Chapter 4 · Portfolio Mathematics

4.2 CAPM, Factors, Asset Allocation and Rebalancing

Add a risk-free asset and one risky portfolio becomes best for everyone: the capital market line, the tangency portfolio, CAPM and the security market line, what size, value and momentum factors are, and how to choose, maintain and rebalance an allocation across equity, debt and gold.

📍 The situation

Arjun, 35, an engineer in Pune, has ₹20,00,000 saved and a monthly SIP. He has narrowed his choices to four building blocks: an equity index fund, a debt fund, a gold fund, and a liquid fund or treasury bills that pay a modest return with almost no risk over a short horizon. Section 4.1 showed him how to compute the risk of any mix and draw the efficient frontier of the three risky funds. But he still has questions. If a nearly risk-free asset is available, does the frontier change? Is there one "best" mix of the risky funds, and how much of his money should go into it? His colleague says a stock with a beta of 1.3 "should earn 14%" — where does such a number come from? A fund advertises a "value" and "momentum" strategy — is that a new law of returns or a marketing label?

And then the practical part: once Arjun picks a mix, say 60% equity, 30% debt and 10% gold, markets will push it away from that mix every month. When should he rebalance, how much should he trade, and does rebalancing earn or cost him money? Is "hold 100 minus your age in equity" a good rule? This section answers these questions with the capital market line, the Capital Asset Pricing Model (CAPM), a descriptive look at factors, and the arithmetic of asset allocation and rebalancing.

⚠️ About the numbers on this page

Risk-free rates, expected returns, volatilities, correlations, factor premiums, betas and yearly returns on this page are illustrative assumptions or simulated numbers, not historical data or forecasts. Tax on gains realised when rebalancing is shown with an illustrative rate only; tax rates, holding periods and exemptions change — check the Income Tax Department's current rules. Funds, companies and people are hypothetical, and nothing here is investment advice.

Adding a Risk-Free Asset: the Capital Allocation Line

A treasury bill held to maturity pays a known rupee amount on a known date; over that horizon its return has (almost) no variance and no covariance with anything else. Mixing such an asset with a risky portfolio turns the curved frontier of §4.1 into straight lines.

📘 Definition 4.5 — Risk-free asset and capital allocation line

A risk-free asset has a certain return $r_f$ over the investment period (variance 0). Investing a fraction $y$ of wealth in a risky portfolio $P$ (expected return $\mu_P$, volatility $\sigma_P$) and $1-y$ in the risk-free asset gives a complete portfolio $C$. As $y$ varies, the points $(\sigma_C,\mu_C)$ form the capital allocation line (CAL) of $P$. For $0\le y\le1$ the investor lends at $r_f$; for $y\gt1$ the investor borrows to invest more than 100% of wealth in $P$ (leverage).

In practice a T-bill or a liquid fund is only approximately risk-free: it is free of default and price risk over a short horizon, not of inflation or reinvestment risk.

📐 Theorem 4.8 (Formula) — The capital allocation line

For $y\ge0$, $$\mu_C=r_f+y(\mu_P-r_f),\qquad\sigma_C=y\,\sigma_P,$$ so the CAL is the straight line $$\mu_C=r_f+\frac{\mu_P-r_f}{\sigma_P}\,\sigma_C.$$ Its slope is the Sharpe ratio $S_P=(\mu_P-r_f)/\sigma_P$ of the risky portfolio (§3.2): the extra expected return per unit of volatility. Every point on the line has the same Sharpe ratio.

Derivation. $r_C=yr_P+(1-y)r_f$, so $\mu_C=y\mu_P+(1-y)r_f=r_f+y(\mu_P-r_f)$. Since $r_f$ is a constant, $\operatorname{Var}(r_C)=y^2\sigma_P^2$ (Theorem 4.1 with $\sigma_2=0$), so $\sigma_C=y\sigma_P$. Eliminate $y=\sigma_C/\sigma_P$. ∎

Every risky portfolio has its own CAL. An investor who can combine any risky portfolio with the risk-free asset wants the steepest line — the highest expected return for each level of risk. Rotating the line upwards about $(0,r_f)$ until it just touches the frontier picks out one portfolio.

📐 Theorem 4.9 — The tangency portfolio

Let $\Sigma$ be positive definite and $r_f\lt B/A$ (the GMV return of §4.1). The risky portfolio with the largest Sharpe ratio — the tangency portfolio $T$ — is $$w_T=\frac{\Sigma^{-1}(\mu-r_f\mathbf 1)}{\mathbf 1^{\mathsf T}\Sigma^{-1}(\mu-r_f\mathbf 1)}.$$ Its CAL is tangent to the efficient frontier at $T$. For two risky assets with excess returns $e_i=\mu_i-r_f$: $$w_1=\frac{e_1\sigma_2^2-e_2\sigma_{12}}{e_1\sigma_2^2+e_2\sigma_1^2-(e_1+e_2)\sigma_{12}}.$$

Derivation. Let $x$ be the vector of rupee fractions in the risky assets, with the rest, $1-\mathbf 1^{\mathsf T}x$, in the risk-free asset. The excess return is $x^{\mathsf T}(\mu-r_f\mathbf 1)$ and the variance $x^{\mathsf T}\Sigma x$. Minimising $\tfrac12x^{\mathsf T}\Sigma x$ for a given excess return $e$ with a multiplier $\gamma$ gives $\Sigma x=\gamma(\mu-r_f\mathbf 1)$, so $x=\gamma\,\Sigma^{-1}(\mu-r_f\mathbf 1)$: every efficient choice holds the risky assets in the same proportions, only scaled. Normalising those proportions to add up to 1 gives $w_T$ (the denominator is positive exactly when $r_f\lt B/A$). Since all efficient portfolios lie on the CAL of $T$, it is the steepest one. ∎

✏️ Example 1 — Equity and gold with a 6% risk-free rate
Take only the equity fund ($\mu=12\%$, $\sigma=20\%$) and the gold fund ($\mu=9\%$, $\sigma=15\%$), uncorrelated, with $r_f=6\%$ (illustrative). Find the tangency weights, its expected return, volatility and Sharpe ratio, and compare with each fund alone.
Inputs $e_E=6\%$, $e_G=3\%$, $\sigma_E^2=0.04$, $\sigma_G^2=0.0225$, $\sigma_{EG}=0$.
Weights $w_E=\dfrac{0.06\times0.0225}{0.06\times0.0225+0.03\times0.04}=\dfrac{0.00135}{0.00255}\approx52.94\%$, $w_G\approx47.06\%$.
T $\mu_T\approx0.5294(12)+0.4706(9)\approx10.59\%$; $\sigma_T^2=0.5294^2(0.04)+0.4706^2(0.0225)\approx0.016194$, $\sigma_T\approx12.73\%$; $S_T=(10.59-6)/12.73\approx0.361$.
Compare Equity alone: $S=6/20=0.30$; gold alone: $3/15=0.20$. The mix beats both: diversification raises the reward per unit of risk, and any risk level Arjun wants is best reached along the line through $T$.
✏️ Example 2 — The tangency portfolio of equity, debt and gold
With the three-asset inputs of §4.1 ($\mu=(12\%,7\%,9\%)$, $\sigma=(20\%,5\%,15\%)$, $\rho_{ED}=0.2$, $\rho_{EG}=0$, $\rho_{DG}=0.1$) and $r_f=6\%$, a spreadsheet gives $\Sigma^{-1}(\mu-r_f\mathbf 1)\approx(1.3737,\ 2.5263,\ 1.2491)^{\mathsf T}$. Find $w_T$, $\mu_T$, $\sigma_T$ and $S_T$, and compare the Sharpe ratio with the 50/30/20 mix.
Weights The entries add up to $5.1491$; dividing, $w_T\approx(26.68\%,\ 49.06\%,\ 24.26\%)$ — all positive.
Return and risk $\mu_T=w_T^{\mathsf T}\mu\approx8.82\%$, $\sigma_T=\sqrt{w_T^{\mathsf T}\Sigma w_T}\approx7.40\%$ ($\sigma_T^2\approx0.005475$).
Sharpe $S_T=(8.82-6)/7.40\approx0.381$. The 50/30/20 mix: $(9.9-6)/10.87\approx0.359$ — good, but not the steepest line.
Sensitivity Repeat with $r_f=6.5\%$ and the tangency becomes about $(44.4\%,\ 19.7\%,\ 36.0\%)$: a half-point change in one input moves about 30 points of weight. Optimal weights are fragile (§4.3).
📐 Theorem 4.10 — Two-fund separation and how much to put in $T$
  1. (Separation) With a risk-free asset, every mean–variance efficient portfolio is a mix of the risk-free asset and the same tangency portfolio $T$; investors differ only in the fraction $y$.
  2. For a target volatility $\sigma^\ast$, $y=\sigma^\ast/\sigma_T$. For an investor who maximises $\mu_C-\tfrac{\lambda}{2}\sigma_C^2$ (risk aversion $\lambda\gt0$), $$y^\ast=\frac{\mu_T-r_f}{\lambda\,\sigma_T^2}.$$

Why. (1) is the derivation of Theorem 4.9: the efficient risky holdings are always proportional to $\Sigma^{-1}(\mu-r_f\mathbf 1)$. (2) The first part is Theorem 4.8. For the second, $f(y)=r_f+y(\mu_T-r_f)-\tfrac\lambda2y^2\sigma_T^2$ is a downward parabola; $f'(y)=(\mu_T-r_f)-\lambda y\sigma_T^2=0$ gives $y^\ast$. ∎ The decision splits into a technical part (find $T$ — the same for all) and a personal part (choose $y$ — your risk tolerance).

✏️ Example 3 — Arjun's two choices: cautious and growth
Using $T$ from Example 2 ($\mu_T=8.82\%$, $\sigma_T=7.40\%$, $r_f=6\%$): (a) A cautious plan targets 4% volatility. Find $y$ and the expected return. (b) A growth plan targets 10% volatility. Find $y$ and the expected return if borrowing were possible at $r_f$, and if it costs an illustrative 9%. (c) Without borrowing, the frontier portfolio with 10% volatility has expected return 9.73% (weights about 41/25/34). Compare. (d) What does $y^\ast$ equal for $\lambda=5$ and $\lambda=10$?
(a) $y=4/7.40\approx0.541$: 54.1% in $T$, 45.9% risk-free; $\mu_C=6\%+0.541\times2.82\%\approx7.52\%$. No portfolio of the risky funds alone has volatility as low as 4% (their GMV has 4.86%, §4.1).
(b) $y=10/7.40\approx1.351$: borrow 35.1% of wealth. At $r_f$: $\mu_C\approx6\%+1.351\times2.82\%\approx9.81\%$. At a 9% borrowing cost: $\mu_C=1.351\times8.82\%-0.351\times9\%\approx8.76\%$.
(c) The unlevered frontier portfolio gives 9.73% at the same 10% volatility — better than borrowing at 9% (8.76%) and only 0.08 points below the ideal at $r_f$. Real investors who cannot borrow cheaply simply move up the frontier past $T$.
(d) $y^\ast=0.0282/(5\times0.005475)\approx1.03$ and $0.0282/(10\times0.005475)\approx0.51$. Doubling risk aversion halves the risky share.
Interactive 1 — The Capital Market Line and the Security Market Line

Illustrative equity, debt and gold funds of §4.1 plus a risk-free asset. In the σ–return view, the green curve is the efficient frontier and the blue line the capital market line from $r_f$ through the tangency portfolio T (dashed beyond $T$: borrowing). The amber dot is your mix of $T$ and the risk-free asset. Switch to the β–return view: with $T$ playing the market, every asset sits exactly on the security market line. Move $r_f$ and watch $T$ swing. Play sweeps your fraction in $T$.

6.0%54%
tangency weights E / D / G—
T: return, volatility—
Sharpe ratio of T (CML slope)—
your mix: return, volatility—
betas E / D / G (vs T)—

CAPM and the Security Market Line

Theorem 4.10 says every mean–variance investor holds the same risky portfolio $T$. If, in addition, all investors agree on $\mu$ and $\Sigma$, then adding up everyone's holdings, the total demand for risky assets is in the proportions of $T$ — and the total holdings of all investors are the market. So in equilibrium $T$ must be the market portfolio. This one idea gives a formula for the expected return of every asset.

📘 Definition 4.6 — Market portfolio, capital market line and beta

The market portfolio $M$ holds every risky asset in proportion to its market value (a broad free-float index such as the Nifty 50 is a practical stand-in, §2.1). The CAL of $M$ is the capital market line (CML): $\mu=r_f+\dfrac{\mu_M-r_f}{\sigma_M}\sigma$. The beta of asset $i$ is $\beta_i=\dfrac{\operatorname{Cov}(r_i,r_M)}{\sigma_M^2}$ (estimated by the market-model regression of §3.2), and $\mu_M-r_f$ is the market risk premium.

📐 Theorem 4.11 — The Capital Asset Pricing Model (CAPM)

Suppose investors are mean–variance optimisers with a common one-period horizon and identical estimates of $\mu$ and $\Sigma$, can lend and borrow at $r_f$, and trade without taxes or costs, so that the market portfolio is the tangency portfolio. Then for every asset (and every portfolio) $i$, $$\mu_i=r_f+\beta_i\,(\mu_M-r_f).$$ More generally, for any set of assets and the tangency portfolio $T$ of Theorem 4.9, $\mu_i-r_f=\beta_{i,T}(\mu_T-r_f)$ with $\beta_{i,T}=\operatorname{Cov}(r_i,r_T)/\sigma_T^2$ — this part is pure algebra.

Derivation. From Theorem 4.9, $\Sigma w_T=k(\mu-r_f\mathbf 1)$ for some $k\gt0$. The $i$-th entry of $\Sigma w_T$ is $\operatorname{Cov}(r_i,r_T)$, so $\operatorname{Cov}(r_i,r_T)=k(\mu_i-r_f)$ for every $i$. Multiply by $w_{T,i}$ and add: $\sigma_T^2=k(\mu_T-r_f)$, so $k=\sigma_T^2/(\mu_T-r_f)$. Substituting, $\mu_i-r_f=\dfrac{\operatorname{Cov}(r_i,r_T)}{\sigma_T^2}(\mu_T-r_f)$. Under the equilibrium assumptions $T=M$. ∎ Check with §4.1's numbers ($r_f=6\%$): $\Sigma w_T\approx(0.011652,\ 0.001942,\ 0.005826)$, so the betas against $T$ are $(2.128,\ 0.355,\ 1.064)$, and $6\%+2.128\times2.82\%=12\%$, $6\%+0.355\times2.82\%=7\%$, $6\%+1.064\times2.82\%=9\%$ — exactly the assumed expected returns.

📘 Definition 4.7 — Security market line and alpha

The security market line (SML) is the line $\mu=r_f+\beta(\mu_M-r_f)$ in the $(\beta,\mu)$ plane; under CAPM every asset lies on it. If an asset's expected return (from a forecast) is $\hat\mu_i$, its alpha relative to CAPM is $$\alpha_i=\hat\mu_i-\bigl[r_f+\beta_i(\mu_M-r_f)\bigr].$$ A positive alpha (above the SML) means a higher expected return than its systematic risk requires; a negative alpha, lower. The CAPM required return is also used as the cost of equity in valuation (§5.1).

📐 Theorem 4.12 (Formula) — Portfolio beta

A portfolio with weights $w_i$ has $\beta_p=\sum_iw_i\beta_i$ (the risk-free asset has $\beta=0$), and under CAPM $\mu_p=r_f+\beta_p(\mu_M-r_f)$. Only systematic risk, $\beta_p^2\sigma_M^2$, is priced: two portfolios with the same beta have the same CAPM expected return however much specific risk they carry.

Why. Covariance is linear in its first argument, so $\operatorname{Cov}(\sum w_ir_i,r_M)=\sum w_i\operatorname{Cov}(r_i,r_M)$; divide by $\sigma_M^2$. Multiply the CAPM equations by $w_i$ and add. ∎ This is the theoretical reason for §4.1's conclusion: unsystematic risk can be diversified away at no cost, so the market does not pay for bearing it.

✏️ Example 4 — Required returns and a portfolio beta
Take an illustrative $r_f=6.5\%$ and expected market return $12.5\%$. (a) Find the CAPM required return of Alpha Bank ($\beta=1.3$) and of a consumer-goods company ($\beta=0.7$). (b) Arjun's ₹10 lakh stock portfolio is 40% in a stock with $\beta=1.3$, 35% with $\beta=0.8$ and 25% with $\beta=1.0$. Find $\beta_p$ and the required return, and the typical portfolio move if the market falls 10%.
Premium $\mu_M-r_f=12.5\%-6.5\%=6\%$.
(a) Alpha Bank: $6.5\%+1.3\times6\%=14.3\%$. Consumer goods: $6.5\%+0.7\times6\%=10.7\%$. The colleague's "14%" is a CAPM required return, not a forecast.
(b) $\beta_p=0.4(1.3)+0.35(0.8)+0.25(1.0)=1.05$; required return $6.5\%+1.05\times6\%=12.8\%$.
Market fall On average the portfolio moves about $1.05\times(-10\%)=-10.5\%$ (≈ −₹1,05,000), plus or minus its specific risk. This is also the beta that sizes an index-futures hedge (§6.1).
✏️ Example 5 — Above or below the SML?
An analyst forecasts one-year expected returns for three hypothetical stocks: Gamma Steel ($\beta=1.4$, forecast 15%), Delta FMCG ($\beta=0.6$, forecast 11%), Epsilon Tech ($\beta=1.1$, forecast 11%). With $r_f=6.5\%$ and a market premium of 6%, find each alpha and interpret.
Gamma Required $6.5+1.4\times6=14.9\%$; $\alpha=+0.1\%$ — essentially on the SML.
Delta Required $6.5+0.6\times6=10.1\%$; $\alpha=+0.9\%$ — above the SML.
Epsilon Required $6.5+1.1\times6=13.1\%$; $\alpha=-2.1\%$ — below the SML: the forecast return is low for its beta.
Interpret carefully Alpha here is only as good as the forecast, the beta estimate and the premium assumed. A 1-point error in the premium moves Gamma's required return by 1.4 points. Under CAPM, a persistent positive alpha would attract buyers, raise the price and lower the future expected return until $\alpha\approx0$.

Factor Ideas: Size, Value and Momentum

Tests of CAPM on many stock markets have found that beta alone does not explain average returns well: groups of stocks sorted by company size, by price relative to accounting value, or by recent performance have shown average returns different from their CAPM predictions over long periods. Researchers responded with multi-factor models. We describe the ideas; we make no claim about what any factor will earn.

📘 Definition 4.8 — Factor model and the size, value and momentum factors

A factor model explains a stock's or fund's excess return by several factor returns: $$r_i-r_f=\alpha_i+\beta_i\,\text{MKT}+s_i\,\text{SMB}+h_i\,\text{HML}+m_i\,\text{WML}+\varepsilon_i.$$

  • MKT — the market excess return $r_M-r_f$ (as in CAPM).
  • Size (SMB, "small minus big") — the return of a portfolio long small-cap and short large-cap stocks.
  • Value (HML, "high minus low") — long stocks with a high book-to-price ratio ("cheap") and short those with a low one ("expensive").
  • Momentum (WML, "winners minus losers") — long stocks with the highest returns over roughly the past year (often skipping the latest month) and short the lowest.

The coefficients $s_i,h_i,m_i$ are factor exposures (loadings), estimated by multiple regression; $\alpha_i$ is the return left unexplained. Index providers, including NSE, publish rule-based "factor" or "smart-beta" indices built on such ideas.

✏️ Example 6 — Is that alpha skill or a factor tilt?
A hypothetical small-and-mid-cap fund has a CAPM alpha of 1.9% a year. A three-factor regression gives (with the same market beta) $s=0.4$, $h=0.3$ and $\alpha=0.2\%$. Over the sample, the SMB and HML factors averaged (hypothetically) 2% and 3% a year. (a) How much of the CAPM alpha is explained by the factor exposures? (b) A second fund has $\beta=1.0$, $s=0.5$, $h=-0.2$, $m=0.3$. With hypothetical premiums MKT 6%, SMB 2%, HML 3%, WML 4% and zero alpha, what excess return does the model predict?
(a) The tilts contribute $0.4\times2\%+0.3\times3\%=1.7\%$; with the remaining $\alpha=0.2\%$ this accounts for the CAPM alpha of 1.9%.
Meaning Most of the "outperformance" came from holding smaller and cheaper stocks — exposures an investor could get from rule-based factor funds — not from unexplained stock-picking skill.
(b) $1.0(6)+0.5(2)-0.2(3)+0.3(4)=6+1-0.6+1.2=7.6\%$ over $r_f$.
Caveat Factor premiums are averages over long samples. Each factor has had long stretches of negative returns; a premium measured in the past is not a promise, and the more rules researchers try, the more likely some "work" by chance in the data (§7.3).

Strategic Asset Allocation and Rebalancing

For a long-term investor the biggest decision is not which stock or fund to buy but how to split money between asset classes. The theory above says "hold $T$ and choose $y$"; in practice, with uncertain inputs, investors fix a simple long-term mix that matches their goals and risk tolerance, and then maintain it.

📘 Definition 4.9 — Strategic allocation, rebalancing and glide paths

A strategic asset allocation (SAA) is a long-term target mix of asset classes, e.g. 60% equity, 30% debt, 10% gold. Tactical allocation means temporary deliberate deviations from it. Market moves make the actual weights drift; rebalancing trades back to the targets, either on a calendar (e.g. once a year) or when a weight leaves a tolerance band (e.g. ±5 percentage points). A glide path is a planned change of the target over time, usually reducing equity as a goal approaches (the idea behind age-based rules and target-date funds).

📐 Theorem 4.13 (Formula) — Weight drift and the rebalancing trade

A portfolio of value $V$ with weights $w_i$ earns returns $R_i$ over a period. Its new value is $V'=V(1+R_p)$ with $R_p=\sum w_iR_i$, and the new weights are $$w_i'=\frac{w_i(1+R_i)}{1+R_p}.$$ To restore the targets, buy $w_iV'-w_i(1+R_i)V$ of asset $i$ (sell if negative); the trades add up to zero. An asset's weight rises exactly when it beat the portfolio ($R_i\gt R_p$), so rebalancing always sells some of what has risen most and buys what has lagged.

Derivation. Asset $i$ grows from $w_iV$ to $w_iV(1+R_i)$, and the total is $\sum w_iV(1+R_i)=V(1+R_p)$; divide. The target holding is $w_iV'$. The trades sum to $V'\sum w_i-V'=0$. Finally $w_i'\gt w_i$ iff $1+R_i\gt1+R_p$. ∎

✏️ Example 7 — Two years of a 60/40 portfolio
₹10,00,000 is invested 60% equity, 40% debt. Year 1: equity +30%, debt +6%. Year 2: equity −25%, debt +7% (hypothetical). (a) Find the weights after year 1 and the rebalancing trade. (b) After year 2, compare the annually rebalanced portfolio with buy-and-hold. (c) What if year 2 had instead been equity +30%, debt +6% again?
(a) Equity ₹7,80,000, debt ₹4,24,000, total ₹12,04,000; equity weight $7.80/12.04\approx64.78\%$. Target equity $0.6\times12.04=$ ₹7,22,400: sell ₹57,600 of equity and buy the same amount of debt.
(b) Rebalanced From (₹7,22,400, ₹4,81,600): equity → ₹5,41,800, debt → ₹5,15,312, total ₹10,57,112 (equity weight 51.25%; next trade: buy ₹92,467 of equity).
(b) Buy-and-hold From (₹7,80,000, ₹4,24,000): equity → ₹5,85,000, debt → ₹4,53,680, total ₹10,38,680. Rebalancing ended ₹18,432 ahead — it had sold equity before the fall.
(c) A trend With a second +30% year: buy-and-hold ₹14,63,440, rebalanced ₹14,49,616 — now buy-and-hold is ₹13,824 ahead. Rebalancing is not a return machine; its job is to keep the risk you chose. It helps when markets reverse and costs a little when they trend.
📐 Theorem 4.14 — Rebalancing and the diversification return

For returns that are not too large, an asset with expected return $\mu_i$ and volatility $\sigma_i$ has long-run growth rate $g_i\approx\mu_i-\tfrac12\sigma_i^2$ (volatility drag, §3.1). A long-only portfolio rebalanced to fixed weights $w$ grows at $$g_p\approx\mu_p-\tfrac12\sigma_p^2=\sum_iw_ig_i+\tfrac12\Bigl(\sum_iw_i\sigma_i^2-\sigma_p^2\Bigr),$$ and the last term, the diversification return ("rebalancing bonus"), is $\ge0$: a rebalanced portfolio grows faster than the weighted average of its assets' growth rates.

Why. $\sum w_ig_i\approx\sum w_i\mu_i-\tfrac12\sum w_i\sigma_i^2=\mu_p-\tfrac12\sum w_i\sigma_i^2$; subtract from $g_p$. For non-negativity, $\sigma_p\le\sum w_i\sigma_i$ (risk is at most the weighted average, Theorem 4.2 extended) and $(\sum w_i\sigma_i)^2\le\sum w_i\sigma_i^2$ (the square of an average is at most the average of the squares). ∎ Caution: the comparison is with the average growth rate, not with buy-and-hold in every market; costs and taxes come off the bonus.

✏️ Example 8 — The seesaw market and a 60/40 bonus
(a) A stock alternately rises 50% and falls one-third each year. Compare 10 years of buy-and-hold in the stock with a 50/50 stock-and-cash portfolio (cash earning 0%) rebalanced every year. (b) Estimate the diversification return of an annually rebalanced 60/40 equity–debt portfolio with $\sigma_E=20\%$, $\sigma_D=5\%$, $\rho=0.2$.
(a) Stock $1.5\times\tfrac23=1$: every two years the stock is back where it started. After 10 years: ₹1 → ₹1. Buy-and-hold 50/50 also ends at ₹1.
(a) Rebalanced Each up-year the portfolio gains $0.5\times50\%=25\%$, each down-year it loses $0.5\times33.3\%=16.7\%$: $1.25\times\tfrac56\approx1.0417$ every two years, so after 10 years $1.0417^5\approx1.226$ — about 2.1% a year from nothing but rebalancing between a volatile asset and cash.
(b) $\sigma_p^2=0.36(0.04)+0.16(0.0025)+2(0.6)(0.4)(0.002)=0.01576$; $\sum w_i\sigma_i^2=0.6(0.04)+0.4(0.0025)=0.025$. Bonus $\approx\tfrac12(0.025-0.01576)\approx0.46\%$ a year.
Perspective Real markets are not seesaws, and the bonus is modest and uncertain; transaction costs and taxes on realised gains can eat it. Its main value is that it comes with keeping risk at the chosen level.

Age-based rules. A popular rule says: hold "100 minus your age" percent in equity (some versions use 110 or 120). At 25 that gives 75% (or 85%), at 35 it gives 65% (75%), at 55 it gives 45% (55%). The rule captures a real idea — a young saver has decades of future salary (human capital, which behaves somewhat like a bond) and time to recover from bad years, so can usually bear more market risk; near a goal, a large fall cannot be waited out. But it ignores almost everything that matters: the goal and its date (a 30-year-old saving for a house in 3 years should not hold 70% equity for that money), income stability (a business owner's income may already be tied to the economy), existing assets such as a house or a provident fund, and the investor's ability to stay invested through a fall. A glide path tied to each goal's date, not to age alone, is the more careful version of the same idea (§8.1).

Interactive 2 — Rebalancing Simulator: Buy-and-Hold, Annual and Band

₹10 lakh in an equity–debt mix, over a simulated market (monthly lognormal returns, illustrative equity 12% ± 20% and debt 7% ± 5% a year, correlation 0.2 — not historical data). Top panel: portfolio value; bottom panel: equity weight. Red: buy-and-hold; blue: rebalanced every year; green: rebalanced only when the equity weight leaves the band. Try several market paths — sometimes buy-and-hold ends ahead, but its equity weight wanders far from the target. Play runs time forward.

#1
buy-and-hold: value, max drawdown—
annual: value, max drawdown—
band: value, trades—
equity weight range, buy-and-hold—
📈 Market case — Arjun's 60/30/10 plan through three hypothetical years

Arjun compares rules of thumb — 100 − 35 = 65% equity, 110 − 35 = 75% — with his own goals (retirement in 25 years, a child's education in 15) and settles on a written plan: 60% equity, 30% debt, 10% gold, rebalanced once a year. He invests ₹20,00,000. The three years of returns below are hypothetical.

yearequitydebtgoldweights before rebalancingrebalancing trade (₹)value (₹)
1+28%+6.5%−4%64.9 / 27.0 / 8.1sell E 1,15,800; buy D 71,100, G 44,70023,67,000
2−18%+7.5%+22%52.5 / 34.4 / 13.0buy E 1,65,453; sell D 98,349, G 67,10522,16,696
3+12%+6%+8%61.2 / 29.0 / 9.8small: sell E 29,260; buy D 25,270, G 3,99024,33,932
  1. The rule in action. After the strong equity year he sells equity; after the fall he buys it — mechanically, without forecasting (Theorem 4.13). In year 3 the weights are within 1.2 points of target; a ±5-point band rule would not trade at all.
  2. Against buy-and-hold. Doing nothing would have ended at ₹23,91,782 with weights 59.0 / 30.4 / 10.6. The rebalanced plan ends ₹42,150 higher on this path, mainly because it bought equity after year 2's fall. On a trending path the result could reverse (Example 7).
  3. Costs and tax. Selling ₹1,15,800 of equity in year 1 realises a gain: the equity holding grew from ₹12,00,000 to ₹15,36,000, so $21.875\%$ of the amount sold, ₹25,331, is gain. With an illustrative tax rate $t$, the tax is $t\times$ ₹25,331 (₹3,166 for $t=12.5\%$) — before any exemption, and depending on the holding period. Check the current rules.
  4. Rebalancing with new money. Arjun's SIP can do much of the work without selling: in year 2 he can direct new contributions to equity until it is back near 60%. This avoids realising gains and saves transaction costs.

The mathematics, not a recommendation. The arithmetic shows what a rebalancing rule does — keeps risk near the chosen level and enforces "sell high, buy low" — and what it costs. Whether 60/30/10 suits Arjun depends on his goals and his capacity to bear losses, which no formula decides for him.

⚠️ Common mistakes and myths
  1. "CAPM says this stock will return 14.3%." CAPM gives a required or equilibrium expected return under strong assumptions, using estimated inputs — not a forecast of next year's return (Example 4).
  2. "A high-volatility stock must have a high expected return." Under CAPM only beta (systematic risk) is rewarded. A volatile stock with a low beta has a low required return; its specific risk can be diversified away for free.
  3. Using the CML for a single stock. Individual assets lie inside the frontier, below the CML; the relation that holds for every asset is the SML, in terms of beta.
  4. "Positive alpha means a great fund." Alpha depends on the benchmark model. A CAPM alpha can be a size or value tilt in disguise (Example 6), and estimated alphas are noisy.
  5. "Factor premiums are guaranteed extra returns." They are long-run historical averages with long losing streaks, and may shrink once widely known.
  6. "Rebalancing always adds return." It controls risk. It helps in choppy markets and lags in strong trends, and taxes and costs can outweigh the diversification return (Example 7, Theorem 4.14).
  7. "Rebalance whenever weights move." Trading every small drift multiplies costs; calendar or band rules, and redirecting new money, are cheaper.
  8. "100 minus age is the right allocation." It is a rough starting point that ignores goals, income, other assets and behaviour.
  9. "Borrowing to reach the CML is free." The theory assumes borrowing at $r_f$; real borrowing costs more, and leverage magnifies losses (Example 3).

Exercises

All rates and returns are illustrative or hypothetical, per year. Round percentages to two decimals and rupee amounts to the nearest rupee.

I
The capital allocation line and the tangency portfolio. Risk-free rate 6%; a risky fund P has $\mu_P=11\%$, $\sigma_P=16\%$.
  • 1 Write the equation of P's CAL and find its Sharpe ratio. Find the expected return and volatility of 60% in P, 40% risk-free.
  • 2 What fraction in P gives a volatility of 12%? What is its expected return?
  • 3 An investor wants an expected return of 13%. Find $y$ and the volatility. If borrowing costs 9%, what expected return does the same position actually give?
  • 4 Two risky funds: A ($\mu=14\%$, $\sigma=25\%$) and B ($\mu=8\%$, $\sigma=10\%$), $\rho=0.3$. Find the tangency weights, $\mu_T$, $\sigma_T$ and $S_T$ for $r_f=6\%$.
  • 5 Compare $S_T$ in part 4 with the Sharpe ratios of A and B alone.
  • 6 Explain in two sentences why, in the theory, a cautious and an aggressive investor hold the risky funds in the same proportions.
1 $\mu_C=6\%+\frac{5}{16}\sigma_C$; Sharpe $=5/16=0.3125$. With $y=0.6$: $\mu=6+0.6(5)=9\%$, $\sigma=0.6(16)=9.6\%$.
2 $y=12/16=0.75$; $\mu=6+0.75(5)=9.75\%$.
3 $y=(13-6)/5=1.4$ (borrow 40% of wealth); $\sigma=1.4\times16=22.4\%$. At a 9% borrowing cost: $1.4(11\%)-0.4(9\%)=11.8\%$, not 13%.
4 $e_A=8\%$, $e_B=2\%$, $\sigma_{AB}=0.0075$: $w_A=\dfrac{0.08(0.01)-0.02(0.0075)}{0.08(0.01)+0.02(0.0625)-0.10(0.0075)}=\dfrac{0.00065}{0.0013}=50\%$. Then $\mu_T=11\%$, $\sigma_T\approx14.79\%$, $S_T\approx0.338$.
5 A: $8/25=0.32$; B: $2/10=0.20$; $T$: 0.338 — higher than both.
6 With a risk-free asset, the best risk–return trade-off is the steepest line, through the same tangency portfolio for everyone (Theorem 4.10). Risk tolerance only decides how far along that line to go — how much to put in $T$ versus the risk-free asset.
II
CAPM and the SML. Use $r_f=6.5\%$ and an expected market return of 12% unless stated.
  • 1 Find the CAPM required return for $\beta=0,\ 0.5,\ 1,\ 1.5$.
  • 2 A stock has $\beta=1.2$ and a forecast return of 13%. Find its alpha; is it above or below the SML?
  • 3 A stock with $\beta=0.9$ has a CAPM required return of 11.9% when $r_f=6.5\%$. What market risk premium is being assumed?
  • 4 A portfolio is 50% in a stock with $\beta=1.4$, 30% in one with $\beta=0.9$ and 20% in T-bills. Find its beta and required return.
  • 5 $\operatorname{Cov}(r_i,r_M)=0.0288$ and $\sigma_M=16\%$. Find $\beta_i$.
  • 6 An asset has $\beta=-0.3$. Find its CAPM required return and explain why it can be below $r_f$.
  • 7 In one or two sentences each: where do individual stocks lie relative to the CML, and relative to the SML, if CAPM holds?
1 Premium 5.5%: 6.5%, 9.25%, 12%, 14.75%.
2 Required $6.5+1.2(5.5)=13.1\%$; $\alpha=13-13.1=-0.1\%$, just below the SML.
3 $(11.9-6.5)/0.9=6\%$.
4 $\beta_p=0.5(1.4)+0.3(0.9)+0.2(0)=0.97$; required $6.5+0.97(5.5)\approx11.84\%$.
5 $\beta=0.0288/0.0256=1.125$.
6 $6.5+(-0.3)(5.5)=4.85\%$, below $r_f$. An asset that tends to rise when the market falls reduces portfolio risk — it acts like insurance, and investors accept a lower expected return for it.
7 Individual stocks lie below the CML (they carry specific risk that is not rewarded, so their Sharpe ratios are lower than the market's). Under CAPM every stock lies on the SML, because its expected return depends only on its beta.
III
Factors (all premiums and coefficients hypothetical).
  • 1 A fund has $\beta=1.0$, $s=0.5$, $h=-0.2$, $m=0.3$. With premiums MKT 6%, SMB 2%, HML 3%, WML 4% and zero alpha, find the predicted excess return. (Compare Example 6.)
  • 2 A fund's CAPM alpha is 2.4%. A three-factor regression with the same market beta gives $s=0.6$, $h=0.5$; SMB and HML averaged 1.5% and 3%. Find the three-factor alpha and interpret.
  • 3 Give two reasons why a factor premium measured over the past 20 years might not be earned over the next 20.
  • 4 A momentum rule ranks stocks by the return from 12 months ago to 1 month ago. Five stocks were priced (₹) 100, 250, 80, 400, 60 twelve months ago and 118, 240, 104, 420, 57 one month ago. Which two does the rule pick?
  • 5 A "value" fund has a market beta of 1.0. Explain why it can still lose 30% in a year when the market falls 30%, whatever its value exposure.
1 $6+0.5(2)-0.2(3)+0.3(4)=7.6\%$ over $r_f$.
2 Tilts explain $0.6(1.5)+0.5(3)=2.4\%$, so the three-factor alpha is $0\%$: the whole CAPM alpha came from small-cap and value exposure, not from unexplained skill.
3 (i) The past average may be partly luck or data-mining (many rules were tried; the best-looking survive). (ii) Once a premium is widely known and cheap to harvest, more money chases it, prices adjust and the future premium can shrink. Also: the premium may be compensation for a risk that shows up in bad times, and bad times can last.
4 Returns: +18%, −4%, +30%, +5%, −5%. The top two are the third stock (+30%) and the first (+18%).
5 With $\beta=1$ the fund carries the full market risk: a 30% market fall predicts about a 30% fall in the fund. Factor tilts add or subtract a few points around that; they do not remove market exposure.
IV
Allocation and rebalancing.
  • 1 ₹5,00,000 at 70/30 equity/debt. Equity returns −20%, debt +7%. Find the new equity weight and the trade that restores 70/30.
  • 2 Repeat part 1 with equity +25%, debt +6%.
  • 3 A 60/40 portfolio uses a ±5-point band. If debt earns 0.5% in a month, what equity return in that month would push the equity weight to 65%? Down to 55%?
  • 4 Estimate the diversification return of an annually rebalanced 50/50 equity–debt mix with $\sigma_E=20\%$, $\sigma_D=5\%$, $\rho=0.2$.
  • 5 Under "100 minus age" and "110 minus age", find the equity share at 30 and at 55. Name two facts about an investor that these rules ignore.
  • 6 A 60/40 target portfolio of ₹12 lakh has drifted to 64/36. How much new money must go into debt alone (ignoring returns meanwhile) to restore 60/40? At ₹40,000 a month, how many months is that?
  • 7 A glide path lowers the equity target linearly from 80% at age 30 to 40% at age 60. Find the target at 45 and the yearly change.
1 Equity ₹2,80,000, debt ₹1,60,500, total ₹4,40,500; equity weight ≈ 63.56%. Target equity ₹3,08,350: buy ₹28,350 of equity, sell the same amount of debt.
2 Equity ₹4,37,500, debt ₹1,59,000, total ₹5,96,500; weight ≈ 73.34%. Target ₹4,17,550: sell ₹19,950 of equity.
3 Up: $\dfrac{0.6(1+x)}{0.6(1+x)+0.4(1.005)}=0.65$ gives $1+x=\dfrac{0.65\times0.402}{0.6\times0.35}$, $x\approx+24.43\%$. Down: $x\approx-18.11\%$. A ±5-point band trades only after big moves.
4 $\sigma_p^2=0.25(0.04)+0.25(0.0025)+2(0.25)(0.002)=0.011625$; $\sum w_i\sigma_i^2=0.02125$; bonus $\approx\tfrac12(0.02125-0.011625)\approx0.48\%$ a year.
5 100 − age: 70% at 30, 45% at 55. 110 − age: 80% and 55%. They ignore, for example, the goal and its date, income stability, other assets (house, provident fund), and the investor's tolerance for losses.
6 Equity stays ₹7.68 lakh, so we need $7.68=0.6(12+X)$: $X=$ ₹80,000 — two months of ₹40,000, with no selling and no realised gains.
7 The target falls by $40/30\approx1.33$ points a year; at 45 it is $80-15\times1.333=60\%$.
🎯 Practice corner NISM-styleCase studyInterview

E1. NISM-style According to the Capital Asset Pricing Model, the expected return of a security depends on (A) its total risk (B) its systematic risk, measured by beta (C) its unsystematic risk (D) its past returns.

Reason $\mu_i=r_f+\beta_i(\mu_M-r_f)$: only beta enters. Unsystematic risk can be diversified away, so it is not rewarded (Theorem 4.12).
Answer (B).

E2. NISM-style The risk-free rate is 7%, the expected market return 12% and a stock's beta 1.2. Its CAPM required return is (A) 13.0% (B) 14.4% (C) 8.4% (D) 19.0%.

Compute $7\%+1.2\times(12\%-7\%)=13\%$.
Distractors (B) is $1.2\times12\%$ (forgets $r_f$); (C) is $1.2\times7\%$; (D) adds the two rates.
Answer (A).

E3. NISM-style An investor's target allocation is 60% equity and 40% debt. After a strong rally in equities the allocation is 68:32. Rebalancing to the target requires (A) buying more equity (B) selling some equity and buying debt (C) no action, since equity is performing well (D) moving everything to cash.

Reason Equity is overweight by 8 points; restoring 60:40 means selling equity worth 8% of the portfolio and buying debt (Theorem 4.13), or directing new money to debt.
Answer (B).

E4. Case study An equity fund has $\mu=13\%$, $\sigma=22\%$; a debt fund $\mu=7.5\%$, $\sigma=6\%$; correlation 0.15; a liquid fund (treated as risk-free) pays 6.5%. (a) Find the tangency weights, return, volatility and Sharpe ratio. (b) Compare with each fund's Sharpe ratio. (c) An investor wants 8% volatility. Give the split of her money among the three funds and her expected return.

(a) $e_E=6.5\%$, $e_D=1\%$, $\sigma_{ED}=0.15(0.22)(0.06)=0.00198$: $$w_E=\frac{0.065(0.0036)-0.01(0.00198)}{0.065(0.0036)+0.01(0.0484)-0.075(0.00198)}\approx37.61\%.$$ Then $\mu_T\approx9.57\%$, $\sigma_T\approx9.58\%$, $S_T\approx0.320$.
(b) Equity: $6.5/22\approx0.295$; debt: $1/6\approx0.167$. The mix beats both.
(c) $y=8/9.58\approx0.835$: equity $0.835\times37.61\%\approx31.4\%$, debt $\approx52.1\%$, liquid fund $\approx16.5\%$. Expected return $6.5\%+0.835\times3.07\%\approx9.06\%$.

E5. Case study ₹8,00,000 is invested 50/50 in equity and debt. Year 1: equity +40%, debt +5%. Year 2: equity −30%, debt +8% (hypothetical). (a) Find the value and weights after year 1 and the rebalancing trade. (b) Compare the values after year 2 with and without annual rebalancing, and the equity weight in each.

(a) Equity ₹5,60,000, debt ₹4,20,000, total ₹9,80,000; equity weight ≈ 57.14%. Rebalance to ₹4,90,000 each: sell ₹70,000 of equity, buy debt.
(b) Rebalanced Equity ₹3,43,000, debt ₹5,29,200: total ₹8,72,200, equity weight ≈ 39.3%.
(b) Buy-and-hold Equity ₹3,92,000, debt ₹4,53,600: total ₹8,45,600, equity weight ≈ 46.4%. Rebalancing is ₹26,600 ahead on this reversing path; both have drifted below 50% equity, and the next rebalance buys equity.

E6. Interview "If CAPM holds, would any rational investor buy an asset whose expected return is below the risk-free rate?" Answer with a number: $r_f=6\%$, market premium 6%, an asset with $\beta=-0.4$.

Key insight Yes. CAPM prices an asset by its contribution to the risk of the market portfolio, not by its own volatility. A negative-beta asset tends to rise when the market falls, so adding it reduces portfolio risk.
Number Required return $6\%+(-0.4)(6\%)=3.6\%$, below $r_f$.
Interview answer "Yes — it is like buying insurance: you accept a lower expected return, even below the risk-free rate, because it pays off in the states where everything else is doing badly. The same logic is why put options have negative expected returns for their buyers yet are still bought (§6.2)."
🧑‍🏫 For the teacher

Suggested plan — 2 lectures × 60 min.

  1. From the frontier to CAPM. Recall the frontier of §4.1. Add $r_f$: derive Theorem 4.8 in two lines and draw several CALs; ask which one an investor wants. Derive the tangency portfolio (Theorem 4.9) and do Examples 1–2; project Interactive 1 and move $r_f$ to show how fragile $T$ is. Theorem 4.10 and Example 3 (cautious vs growth, and why leverage at a realistic cost loses). Then the CAPM derivation — the key line is $\operatorname{Cov}(r_i,r_T)=k(\mu_i-r_f)$ — and switch Interactive 1 to the SML view. Examples 4–5.
  2. Factors and allocation. Factor ideas descriptively (Definition 4.8, Example 6). Then allocation: Theorem 4.13 with Example 7 on the board; Theorem 4.14 and the seesaw (Example 8 — students are usually amazed). Project Interactive 2 and try several market paths: count how often buy-and-hold wins and how far its equity weight wanders. Close with age rules and the 📈 Market case.

Class activity. Each student writes a one-page "investment policy" for a hypothetical person (age, goals with dates, income type): target mix, rebalancing rule (calendar or band), and how new SIP money is used. Pairs then run each other's rule through Interactive 2 on three market paths and report the range of outcomes.

Misconceptions to watch for. CAPM return = forecast; volatility is rewarded; CML applies to single stocks; alpha = skill; factors guarantee premiums; rebalancing always adds return; 100 − age is a law.

Suggested homework. I(1, 3, 4), II(1, 2, 4, 6, 7), III(1, 2, 4), IV(1, 3, 4, 6); Practice corner E2, E4, E5.

🛠️ Do it yourselfSpreadsheet mini-project · about 90 minutes

Project. Build a rebalancing backtest for a two- or three-asset mix on simulated (or downloaded) monthly returns, and compare buy-and-hold, annual and band rebalancing — including a simple cost and tax line.

Why it matters. A rebalancing rule is the part of an allocation plan you actually execute. Seeing its effect on risk, return, turnover and tax on many paths is the honest way to choose one.

What to do.

  1. Returns. Either download month-end values of an equity index and a debt index fund NAV (from the index provider's or fund house's website) and compute monthly returns =B3/B2-1, or simulate: equity =EXP((0.12-0.02)/12+0.2/SQRT(12)*NORM.S.INV(RAND()))-1, debt =EXP((0.07-0.00125)/12+0.05/SQRT(12)*NORM.S.INV(RAND()))-1 (copy, then paste as values so they stop changing).
  2. Buy-and-hold columns: equity value =prev*(1+rE), debt value =prev*(1+rD), total, equity weight.
  3. Annual rule: same, but in every 12th row set equity =w*total and debt =(1-w)*total, after subtracting a cost =c*ABS(trade) with an illustrative $c=0.1\%$.
  4. Band rule: rebalance in a row only if ABS(weight-w)>band; count trades with =COUNTIF(flag,TRUE).
  5. For each rule: final value, CAGR =(final/initial)^(12/months)-1, volatility of monthly returns =STDEV.S(...)*SQRT(12), maximum drawdown (running peak =MAX(prev_peak,total), drawdown =total/peak-1, then =MIN(...)), and total amount of equity sold.
  6. Tax line: for each equity sale, gain ≈ amount sold × (1 − cost/value) — keep it symbolic with an illustrative rate $t$, and note that real rules depend on holding period and exemptions.

Python version: loop over months; keep E, D; E*=1+rE; D*=1+rD; if rule(k, E/(E+D)): T=E+D; E,D = w*T, (1-w)*T; record values; repeat for 1,000 seeded paths and compare the distributions of final value and maximum drawdown.

Hand in: one chart of values and one of equity weights (like Interactive 2), a table of the four statistics per rule, and three sentences on what the rule buys you and what it costs.

Going further. Replace selling with "rebalancing by new money": direct a monthly SIP into whichever asset is under target. How close does it keep the weights, and how much tax does it avoid?

Instructor note: as a check, Example 7's two years must reproduce ₹10,57,112 (rebalanced) and ₹10,38,680 (buy-and-hold) with no costs. On simulated paths, students should find that annual rebalancing narrows the spread of maximum drawdowns more reliably than it raises the average final value.

🧪 Self-check quiz

0/5 correct

Q1. The slope of the capital allocation line of a risky portfolio $P$ is

Q2. With $r_f=6\%$ and a market risk premium of 5%, a stock with $\beta=1.4$ has a CAPM required return of

Q3. Under CAPM, which kind of risk earns a higher expected return?

Q4. A 60/40 equity–debt portfolio earns equity +30% and debt +6% in a year. The new equity weight is about

Q5. The main purpose of rebalancing a strategic allocation is to