theory of market equilibrium under conditions of risk, Journal of Finance.
Fama and French (1993) — Common risk factors in the returns on stocks and bonds, Journal of Financial Economics.
Brealey, Myers, Allen — Principles of Corporate Finance, chapters on risk and return and on the term structure of interest rates.
Independent educational material. Not investment, legal or tax advice. All companies and figures in the running case are fictional.