Unit 1, lesson 4

Returns and self-financing portfolio accounting

Defines simple, gross, price, total and log returns, explains why log returns add over time only while wealth stays positive, and derives dividend-inclusive self-financing accounting through a rebalance.

  • Video 4
  • Contracts, arbitrage and returns

A portfolio worth 250250 currency units receives a dividend, borrows to buy one more share and is worth 284.1284.1 two periods later. How much of that change is investment gain, and what return does it represent? The answer depends on the denominator, on how the dividend is counted and on whether any money came from outside the portfolio. This lesson defines simple, gross, price, total and log returns together with the conditions they need, and derives the dividend-inclusive accounting identity that a self-financing portfolio satisfies at every rebalance.

Key terms

Simple return and gross return

Definition Consider a position over one period, from date tt to date t+1t + 1, with value Vt>0V_t > 0 at the start and value Vt+1V_{t+1} at the end. The end value includes every cash distribution paid to the position during the period, carried to date t+1t + 1 by a stated reinvestment convention, and no money is added or withdrawn from outside. The simple return and the gross return over the period are

Rt+1=Vt+1−VtVt,1+Rt+1=Vt+1Vt.\begin{gathered} R_{t+1} = \frac{V_{t+1} - V_t}{V_t}, \\[0.4em] 1 + R_{t+1} = \frac{V_{t+1}}{V_t}. \end{gathered}

Both are pure numbers per period, not annualized rates. The denominator is the capital committed at the start, so a simple return is defined only when that capital is strictly positive. A position without borrowing or short sales, in assets with nonnegative prices and payouts, has Vt+1≥0V_{t+1} \ge 0 and so Rt+1≥−1R_{t+1} \ge -1, while a position with borrowing or short sales can lose more than its capital and have Rt+1<−1R_{t+1} < -1. Over consecutive periods k=1,…,nk = 1, \dots, n in which each period starts with the end value of the one before, all values are positive and nothing flows in or out, the gross returns multiply,

1+R0,n=VnV0=∏k=1n(1+Rk),\begin{aligned} 1 + R_{0,n} &= \frac{V_n}{V_0} \\[0.4em] &= \prod_{k=1}^{n} (1 + R_k), \end{aligned}

because the intermediate values cancel in the product of the ratios Vk/Vk−1V_k / V_{k-1}.

In plain words A return is a gain measured per unit of the capital that produced it. The gross return is the factor by which that capital grows, and growth over several periods is the product of the factors, not the sum of the returns.

Example A gain of 55 currency units is a return of 5/100=0.055/100 = 0.05 on a capital of 100100 and of 5/250=0.025/250 = 0.02 on a capital of 250250. Now let a position worth 100100 gain 10%10\% and then lose 10%10\%. It is worth 110110 after one period and 110⋅0.9=99110 \cdot 0.9 = 99 after two, so the two-period gross return is 1.1⋅0.9=0.991.1 \cdot 0.9 = 0.99 and the two-period simple return is −0.01-0.01, although the returns of the two periods add to zero. A position that buys one share at 100100 entirely with borrowed money has zero starting capital and therefore no simple return, even though its gain in currency units is well defined.

Price return and total return

Definition Let St>0S_t > 0 be the ex-dividend price of one share at date tt and Dt+1≥0D_{t+1} \ge 0 the cash dividend per share paid at date t+1t + 1 to holders of the share over the period. The price return and the total return of the share over the period are

Rt+1pr=St+1−StSt,Rt+1tot=St+1−St+Dt+1St.\begin{aligned} R^{\text{pr}}_{t+1} &= \frac{S_{t+1} - S_t}{S_t}, \\[0.4em] R^{\text{tot}}_{t+1} &= \frac{S_{t+1} - S_t + D_{t+1}}{S_t}. \end{aligned}

The total return is the simple return of the position that holds the share and keeps its dividend. If the dividend is paid at an earlier date inside the period, then Dt+1D_{t+1} is replaced by the value at date t+1t + 1 of the dividend reinvested by a stated convention, for example in the cash account or in further shares. If a split or another corporate action changes the number of shares or the quoted price, prices and dividends are first put on the same share basis, so that a mechanical drop in the quote is not read as a loss.

In plain words The price return reads only the quote, while the total return counts everything the holder receives. An ex-dividend price no longer contains the dividend that has just been paid, so whenever a dividend is paid, leaving it out understates what the holder earned.

Example A share bought at 100100 ends the period at 102102 and pays a dividend of 33 at the end. The price return is 2/100=0.022/100 = 0.02 and the total return is (2+3)/100=0.05(2 + 3)/100 = 0.05. If the same dividend had been paid halfway through and deposited in a cash account that grows by the factor 1.011.01 over the second half, it would be worth 3⋅1.01=3.033 \cdot 1.01 = 3.03 at the end and the total return would be 0.05030.0503. These numbers are constructed. They show that the payment date changes the total return even when the dividend amount does not.

Log return

Definition For a period with Vt>0V_t > 0 and Vt+1>0V_{t+1} > 0, so that the gross return is strictly positive, the log return is

ℓt+1=ln⁡(1+Rt+1)=ln⁡Vt+1Vt,\begin{aligned} \ell_{t+1} &= \ln(1 + R_{t+1}) \\[0.4em] &= \ln \frac{V_{t+1}}{V_t}, \end{aligned}

with the natural logarithm. Over consecutive periods with positive values at every date and no outside flows, taking logarithms in the product of gross returns gives

ℓ0,n=ln⁡VnV0=∑k=1nℓk.\ell_{0,n} = \ln \frac{V_n}{V_0} = \sum_{k=1}^{n} \ell_k .

This is an exact algebraic identity. It needs no independence and no distributional assumption. For ∣R∣<1\lvert R \rvert < 1 the series ln⁡(1+R)=R−R2/2+R3/3−…\ln(1 + R) = R - R^2/2 + R^3/3 - \dots converges, and ln⁡(1+R)<R\ln(1 + R) < R for every R>−1R > -1 other than 00. The log return is not defined when the end value is zero, since ln⁡(Vt+1/Vt)→−∞\ln(V_{t+1}/V_t) \to -\infty as Vt+1V_{t+1} decreases to 00, and there is no real log return when the end value is negative. Across assets, the simple return of a portfolio over one period is Rp=∑iwiRiR_p = \sum_i w_i R_i, where wiw_i is the fraction of the starting capital Vt>0V_t > 0 placed in asset ii and the weights add to 11. The portfolio log return is ln⁡(1+Rp)\ln(1 + R_p), which in general differs from ∑iwiln⁡(1+Ri)\sum_i w_i \ln(1 + R_i). For nonnegative weights and positive gross returns 1+Ri1 + R_i, the concavity of the logarithm gives

ln⁡(1+Rp)≥∑iwiln⁡(1+Ri),\begin{aligned} &\ln(1 + R_p) \\[0.4em] &\ge \sum_i w_i \ln(1 + R_i), \end{aligned}

with equality exactly when all the RiR_i with positive weight are equal.

In plain words Log returns turn the multiplication of growth factors into addition, so a whole history can be summarized by adding period numbers. The cost of this convenience is the domain, because wealth must stay strictly positive, and the fact that log returns do not average across assets the way simple returns do.

Example For the gain of 10%10\% followed by the loss of 10%10\%, the two-period log return is ln⁡0.99≈−0.01005\ln 0.99 \approx -0.01005, and indeed

ln⁡1.1+ln⁡0.9≈0.09531−0.10536=−0.01005.\begin{aligned} &\ln 1.1 + \ln 0.9 \\[0.4em] &\approx 0.09531 - 0.10536 \\[0.4em] &= -0.01005 . \end{aligned}

The logs add because the logarithm of a product is a sum, not because the two percentages cancel. Now split a capital equally between two assets that return 10%10\% and −10%-10\% over the same period. The portfolio simple return is 0.5⋅0.1+0.5⋅(−0.1)=00.5 \cdot 0.1 + 0.5 \cdot (-0.1) = 0, so the portfolio log return is 00, while the average of the two log returns is about −0.00503-0.00503. Averaging component log returns understates the portfolio’s growth here, as the concavity inequality predicts.

Self-financing portfolio

Definition Fix trading dates t=0,1,…,nt = 0, 1, \dots, n. A stock has ex-dividend price St>0S_t > 0 and pays the dividend Dt≥0D_t \ge 0 per share at date tt to the shares held over the period from t−1t - 1 to tt, and a cash account, the money-market account, has unit value Bt>0B_t > 0. At each date t<nt < n a strategy chooses, using only information available at date tt, the number θt\theta_t of shares and the number βt\beta_t of cash-account units it holds over the period from tt to t+1t + 1. Negative numbers are allowed, a negative βt\beta_t being borrowing in the cash account and a negative θt\theta_t a short sale. The value just after trading at date tt is

Vt=θtSt+βtBt,V_t = \theta_t S_t + \beta_t B_t ,

and the value just before trading at a date t≥1t \ge 1, after the dividend on the old holding is paid, is

Vt−=θt−1(St+Dt)+βt−1Bt.\begin{aligned} V_{t^-} &= \theta_{t-1}(S_t + D_t) \\[0.4em] &\quad + \beta_{t-1} B_t . \end{aligned}

The strategy is self-financing if Vt=Vt−V_t = V_{t^-} at every trading date 1≤t≤n−11 \le t \le n - 1, which is the same as

(θt−θt−1)St+(βt−βt−1)Bt=θt−1Dt.\begin{aligned} &(\theta_t - \theta_{t-1}) S_t \\[0.4em] &\quad + (\beta_t - \beta_{t-1}) B_t \\[0.4em] &= \theta_{t-1} D_t . \end{aligned}

The dividend term uses the old holding θt−1\theta_{t-1}, the shares actually held over the period that earned it. Writing Vn=Vn−V_n = V_{n^-} for the value of the holdings carried into the last date, every self-financing strategy satisfies the wealth recursion

Vn=V0+∑k=1nGk,Gk=θk−1(Sk+Dk−Sk−1)+βk−1(Bk−Bk−1),\begin{aligned} V_n &= V_0 + \sum_{k=1}^{n} G_k , \\[0.4em] G_k &= \theta_{k-1}(S_k + D_k - S_{k-1}) \\[0.4em] &\quad + \beta_{k-1}(B_k - B_{k-1}) , \end{aligned}

since Vk−−Vk−1=GkV_{k^-} - V_{k-1} = G_k for every kk and Vk=Vk−V_k = V_{k^-} at every date. This is a discrete sum of holdings times increments, and its continuous-time counterpart, a stochastic integral, comes later in the course. The accounting is frictionless, with trades at the quoted prices, no transaction costs and one cash account for both lending and borrowing. A deposit or withdrawal CtC_t from outside changes the identity to Vt=Vt−+CtV_t = V_{t^-} + C_t, so a strategy with a nonzero CtC_t at some date is not self-financing.

In plain words A self-financing portfolio pays for every purchase with a sale, with cash it already holds, with borrowing in its own cash account or with a dividend it has just received. Nothing comes in or goes out from outside, so every change in its wealth comes from price moves, dividends and the growth of the cash account. Borrowing is allowed and outside money is not.

Example A strategy is short 22 shares and holds 55 cash-account units when a dividend of 33 per share is paid at a date with St=102S_t = 102 and Bt=60B_t = 60. A short seller owes the dividend, so the right side of the identity is θt−1Dt=−6\theta_{t-1} D_t = -6. If the strategy leaves its stock holding unchanged, its cash holding must fall by 6/60=0.16/60 = 0.1 units to βt=4.9\beta_t = 4.9. Its value is −2⋅105+5⋅60=90-2 \cdot 105 + 5 \cdot 60 = 90 before the trade and −2⋅102+4.9⋅60=90-2 \cdot 102 + 4.9 \cdot 60 = 90 after it.

Portfolio rebalance

Definition A portfolio rebalance at a trading date tt replaces the holdings (θt−1,βt−1)(\theta_{t-1}, \beta_{t-1}) carried into the date by new holdings (θt,βt)(\theta_t, \beta_t), valued at the date-tt prices after the dividend is paid. A self-financing rebalance keeps the value fixed, Vt=Vt−V_t = V_{t^-}, so once a new stock holding θt\theta_t is chosen the cash holding is determined,

βt=Vt−−θtStBt.\beta_t = \frac{V_{t^-} - \theta_t S_t}{B_t} .

Under the frictionless hypotheses the rebalance itself creates no gain or loss. It changes exposure, which is how much the next period’s moves in the stock and the cash account will change the wealth. The number of cash-account units βt\beta_t is distinct from the currency value βtBt\beta_t B_t of the cash position.

In plain words Rebalancing moves wealth between the stock and the cash account at today’s prices. It decides what the next period will do to the portfolio but does not change what the portfolio is worth today. Extra exposure must be paid for by a sale, a dividend or borrowing, and money brought in from outside is a deposit, not a gain.

Example In a constructed example, a portfolio holds 22 shares at S0=100S_0 = 100 and 11 cash-account unit at B0=50B_0 = 50, so V0=250V_0 = 250. At date 11 the share is quoted at S1=102S_1 = 102 and pays D1=3D_1 = 3, and the cash account is worth B1=51B_1 = 51. Before trading,

V1−=2⋅102+2⋅3+1⋅51=261.\begin{aligned} V_{1^-} &= 2 \cdot 102 + 2 \cdot 3 + 1 \cdot 51 \\[0.4em] &= 261 . \end{aligned}

Rebalancing to θ1=3\theta_1 = 3 shares costs 306306, so the cash position is worth 261−306=−45261 - 306 = -45 and β1=−45/51=−15/17\beta_1 = -45/51 = -15/17 units. The third share costs 102102 and is paid for by the dividends of 66, the cash of 5151 and a loan of 4545, and the value stays 306−45=261306 - 45 = 261. Over the second period the share rises to S2=110S_2 = 110 with no dividend and the cash account grows by the factor 1.021.02 to B2=52.02B_2 = 52.02, so

V2=3⋅110−45⋅1.02=330−45.9=284.1.\begin{aligned} V_2 &= 3 \cdot 110 - 45 \cdot 1.02 \\[0.4em] &= 330 - 45.9 = 284.1 . \end{aligned}

The second-period gain of 23.123.1 is a price gain of 88 on each of three shares less 0.90.9 of interest on the loan. Because the strategy is self-financing, its period returns 261/250−1=0.044261/250 - 1 = 0.044 and 284.1/261−1≈0.0885284.1/261 - 1 \approx 0.0885 compound to the two-period return 284.1/250−1=0.1364284.1/250 - 1 = 0.1364. Had the portfolio bought the third share and kept its cash unit worth 5151, it would have held 306+51=357306 + 51 = 357 while only 261261 was available, so 9696 would have come from outside. Reporting 357/250−1=0.428357/250 - 1 = 0.428 as a return would count that deposit as investment gain.

Check yourself

  1. A share is bought at 8080, ends the period at 7676 and pays a dividend of 22 at the end of the period. Find the price return and the total return.
Show the answer

The price return is (76−80)/80=−0.05(76 - 80)/80 = -0.05. The total return adds the dividend to the gain,

76−80+280=−0.025.\frac{76 - 80 + 2}{80} = -0.025 .

The holder lost 2.5%2.5\% of the capital, not 5%5\%, because the ex-dividend price no longer contains the 22 that was paid out.

  1. A position worth 100100 has simple returns 0.20.2, −0.2-0.2 and 0.050.05 in three consecutive periods with no outside flows. Find its three-period simple return and log return, and check that the log returns add. A second position starts with capital 100100, borrows 100100 at zero interest and buys 200200 of a stock whose price then falls by 60%60\%. What are its simple return and its log return?
Show the answer

The gross returns multiply, 1.2⋅0.8⋅1.05=1.0081.2 \cdot 0.8 \cdot 1.05 = 1.008, so the three-period simple return is 0.0080.008, far from the sum 0.050.05 of the period returns. The log return is ln⁡1.008≈0.007968\ln 1.008 \approx 0.007968, and the period log returns add to the same number,

ln⁡1.2+ln⁡0.8+ln⁡1.05≈0.18232−0.22314+0.04879=0.00797.\begin{aligned} &\ln 1.2 + \ln 0.8 + \ln 1.05 \\[0.4em] &\approx 0.18232 - 0.22314 + 0.04879 \\[0.4em] &= 0.00797 . \end{aligned}

The second position ends with stock worth 8080 and a loan of 100100, so its value is −20-20. Its simple return is (−20−100)/100=−1.2(-20 - 100)/100 = -1.2, which is defined because the starting capital is positive. Its gross return −0.2-0.2 is negative, so it has no real log return. The domain condition of the log return fails, and no identity with logarithms can be applied to this period.

  1. A strategy arrives at a trading date with θt−1=10\theta_{t-1} = 10 shares and βt−1=2\beta_{t-1} = 2 cash-account units. At that date St=50S_t = 50, Dt=1D_t = 1 and Bt=100B_t = 100, and the strategy buys 44 shares in a self-financing trade. Find the new cash holding and check the value before and after the trade. What goes wrong if the dividend is computed on the new holding of 1414 shares?
Show the answer

The self-financing identity gives 4⋅50+(βt−2)⋅100=10⋅14 \cdot 50 + (\beta_t - 2) \cdot 100 = 10 \cdot 1, so (βt−2)⋅100=−190(\beta_t - 2) \cdot 100 = -190 and βt=0.1\beta_t = 0.1. Before the trade the value is 10⋅51+2⋅100=71010 \cdot 51 + 2 \cdot 100 = 710, and after it the value is 14⋅50+0.1⋅100=71014 \cdot 50 + 0.1 \cdot 100 = 710. The purchase of 200200 is paid for by the dividend of 1010 and by giving up 1.91.9 cash-account units worth 190190.

Using 1414 shares on the dividend side gives (βt−2)⋅100=14−200(\beta_t - 2) \cdot 100 = 14 - 200, so βt=0.14\beta_t = 0.14 and a value of 714714 after the trade. That exceeds the 710710 available by 44, the dividend on the four shares that were bought only at this date and did not earn it. The holding that earns a dividend is the one carried over the period that ends at the payment date.

How this lesson connects