A dollar a year from now is worth less than a dollar today, and a risky dollar is worth less than a certain one. Almost everything in finance is an answer to one of two questions that follow from those two facts: what is a future payment worth, and what is a risky claim worth? Consumer theory already taught the first idea in disguise: choosing between consumption now and consumption later is a choice along a budget line whose slope is set by the interest rate. Finance picks up exactly there and pushes into the harder territory: claims whose value is uncertain.
The apparatus assembled here is the modern teaching core of asset pricing. It runs from Fisher's present value, through Markowitz's insight that risk is a property of a portfolio rather than of a single asset, to the Capital Asset Pricing Model's verdict that only one kind of risk is ever paid for, and then to the most general statement of all, that every price in a market obeys a single pricing rule. From that rule the chapter recovers two things that look like separate subjects: option pricing, which turns out to need no forecast of returns at all, and the long argument over whether prices are right.
This is where the economics book owns the asset-pricing model. The intellectual history (Fisher in 1907, Markowitz in 1952, Sharpe in 1964, Black, Scholes, and Merton in 1973) belongs to the history-of-economic-thought book and is named here only in passing. The crisis chronology belongs to the economic-history book. What follows is the machinery itself, taught once and cleanly, in the form a working economist uses it.
Prerequisites: intertemporal choice, utility maximization, and the Lagrangian (Chapter 5); the interest rate and the no-arbitrage discipline within the government budget constraint (Chapter 16). Econometrics (Chapter 10) and the expected-utility benchmark (Chapter 19) are helpful for the efficiency section but are not assumed.
This chapter is the home of the asset-pricing apparatus that several of the book's Walkthroughs have needed and never had a single place to point to. The pricing equation, the efficiency debate, and the return-to-capital intuitions all live here.
Consumer theory framed saving and borrowing as a choice between consumption today and consumption later. The two are traded at a price, the interest rate. Lend a dollar today and a year on you have $(1+r)$ dollars; to obtain a dollar a year from now you must set aside $1/(1+r)$ dollars today. That ratio, $1/(1+r)$, is the price of a future dollar in terms of present dollars, and it is the seed of everything in this chapter.
Meet Lena, who has just received a settlement she can spend, save, or invest. Across this chapter she faces the questions the apparatus answers, one at a time, and the first is the most basic. A contract promises her payments spread over several years: what is it worth to her now? The answer is to value each future payment at its present-dollar price and add them up.
Discount each payment in a stream by its own discount factor and sum. For payments $C_1, C_2, \dots$ arriving in years $1, 2, \dots$ at a constant rate $r$, the present value is the sum of the discounted amounts.
Why it matters: The interest rate is an exchange rate between two currencies, "money now" and "money later." A payment far in the future is converted into present money at a worse rate the further out it sits and the higher the interest rate climbs, because the alternative (putting money aside now to grow on its own) gets better. Present value is just the total bill once every future payment has been converted into today's money. Lena does not compare a payment next year to a payment in five years directly; she converts both to now and then they are comparable.
Fisher's deeper result is that once a capital market exists, what a person should invest in has nothing to do with how patient or impatient that person is. Lena and a far less patient acquaintance with the same opportunities should make the identical investment choice (take every project whose present value exceeds its cost) and then satisfy their different tastes for spending now versus later by borrowing or lending at the market rate. The objective ranking of investments (maximize present value) is one problem; the personal pattern of consumption over time is a separate problem solved afterward against the budget line.
Formally, with a two-period endowment and the ability to borrow and lend at rate $r$, the consumer chooses present and future consumption to maximize utility subject to the intertemporal budget constraint. The first-order condition equates the marginal rate of substitution between today and tomorrow to the gross return:
At the optimum the rate at which the consumer is willing to trade present for future consumption equals the rate the market offers. The production side, choosing investment to maximize the present value of the endowment, is solved without reference to $u(\cdot)$ at all; that is the separation. Wealth here is written $W$ and income $y$; the symbol $m$ is reserved for §24.4.
Why it matters: Picture the budget line that consumer theory drew between consuming now and consuming later, its slope fixed by the interest rate. A good investment pushes that whole line outward, making Lena richer in present-value terms, and a richer line is better whatever her taste for spending sooner or later. So she should chase the investment that pushes the line out the most, then slide along the new line to whatever now-versus-later mix she prefers, borrowing or lending to get there. Patience decides where she sits on the line; the market decides which line she gets, and present value is how she ranks the lines.
Figure 24.0. The intertemporal budget line. The endowment point is fixed; a productive project shifts wealth from today into a larger future payoff, pushing the whole budget line outward whenever the project's present value is positive. The interest rate sets the slope. Drag the sliders. The economics book's Chapter 5 derives this line from preferences in full.
Problem. The interest rate is $r = 6\%$. (a) A contract pays Lena \$1,000 at the end of each of the next three years. What is its present value? (b) Project A returns \$3,200 in one year; Project B returns \$3,600 in three years. Both cost \$3,000 today. By present value, which should she take, and does her impatience change the answer?
Solution.
(a) $PV = \dfrac{1000}{1.06} + \dfrac{1000}{1.06^2} + \dfrac{1000}{1.06^3} = 943.4 + 890.0 + 839.6 = \$2{,}673.0$. The three certain payments are worth about \$2,673 today, not \$3,000.
(b) $PV_A = 3200/1.06 = \$3{,}018.9$, net of cost $+\$18.9$. $PV_B = 3600/1.06^3 = \$3{,}022.6$, net of cost $+\$22.6$. Both are worth taking; B has the higher present value, so Project B is preferred.
By Fisher separation, her impatience does not change this ranking. If she needs cash sooner she takes B anyway and borrows against its larger present value; the investment decision (maximize present value) and the consumption-timing decision (borrow or lend) separate.
The idea that the interest rate is the price knitting present to future, and that valuation runs through discounting, descends from Irving Fisher's work of 1907 and 1930. That lineage, how the marginalist treatment of capital and interest gave finance its starting point, is traced in the history-of-economic-thought book's chapter on the marginalist revolution. The interest rate itself, and the no-arbitrage discipline that governs it, is built in Chapter 16 on monetary and fiscal theory.
Present value answers what a certain future claim is worth. Real claims, like a share of a company or a bond that might default, are not certain. The next question is how to hold them, and the answer overturns the natural instinct to find the single best one.
The sensible-sounding way to invest is to study companies, find the ones with the best prospects, and put your money there. Lena's instinct is to do exactly this: pick the strongest assets, avoid the weak ones, and her portfolio will be as strong as the assets in it. The reasoning is not foolish. A better asset really does offer a better deal, and the work of judging assets one by one is the work most of the industry does.
Markowitz's contribution in 1952 was to show that this frame, while not wrong about individual assets, misses where risk actually lives. Risk is not a property an asset carries into a portfolio unchanged. The risk of a portfolio depends on how its assets move together, and combining assets that do not move in lockstep destroys risk that neither asset could shed alone. Diversification, in his phrase, is the one free lunch in finance: a reduction in risk that costs nothing in expected return.
A portfolio is a set of weights $w_i$ summing to one, the share of wealth in each asset. Its expected return is simply the weighted average of the assets' expected returns:
The reward combines linearly. Risk does not. Portfolio variance includes not only each asset's own variance but every pair's covariance, and it is the covariance terms that make diversification work:
For two assets with weights $w$ and $1-w$, this is the relation whose dependence on correlation $\rho$ is the whole point:
When $\rho = 1$, the last term is at its maximum and the standard deviation is just the weighted average $w\sigma_1 + (1-w)\sigma_2$, with no risk reduction at all. As $\rho$ falls, the cross term shrinks and the portfolio's standard deviation drops below that weighted average. At $\rho = -1$ the two risks can be made to cancel exactly. Expected return is untouched throughout; only risk falls. That is the free lunch stated in symbols.
Why it matters: Two assets that lurch the same way at the same time give you no protection: when one is down the other is down, so holding both is no safer than holding either. But if one tends to be up when the other is down, their bad days partly cancel, and the combined holding swings less than either piece. You did not pick better assets to get this; you only stopped putting all the risk in one direction. The reward, the average return, is exactly the blend of the two averages, untouched. So the risk drops while the reward holds. On the figure below, drag the correlation slider down from $+1$ toward $-1$ and watch the curve of possible portfolios bow leftward toward less risk. Nothing about the assets changed; only how they are combined.
Trace out every portfolio of the available assets and the achievable combinations fill a region in risk-return space. Its upper-left edge, the place of most return per unit of risk, is the efficient frontier. Every portfolio strictly inside the frontier is dominated: there is another with the same risk and more return, or the same return and less risk. The lower the correlations among the assets, the further the frontier bows toward the low-risk axis, because diversification is doing more.
Introduce a risk-free asset and the investor can mix it with any risky portfolio along a straight line, the capital allocation line, whose steepness is the reward-to-risk ratio of that risky portfolio. The best line is the steepest one still touching the frontier, and it touches at a single point: the tangency portfolio. Every investor, cautious or bold, wants that same risky portfolio; they differ only in how they split between it and the risk-free asset. A timid Lena lends most of her wealth at $R_f$ and holds a sliver of the tangency portfolio; a bold one borrows to hold more than her wealth in it. Neither holds any other risky mix. The question "which risky assets should I own?" separates cleanly from "how much risk do I want?"
Why it matters: Once a safe asset is on the table, raising and lowering your risk no longer means trading one risky bet for another. It means dialing one knob: how much of your money sits safe versus how much rides on the best risky bundle there is. And the best risky bundle is the same for everyone, because it is the one with the highest reward per unit of risk — preferences do not enter that comparison. So the hard part of investing (which risky things to own) has one answer for all, and the personal part (how bold to be) is just where you set the dial. On the figure, raise the risk-free slider and watch the straight allocation line pivot to a new tangency point — that point is the bundle everyone should hold.
Figure 24.1. The efficient frontier of two risky assets, with the capital allocation line drawn to the risk-free asset. Lower correlation bows the frontier leftward — diversification at work. The straight line from $R_f$ touches the frontier at the tangency portfolio every investor holds. Drag the correlation slider from +1 toward −1 to see the free lunch appear.
Problem. Two assets each have $\sigma = 20\%$ and equal expected return; Lena holds them 50/50. Compute the portfolio standard deviation at (a) $\rho = 1$, (b) $\rho = 0$, (c) $\rho = -1$.
Solution. With $w = 0.5$ and $\sigma_1 = \sigma_2 = 0.20$, Eq. 24.5 gives $\sigma_p^2 = 0.25(0.04) + 0.25(0.04) + 2(0.25)\rho(0.04) = 0.02 + 0.02\rho$.
(a) $\rho = 1$: $\sigma_p^2 = 0.04$, so $\sigma_p = 20\%$ — the weighted average, no reduction. (b) $\rho = 0$: $\sigma_p^2 = 0.02$, so $\sigma_p = 14.1\%$ — risk cut by nearly a third for free. (c) $\rho = -1$: $\sigma_p^2 = 0$, so $\sigma_p = 0\%$ — the two risks cancel exactly. Expected return is identical in all three cases; only the correlation moved.
Markowitz's mean-variance framework, set out in 1952, is one of the founding results of the postwar formalization of economics. Its place in that story, alongside the equilibrium asset-pricing models it made possible, is traced in the history-of-economic-thought book's forthcoming chapter on the postwar synthesis.
Markowitz tells one investor what to hold. If everyone solves this same problem and holds the same tangency portfolio, the prices of assets cannot be arbitrary; they must adjust until the market clears. That equilibrium reasoning is the next step, and it delivers a sharp verdict about which risks earn a reward.
If two-fund separation holds for everyone, then in equilibrium the tangency portfolio is the market portfolio, the value-weighted bundle of all risky assets. This single assumption, run through the demand of investors who all hold the market, produces the Capital Asset Pricing Model and its one memorable conclusion: an asset's expected return depends only on how it moves with the market, not on how variable it is on its own.
Lena's next question is concrete. She is considering a single stock. What return must it offer her to be worth holding? Not, it turns out, a return that compensates her for the stock's total variability. Most of that variability she can diversify away by holding the stock inside the market portfolio, and the market does not pay anyone for risk they could have shed for free.
The CAPM relation states the required expected return of any asset:
where beta is the asset's covariance with the market scaled by the market's variance:
Total variance decomposes into a systematic part, $\beta_i^2 \sigma_m^2$, and an idiosyncratic part, $\sigma_{\varepsilon}^2$, that is uncorrelated with the market. Only the first carries a premium. Two stocks with identical total variance but different betas command different expected returns; the high-beta one is paid more because its risk cannot be diversified away, while the other's larger idiosyncratic share earns nothing.
Why it matters: A stock's own wildness is not what the market pays you to bear, because you can cancel most of it by owning many stocks at once, and anything you can get rid of for free is not worth a reward. What you cannot get rid of is the part that drops when the whole market drops; everyone is exposed to that at the same time, so no amount of diversifying escapes it. That undiversifiable part is beta, and it is the only thing priced. A stock that lurches violently on its own news but barely tracks the market is, to a diversified investor, almost as good as safe, and the market prices it that way. On the figure, slide a stock's beta up and watch its required return climb along the line; slide its idiosyncratic risk and watch the required return not move at all.
Figure 24.2. The security market line and the split of total risk. The stock's required return is read off the line at its beta; moving idiosyncratic risk changes total risk but leaves the required return untouched, because only the systematic part is priced. Drag the sliders.
Problem. $R_f = 3\%$, the market premium is $6\%$, a stock has $\beta = 1.4$. (a) What expected return must it offer? (b) Its total standard deviation is $30\%$ and the market's is $16\%$. What fraction of its variance is systematic?
Solution.
(a) $E[R] = 0.03 + 1.4(0.06) = 0.03 + 0.084 = 11.4\%$.
(b) Systematic variance $= \beta^2\sigma_m^2 = 1.96 \times 0.0256 = 0.0502$. Total variance $= 0.30^2 = 0.09$. Systematic share $= 0.0502/0.09 = 56\%$; the remaining $44\%$ is idiosyncratic and earns no premium. Two stocks with this same $30\%$ total risk but different betas would command different returns — only the systematic slice is paid for.
The equilibrium step from Markowitz's frontier to a pricing model was taken independently by William Sharpe, John Lintner, and Jan Mossin in the mid-1960s. Their place in the postwar formalization of finance is traced in the history-of-economic-thought book's forthcoming chapter on the postwar synthesis.
CAPM is a specific equilibrium model, resting on mean-variance preferences and a particular market structure. It works, but a question lurks beneath it: is the security market line a fundamental law of prices, or one special case of something more general? There is a more general law, and it needs almost no assumptions at all.
Strip away mean-variance preferences, the market portfolio, and every other special assumption, and one premise remains that no functioning market can violate: two claims that pay the same thing in every state of the world must cost the same. If they did not, a trader could buy the cheap one, sell the dear one, and pocket a sure profit at no risk. That single premise, the law of one price enforced by arbitrage, is enough to pin down a complete theory of prices.
Lena's question now is the most general one she can ask: what is any claim worth, be it a stock, a bond, an insurance contract, or a lottery on next year's weather? The answer is that there exists a single object, common to the whole market, that prices everything. Multiply a claim's random payoff by this object, take the expectation, and you have its price.
The fundamental pricing equation is the entire apparatus in one line:
Applied to a gross return $R_i = x_i/p_i$, the equation reads $1 = E[m R_i]$. Expanding the expectation of a product and rearranging gives the risk-return relation: an asset's expected excess return over the risk-free rate is set by how its return covaries with the discount factor.
An asset that pays off in bad states, when $m$ is high, has positive covariance with $m$ and therefore a negative risk premium: it is insurance, and investors accept a low return to hold it. An asset that pays off in good states has negative covariance with $m$ and a positive premium. The CAPM falls out as the special case in which $m$ is linear in the market return, $m = a - b R_m$: substitute and the covariance with $m$ becomes proportional to the covariance with the market, which is exactly beta. The security market line is one reading of the pricing equation; it is not the law itself.
The same price can be written a second way. Repackage the discount factor and the true probabilities into a new set of probabilities $q$, and the price becomes a plain discounted expectation at the risk-free rate:
The two formulas, discounting the physical-probability expectation with $m$ or discounting the risk-neutral expectation at $R_f$, are the same statement in different probabilities. The risk-neutral probability $q$ is not a belief anyone holds; it is a derived quantity, read straight off the prices of traded assets via no-arbitrage.
Why it matters: There is a single set of prices-of-risk in a market, and once you know it you can value anything; you do not need a special theory for stocks and another for options. A dollar is worth more to the market in bad times than in good times, and that one fact, attached to every state of the world, is the whole pricing rule. The model from the previous section is just one way of reading those prices off the market return; risk-neutral pricing is the same rule rewritten so that the prices-of-risk hide inside the probabilities and everything discounts at the safe rate. On the figure below, set a one-period payoff and watch two methods, one using the true odds and the price of risk, one using the adjusted "risk-neutral" odds, land on the identical price. They must: they are the same statement. Notice too that the adjusted odds are computed, not believed; the market hands them to you.
Figure 24.3. A one-period claim priced two ways. The bars show the price obtained by discounting the physical-probability expectation with the stochastic discount factor, and by discounting the risk-neutral expectation at the risk-free rate. They agree, because no-arbitrage forces them to. The implied risk-neutral probability $q$ is derived from the underlying's price, not assumed. Drag the sliders.
Problem. A stock trades at \$100 and in one period is worth \$120 (up) or \$90 (down). The risk-free rate is $5\%$. A claim pays \$10 if the stock goes up and \$0 if it goes down. Price it (a) by finding the risk-neutral probability, and (b) confirm the implied prices of risk are consistent.
Solution.
(a) The risk-neutral probability solves $100 = \frac{1}{1.05}(q \cdot 120 + (1-q)\cdot 90)$, giving $105 = 90 + 30q$, so $q = 0.5$. The claim's price is $\frac{1}{1.05}(0.5 \cdot 10 + 0.5 \cdot 0) = \frac{5}{1.05} = \$4.76$.
(b) Note $q = 0.5$ regardless of the true probability of an up move — the physical odds never entered the claim's price. They enter only through the price of the stock itself, which already embeds them. This is the same fact the option section turns into a full pricing formula: once the underlying's price is known, the claim's value follows from no-arbitrage alone.
A stop on the question of what money actually is. The pricing equation supplies the missing piece: money, like any claim, is valued by the same prices of risk.
The fundamental pricing equation (§24.4) says every claim is worth the expected product of its payoff and a single price-of-risk operator. Money is a claim, on goods, on settlement, on the issuer. Its value is set by the same machinery that prices a stock: what it pays in which states, weighted by how much the market values a dollar in those states. The discount factor is high exactly when liquidity is scarce, which is when money's services are worth most.
Pricing theory does not settle what money is, whether commodity, credit, or state liability, but it does say money is not exempt from the rule that values everything else. The premium people pay to hold a zero-return safe asset is, in this language, the price of its payoff in bad states. That is one input the larger walkthrough assembles alongside the historical and institutional accounts of money.
A stop on whether inequality is a problem economics can solve. Piketty's r > g rests on a return to capital, and asset pricing is where that return comes from.
The "$r$" in Piketty's $r > g$ is the average return on wealth, and the apparatus here says what determines it: the risk-free rate plus a premium for bearing systematic, undiversifiable risk (§24.3, §24.4). Diversified capital captures the market risk premium; it is paid for bearing the risk no one can shed, not for any special virtue. That this premium has historically exceeded the growth rate of the economy is what makes accumulated wealth grow faster than wages.
Asset pricing explains the level of $r$ as a risk premium rather than a windfall, which is useful for cutting through the claim that capital income is simply unearned. What it does not settle is the distributional question: who holds the diversified capital that earns the premium, and whether the resulting concentration is a problem. That is the work the inequality walkthrough does with this return as an input.
The pricing equation works on any claim. The hardest claim to value in all of finance is an option, whose payoff bends nonlinearly with the price of something else, and the no-arbitrage idea, pushed into continuous time, solves it with a result that startled the field: the expected return of the underlying never enters.
An option to buy a stock at a fixed price pays off only if the stock rises above that price; otherwise it expires worthless. The natural way to value it would be to forecast where the stock is heading, weight the payoffs by those odds, and discount. Lena's intuition runs this way: a stock she expects to soar should make its call options expensive. The Black-Scholes argument shows this intuition is wrong, and wrong in a way that is genuinely surprising.
Here is the move. Hold a precisely chosen number of shares of the stock, financed partly by borrowing, and rebalance the holding continuously as the stock moves. It is possible to choose the share count so that this stock-and-bond portfolio gains and loses exactly what the option does, instant by instant. If the portfolio reproduces the option's payoff in every state, then by the law of one price the option must cost what the portfolio costs to assemble. No forecast of the stock's direction is needed, only the recipe for tracking it.
And that recipe depends on how much the stock jumps around, not on which way it is expected to drift. The drift, the stock's expected return, drops out entirely. Two analysts who disagree violently about whether the stock will rise or fall must still agree on the option's price, provided they agree on its volatility.
The Black-Scholes call price is:
Here $S$ is the stock price, $K$ the strike, $r$ the continuously-compounded risk-free rate, $T$ the time to expiry, $\sigma$ the volatility, and $N(\cdot)$ the standard-normal cumulative distribution. The stock's expected return $\mu$ appears nowhere. This is the formal echo of the risk-neutral pricing of §24.4: the option is valued as if the world were risk-neutral, discounting at $r$, because the replicating portfolio has stripped the risk premium out.
Why it matters: You can build a moving mixture of the stock and a loan that mimics the option tick for tick: when the stock rises you hold more of it, when it falls you hold less, rebalancing as you go. If your homemade mixture pays exactly what the option pays, it must cost exactly what the option costs, or else someone sells the dear one, buys the cheap one, and banks the difference for free. The startling part is what the recipe needs. To track the option you need to know how much the stock wobbles, because that sets how often and how hard you rebalance. You do not need to know which way it is heading. So the forecast that feels like it should matter most, whether the stock will go up, does not enter the price at all. Volatility does all the work. On the figure, slide volatility up and watch the option get dearer; look for a place to set the expected return and find there is none.
Figure 24.4. The Black-Scholes call price as a function of volatility, with the current setting marked. The price rises monotonically with volatility. Nowhere among the inputs is the stock's expected return $\mu$ — there is no slider for it, because the formula has no place for it. Drag the sliders.
Problem. $S = 100$, $K = 100$, $r = 4\%$, $T = 1$ year. (a) Price the call at $\sigma = 20\%$. (b) Reprice at $\sigma = 40\%$ and confirm the price rises. (c) The stock's manager privately expects a $30\%$ return this year. By how much does that change the price?
Solution.
(a) $d_1 = \frac{\ln 1 + (0.04 + 0.02)(1)}{0.20} = 0.30$, $d_2 = 0.10$. $N(0.30) = 0.618$, $N(0.10) = 0.540$. $C = 100(0.618) - 100 e^{-0.04}(0.540) = 61.8 - 51.9 = \$9.9$.
(b) At $\sigma = 40\%$: $d_1 = \frac{0.04 + 0.08}{0.40} = 0.30$, $d_2 = -0.10$. $N(0.30) = 0.618$, $N(-0.10) = 0.460$. $C = 61.8 - 96.08(0.460) = 61.8 - 44.2 = \$17.6$. Doubling volatility nearly doubled the price.
(c) By exactly zero. The expected return $\mu$ does not appear in the Black-Scholes formula. The manager's optimism would change what the stock is worth, but not what a contract on the stock is worth relative to it — the replicating portfolio prices the option off the stock's current price and volatility alone.
The formula is powerful and its assumptions are fragile, and both have always been true together. Black-Scholes assumes constant volatility and continuous, costless rebalancing in a market that never gaps. Reality supplies none of these. The market's own option prices reveal the strain: plug observed prices back through the formula and the implied volatility is not constant across strikes but smiles upward at the extremes, a standing admission that the constant-volatility assumption is wrong. And in a crisis, the continuous-rebalancing premise fails exactly when it is needed most — prices gap, liquidity vanishes, and the correlations the hedger relied on snap toward one. Long-Term Capital Management discovered this in 1998, when positions that were near-riskless under the model's assumptions became ruinous once those assumptions broke. The honest statement is that the model gives a disciplined first answer and a clear map of its own failure points; it is not extended here into the stochastic-volatility machinery that tries to patch the smile.
Long-Term Capital Management. A fund built around the no-arbitrage pricing of fixed-income and derivative claims, advised by two of the economists who built the option-pricing theory, lost most of its capital in weeks in 1998 when a Russian default drove correlations toward one and liquidity out of the markets it traded — the tail behavior the constant-volatility model does not see. The episode is the canonical illustration that the math and the fragility are inseparable. The economic-history book's chapter on globalization and the great moderation sets it in context.
The option-pricing breakthrough of Fischer Black, Myron Scholes, and Robert Merton in 1973 belongs to the wave of formal modeling that the history-of-economic-thought book treats in its chapter on the counter-revolution in economic thought. The derivatives markets the formula helped create scaled enormously in the decades after; that growth is part of the economic-history book's account of late-twentieth-century financial globalization.
The thread that runs from Fisher's rational investor to behavioral finance. Its apparatus (present value, the frontier, CAPM, option pricing) is owned here; the thread traces the lineage and compresses to these sections.
This chapter is where the four apparatus moves of the Fisher-to-behavioral thread are built in full: Markowitz's frontier (§24.2), the CAPM (§24.3), the no-arbitrage pricing equation (§24.4), and the option-pricing argument (§24.5). The thread originally derived each inline, for want of a home; with the finance chapter in place, those derivations compress to references here and the thread can spend its length on the lineage and the eventual behavioral challenge rather than re-teaching the models.
The chapter owns the model; the thread owns the arc: how the rational-investor program was built, where it strained, and how behavioral finance (excess volatility, limits to arbitrage) became a coherent challenge to it. The behavioral terminus lives in the economics book's behavioral and experimental chapter. This stop hands the thread its assembled tools.
Every model so far has assumed that the prices feeding into it are sensible: that the stock price the option is hedged against, the market return CAPM uses, the cash flows present value discounts, all reflect available information. That assumption was never proved. It is the efficient-markets hypothesis, and it is the most fought-over claim in the whole subject.
Start with the strongest version of the case for efficiency, because it is far more powerful than its caricature. If prices already reflect all available information, then no analysis of that information can systematically beat the market, because everything knowable is already in the price. A new piece of news moves the price immediately and fully; what comes next is, by construction, unpredictable. This is not a claim that markets are wise or that prices are "correct" in some cosmic sense. It is the claim that there is no free money lying on the ground, and the reason is almost tautological: if there were, someone would already have picked it up, and the act of picking it up would have moved the price until the opportunity was gone.
Run as a null hypothesis, this is formidable. The overwhelming majority of actively managed funds underperform a passive index over any long horizon, after fees. Public information (earnings, ratios, news) is priced so fast that trading on it earns nothing reliable. The burden of proof sits, correctly, on anyone who claims to beat the market: the default expectation is that they cannot, and most who try cannot. Lena, surveying the evidence, should start as an efficient-markets believer and demand strong evidence to move off it.
Stated formally, efficiency requires that returns be unpredictable given the information set $I_t$ beyond their equilibrium expectation:
Here is the difficulty that keeps the question alive. To say a return is abnormally high, which is the evidence against efficiency, you must first say what a normal return would have been, which requires an asset-pricing model. If you find an abnormal return, you cannot tell whether the market was inefficient or whether your model of normal returns was simply wrong. Every test of efficiency is a joint test of efficiency and the pricing model, and the two cannot be pried apart. This is not a technicality; it is the reason the debate cannot be settled by data alone.
Through the 1980s, researchers documented patterns CAPM could not account for: small companies, cheap "value" stocks, and recent winners earned more than their betas predicted. Fama's own response was not to abandon efficiency but to enrich the model. If small and value stocks reliably out-earn, perhaps they bear some additional systematic risk the market factor missed, and the extra return is its price. The Fama-French three-factor model formalizes this:
The behavioral reading of the very same evidence is different. To Robert Shiller and the behavioral-finance school, the anomalies are not hidden risk factors but the footprints of mispricing: prices that overreact, herd, and swing with sentiment more than fundamentals can justify. Excess volatility, the finding that stock prices move far more than the eventual stream of dividends warrants, is their central exhibit. And mispricing can persist because arbitrage is limited.
Why it matters: The hard part of judging market efficiency is that you can never measure "too high a return" without first deciding what a fair return would have been, and that decision is itself a theory that might be wrong. So when small or cheap stocks beat the market for decades, two honest stories fit the same data. One says: those stocks must carry some extra risk the simple model missed, and the higher return is just payment for it. The other says: the market is simply getting those prices wrong, and the mispricing survives because betting against it is dangerous and slow. The data alone cannot referee between them, which is why the dispute is real and not a sign that one side is confused. On the figure, set a portfolio's exposures and watch the extra return get attributed to factor risk; then flip the toggle and watch the identical number get called mispricing instead. The number does not change — only the story you are willing to tell about it.
Figure 24.5. Factor attribution of a portfolio's excess return. The bars decompose the observed return into market, size, and value contributions plus a residual "alpha." The toggle re-labels that same residual — skill or mispricing (behavioral) versus an unmeasured risk factor (Fama) — without changing the number. That ambiguity is the joint-hypothesis problem. Drag the sliders; flip the toggle.
Problem. A fund earns $11\%$ in excess of the risk-free rate. Its loadings are $\beta_M = 1.0$, $\beta_S = 0.5$, $\beta_V = 0.4$. Factor premia are $\mathrm{MKT} = 6\%$, $\mathrm{SMB} = 2.5\%$, $\mathrm{HML} = 3.5\%$. (a) How much of the return is explained by factor exposure? (b) What is the residual alpha, and what are the two competing interpretations?
Solution.
(a) Explained $= 1.0(6) + 0.5(2.5) + 0.4(3.5) = 6 + 1.25 + 1.4 = 8.65\%$. Most of the apparent out-performance is exposure to priced risks, not skill.
(b) Residual $\alpha = 11 - 8.65 = 2.35\%$. The joint-hypothesis problem means this number is unidentified: it is either genuine skill or mispricing the fund captured (the behavioral reading), or it is compensation for a risk factor the three-factor model still omits (the Fama reading). The data cannot say which; the same $2.35\%$ supports both stories.
The honest state of the question has three layers, and collapsing them into "markets are sort of efficient" loses what matters. First, there is broad consensus that the asset-pricing apparatus is durable and that prices are genuinely hard to beat: the efficient-markets hypothesis is a powerful, largely-correct first approximation, and the failure of active management to beat indexes is its standing confirmation. Second, there is broad consensus that the strong form is false: prices do not reflect everything, sentiment moves them, and the 2013 Nobel Prize awarded jointly to Eugene Fama and Robert Shiller made the dual verdict official: the man who formalized efficiency and the man who documented its excess-volatility failures were honored together, which is the discipline's way of saying both are right about different things. Third, and genuinely unsettled, is the magnitude and the mechanism: how much efficiency survives, and whether the residual patterns are risk premia (Fama) or mispricing (Shiller). This last layer is a live disagreement, not a settled fact dressed up as debate. The point is not to split the difference but to be precise about which layer is consensus and which is open.
The behavioral side of this argument rests on departures from the rational-choice benchmark (loss aversion, overreaction, herding) whose foundations are laid in the economics book's chapter on behavioral and experimental economics. The finance-specific results, excess volatility and the limits to arbitrage that let mispricing persist, are what this section teaches; the general psychology of decision under risk is not re-derived here.
There is a related but distinct sense in which markets can be "efficient": allocative efficiency, whether prices steer capital to its most productive uses. That is a different question from the informational efficiency treated here, and the two should not be run together; the walkthrough on whether financial markets are efficient keeps them separate.
The public stress test. The 2008 financial crisis became the popular indictment of efficient markets and of the models built on them. Prices that had reflected a consensus of safety collapsed, and the limits to arbitrage were on open display as mispricings widened before anyone could correct them. The crisis chronology, and the question of how much the models themselves were to blame, belong to the economic-history book's chapter on the global financial crisis and after and to the walkthroughs that argue it; this section names 2008 only as the moment the efficiency debate left the seminar room.
The efficient-markets hypothesis was given its canonical form by Eugene Fama in 1970, and the doctrinal home of the Chicago efficiency tradition, together with its critics, is the history-of-economic-thought book's chapter on the counter-revolution; the behavioral challenge that grew up against it is traced in that book's treatment of modern pluralism in economic thought.
The informational-efficiency stop. You now hold the EMH at full strength, the joint-hypothesis problem, factor models, and limits to arbitrage — here is what they add up to.
Prices reflect available information because anyone who finds an exploitable pattern trades it away. The empirical backbone is real: active managers cannot reliably beat passive indexes after fees, and public information is priced near-instantly. Run as a null hypothesis, the EMH is the right default, and the burden sits on anyone claiming to beat the market (§24.6).
Prices move far more than fundamentals justify (excess volatility), persistent anomalies appear, and mispricing survives because arbitrage is limited: short-selling is costly, capital is constrained, and being right too early can ruin a fund before the correction arrives. The strong form is plainly false, and sentiment demonstrably moves prices.
The joint-hypothesis problem means every test of efficiency is also a test of the assumed pricing model. An anomaly is either inefficiency or a missing risk factor, and no dataset distinguishes them. This is why the 2013 Nobel went jointly to Fama and Shiller: efficiency as a discipline and the documentation of its limits are both correct, about different things.
Markets are hard to beat and not perfectly efficient, both at once, with no contradiction. The settled part: the EMH is a powerful first approximation and the strong form is false. The open part: how much efficiency survives, and whether the residual is risk premium or mispricing. The right posture is the calibrated one: name which layer is consensus and which is live, rather than averaging to "somewhat efficient."
The magnitude-and-mechanism question is genuinely unresolved and may be unresolvable from return data alone, given the joint-hypothesis problem. Separately, informational efficiency (treated here) is distinct from allocative efficiency (whether prices steer capital well); the walkthrough keeps the two apart, and the allocative question routes elsewhere.
A stop on whether economics caused 2008. The apparatus on trial (efficient markets, diversification, CAPM, no-arbitrage pricing) is the apparatus this chapter builds.
The crisis indictment names specific tools: the efficient-markets hypothesis (§24.6) that licensed treating prices as informative, diversification and correlation (§24.2) that under-counted tail co-movement, CAPM and factor models (§24.3, §24.6) used to gauge risk, and the no-arbitrage pricing of structured claims (§24.4, §24.5) whose assumptions broke when liquidity vanished. Each is built cleanly in this chapter; the walkthrough asks which, if any, deserves blame.
The chapter supplies the models at full strength, including their honestly-named failure points: the volatility smile, the correlations that snap toward one in a crisis (§24.5), the limits to arbitrage (§24.6). That lets the walkthrough argue from the actual machinery rather than a caricature of it. The crisis chronology itself belongs to the economic-history account; the apparatus belongs here.
One question of Lena's remains, and it concerns the firm rather than the market. She is weighing whether a company she might invest in should fund itself with debt or equity. The surprising benchmark answer is that, in a frictionless world, it does not matter at all, and seeing exactly why isolates what does.
A firm can raise money by issuing equity or by borrowing. It feels as though the choice should matter: debt is cheaper than equity, so surely loading up on debt lowers the cost of capital and raises the firm's value. Modigliani and Miller showed in 1958 that, in a world with no taxes, no bankruptcy costs, and symmetric information, the choice does not change the firm's value at all. The total value of the firm is set by the cash flows its assets produce; how that value is sliced between debtholders and shareholders does not change the size of the pie.
The argument is pure no-arbitrage, which is why it belongs at the end of this chapter. Suppose a levered firm were worth more than an identical unlevered one. An investor could replicate the levered firm's payoff by buying the unlevered firm and borrowing on personal account in the same proportion (this is "homemade leverage") and would pay less for the same cash-flow stream. The arbitrage would continue until the two values converged. Leverage adds nothing the investor cannot manufacture; so it is worth nothing.
Why it matters: A firm is worth whatever its assets earn. Splitting that income into a "safe" slice for lenders and a "risky" slice for shareholders does not make the income larger; it just relabels who gets what. An investor who wants more leverage can borrow in their own account; an investor who wants less can hold some bonds. So the firm cannot sell leverage at a premium, because every investor can make their own for free. The value of the result is not that financing never matters (it usually does) but that it tells you exactly where to look. If capital structure changes value, some assumption of the frictionless world has been violated, and the violation is the real story: a tax that favors debt, a bankruptcy cost that punishes it, or an information gap between insiders and the market. On the figure, raise leverage and watch firm value stay flat — then switch on a tax and watch it climb, switch on bankruptcy costs and watch an interior best point appear.
The power of the irrelevance result is that it turns every real-world capital-structure question into "name the friction." Relax the no-tax assumption: interest is tax-deductible while dividends are not, so debt creates a tax shield worth roughly the tax rate times the debt, and value rises with leverage. Relax the no-bankruptcy assumption: high leverage raises the chance of costly financial distress, an expected cost that grows with debt and eventually overwhelms the tax shield, producing an interior optimum: the trade-off theory of capital structure. Relax symmetric information: the choice to issue equity rather than debt signals that managers think the shares are overvalued, so the act of financing itself moves prices, which is the pecking-order theory. Each named imperfection is one assumption removed from the perfect-markets benchmark, and each is a separate, well-defined piece of corporate finance. The benchmark is valuable precisely because it is false in a specific, diagnosable way.
Figure 24.6. Firm value as a function of leverage. With taxes and bankruptcy costs both off, the line is flat — Modigliani-Miller irrelevance. Turn on the tax shield and value rises with leverage; add bankruptcy costs and value humps at an interior optimum. Each friction is a relevance result, switched on one at a time. Drag the sliders.
Problem. An unlevered firm is worth \$100M. (a) Under perfect markets, what is it worth if it recapitalizes to \$40M of debt? (b) Now introduce a $30\%$ corporate tax. What is the levered value, and where does the gain come from?
Solution.
(a) $V_L = V_U = \$100$M. The recapitalization shifts \$40M of value from equityholders to debtholders but leaves the total untouched — homemade leverage means no investor will pay a premium for the firm's leverage.
(b) With the tax shield, $V_L = V_U + \tau D = 100 + 0.30(40) = \$112$M. The \$12M gain is exactly the present value of the tax deductions on the interest — value created not by leverage itself but by the government's subsidy of debt finance. Add expected bankruptcy costs and the firm would stop short of all-debt financing at the leverage where the marginal shield equals the marginal expected distress cost.
The capital-structure irrelevance result is Franco Modigliani and Merton Miller's, from 1958, another keystone of the postwar formalization of economics traced in the history-of-economic-thought book's forthcoming chapter on the postwar synthesis. The questions of whether firms should return cash to shareholders through buybacks, and whom the firm is run for, are taken up in the platform's walkthroughs on corporate purpose rather than expanded here.
Value a future claim (§24.1). Lena began with a settlement and a contract promising payments over several years. She discounted each payment at the interest rate and added them, converting "money later" into "money now," and learned that her impatience did not change which investments were worth taking, only how she would borrow or lend around them.
Hold one or many (§24.2)? Facing risky claims, her instinct was to pick the best stock. Markowitz redirected her: the risk that matters is the portfolio's, and combining imperfectly correlated assets cut her risk for free. With a safe asset added, she found she should hold the one best risky bundle and set her boldness by how much of it she held.
What return must a stock offer (§24.3)? For a single stock, only its market-linked risk, its beta, earned a premium; the firm-specific wobble she could diversify away earned nothing. What is any claim worth (§24.4)? Beneath CAPM she found the deeper rule: one price-of-risk operator values everything, and CAPM is just one reading of it.
What is an option worth (§24.5)? She could price an option by building a portfolio that tracks it, needing its volatility, not a forecast of its return. Can she beat the market (§24.6)? Mostly not: prices are hard to beat, the strong-form claim is false, and whether the leftover patterns are risk or mispricing is genuinely open. Does financing matter (§24.7)? For the firm she might invest in, not in a frictionless world, and naming the friction that makes it matter is the whole of the subject from there.
| Label | Equation | Description |
|---|---|---|
| Eq. 24.1 | $PV = \sum_t C_t/(1+r)^t$ | Present value of a cash-flow stream |
| Eq. 24.2 | $\text{MRS} = 1+r$ | Fisher two-period optimum |
| Eq. 24.3 | $E[R_p] = \sum_i w_i E[R_i]$ | Portfolio expected return |
| Eq. 24.4 | $\sigma_p^2 = \sum_i\sum_j w_i w_j \sigma_{ij}$ | Portfolio variance (covariance core) |
| Eq. 24.5 | $\sigma_p^2 = w^2\sigma_1^2 + (1{-}w)^2\sigma_2^2 + 2w(1{-}w)\rho\sigma_1\sigma_2$ | Two-asset frontier with correlation |
| Eq. 24.6 | $E[R_i] = R_f + \beta_i(E[R_m]-R_f)$ | CAPM / security market line |
| Eq. 24.7 | $\beta_i = \mathrm{Cov}(R_i,R_m)/\mathrm{Var}(R_m)$ | Beta (systematic risk) |
| Eq. 24.8 | $p = E[m\cdot x]$ | Fundamental pricing equation (SDF) |
| Eq. 24.9 | $E[R_i]-R_f = -\mathrm{Cov}(m,R_i)/E[m]$ | SDF risk-return decomposition |
| Eq. 24.10 | $p = \frac{1}{1+R_f}E^Q[x]$ | Risk-neutral pricing |
| Eq. 24.11 | $C = S\,N(d_1) - Ke^{-rT}N(d_2)$ | Black-Scholes call ($\mu$ absent) |
| Eq. 24.12 | $E[R_{t+1}-E(R_{t+1}\mid I_t)\mid I_t]=0$ | EMH unpredictability condition |
| Eq. 24.13 | $E[R_i]-R_f = \beta_M\mathrm{MKT}+\beta_S\mathrm{SMB}+\beta_V\mathrm{HML}$ | Fama-French three-factor model |
| Eq. 24.14 | $V_L = V_U$ | Modigliani-Miller Proposition I |
Fisher (1907, 1930); Markowitz (1952); Sharpe (1964); Lintner (1965); Mossin (1966); Black & Scholes (1973); Merton (1973); Cochrane (2005), Asset Pricing; Fama (1970); Fama & French (1993); Shiller (1981); Shleifer & Vishny (1997); Modigliani & Miller (1958).