Why Subperiod Boundaries Matter
An investment account’s balance can change for two different reasons: the investments can gain or lose value, and the investor can add or remove money. A time-weighted return measures performance across intervals separated by external contributions and withdrawals. The first task is therefore to identify the correct interval boundaries—not to treat every balance change as investment performance.
A boundary is placed at each time an external cash flow enters or leaves the account being measured. Each resulting subperiod runs from one boundary to the next. This tutorial focuses on identifying and ordering those boundaries. It does not calculate the return within an interval or combine interval returns.
“External” is defined relative to the account being measured. If a portfolio sells one security and uses the proceeds to buy another, the money has moved within the portfolio; that trade is not an external cash flow. Interest or a dividend retained in the account is also internal. If the dividend is paid out of the account to the investor, it is an external outflow. The same payment can therefore be internal or external depending on what account is being measured and whether the money remains in it.
For boundary identification, use the time at which the flow is credited to or removed from the measured account. If the record supplies only a calendar date, use that date as the boundary and follow the stated timing convention consistently. In a precise valuation record, the interval before the flow ends immediately before it is applied, and the interval after the flow begins immediately after it is applied.
Why an External Flow Creates a Boundary
Suppose the account is valued immediately before and after a flow at time \(t_j\). Let \(V(t_j^-)\) denote its value just before the flow and \(V(t_j^+)\) its value just after. If \(F_j\) is the net amount transferred into the account, with an outflow recorded as a negative amount, then under the simple assumption that the transfer itself does not change the account’s investment value,
This is an accounting identity, not a return formula. It shows why the flow must be separated from investment performance: the balance can jump by \(F_j\) even if no investment gain or loss occurs at that instant. Ending one interval just before the flow and beginning the next just after it keeps the flow from being mistaken for investment growth or decline.
The boundary does not imply that a return has already been calculated. It identifies which dates belong together so that the performance calculation for each interval can use the appropriate account values and exclude the external transfer from that interval’s investment result.
Write down the exact start and end times. Do not assume that a calendar month or year is the measurement period unless the problem says so.
Ask whether money crosses the boundary of the measured account. Mark contributions and withdrawals; do not mark internal trades or income retained in the account.
Put each distinct external-flow time between the start and end. Combine flows at the same instant for partitioning purposes.
Each pair of adjacent times in the ordered list gives one subperiod. Check that the intervals cover the measurement window without overlap or gaps.
Worked Examples
Worked Example: Several Flows in a One-Year Record
A hypothetical investment account is measured from January 1, 2025, through January 1, 2026. The investor contributes money on March 31 and September 30, then withdraws money on December 15. Identify the subperiods and their lengths in days. Use ordinary elapsed calendar days, so the start date is not counted as a full elapsed day.
Method. Create one boundary at the measurement start, one at each distinct external-flow date, and one at the measurement end. The amounts of the flows are not needed to locate boundaries.
Step 1: Order the boundary dates. They are January 1, March 31, September 30, December 15, and January 1, 2026. There are four consecutive intervals, so there are four subperiods.
Step 2: Find the first length. January 1 to March 31 is 31 days in January, 28 in February, and 30 days from March 1 to March 31: \(31+28+30=89\) days.
Step 3: Find the second length. March 31 to September 30 is 30 days in April, 31 in May, 30 in June, 31 in July, 31 in August, and 30 in September: \(30+31+30+31+31+30=183\) days.
Step 4: Find the third length. September 30 to December 15 is 31 days in October, 30 in November, and 15 days in December: \(31+30+15=76\) days.
Step 5: Find the final length. December 15 to January 1, 2026, is 16 elapsed days to December 31 plus one day to January 1, for \(17\) days.
The partition is therefore January 1–March 31 (89 days), March 31–September 30 (183 days), September 30–December 15 (76 days), and December 15–January 1 (17 days). As a check, \(89+183+76+17=365\) days, the full length of the measurement window. The common mistake avoided here is splitting at arbitrary month-end dates: only the stated flow dates create internal boundaries.
Worked Example: Irregular Dates and a Mid-Record Start
A hypothetical account is measured from February 10, 2025, through June 25, 2025. It receives an external contribution on April 5 and makes an external withdrawal on May 20. Determine the ordered subperiods and their elapsed lengths.
Method. Use the actual dates in the measurement window rather than assuming equal monthly periods. The record does not include a flow at the beginning or end, so those dates remain the first and last boundaries.
Step 1: List the dates in order. The boundaries are February 10, April 5, May 20, and June 25. This gives three subperiods.
Step 2: Count February 10 to April 5. From February 10 to March 1 is 19 elapsed days; March has 31 days; April 1 to April 5 adds 4 days. Thus the first interval is \(19+31+4=54\) days.
Step 3: Count April 5 to May 20. April 5 to May 1 is 26 days, and May 1 to May 20 is 19 days. The second interval is \(26+19=45\) days.
Step 4: Count May 20 to June 25. May 20 to June 1 is 12 days, and June 1 to June 25 is 24 days. The final interval is \(12+24=36\) days.
The subperiods are February 10–April 5 (54 days), April 5–May 20 (45 days), and May 20–June 25 (36 days). The check is \(54+45+36=135\) days; counting directly from February 10 to June 25 also gives 135 elapsed days. The trap is treating “monthly” as equal-length intervals. The actual boundary dates determine the intervals, which need not have equal lengths.
Worked Example: Separate External Flows From Internal Activity
A hypothetical portfolio is measured from January 1, 2025, through January 1, 2026. On May 15, a dividend is received and retained as cash in the portfolio. On May 20, one holding is sold and another is purchased using the proceeds. On June 30, the investor contributes $10,000 and withdraws $2,000 at the same recorded time. On August 1, the investor makes a separate withdrawal. Identify the boundaries and subperiod lengths.
Method. Classify each event relative to the portfolio, then use only the distinct external-flow times as internal boundaries.
Step 1: Classify the May events. The dividend remains in the portfolio, so it is not an external flow. The May 20 sale and purchase are transactions within the portfolio, so they also do not create boundaries.
Step 2: Classify the June 30 events. Both the contribution and withdrawal cross the account boundary. They occur at the same recorded time, so June 30 is one distinct boundary; there is no positive-length interval between them.
Step 3: List all boundaries. The ordered dates are January 1, June 30, August 1, and January 1, 2026. There are three subperiods.
Step 4: Count the lengths. January 1 to June 30 is 180 days. June 30 to August 1 is \(31\) days in July plus \(1\) day to August 1, or 32 days. August 1 to January 1, 2026, is \(31+30+31+30+31=153\) days for August through December.
The subperiods are January 1–June 30 (180 days), June 30–August 1 (32 days), and August 1–January 1 (153 days). Their lengths sum to \(180+32+153=365\), matching the measurement window. A common mistake would be to split at the retained dividend or at the internal trade. Those events change the portfolio’s composition or holdings, but they do not transfer money across the measured account boundary.
Boundary Conventions and Common Traps
The date of an external flow is a boundary, not an extra interval. For example, a contribution on June 30 means that the interval before it ends at that boundary and the next interval begins there. It does not create a zero-length subperiod on June 30. If two flows share a timestamp, list the timestamp once for partitioning, even if their amounts are recorded separately.
A flow at the measurement start or end does not add an internal boundary. The start and end are already boundaries. A flow on the starting date belongs at the start of the record under the stated timing convention; a flow on the ending date belongs at the end. If the problem does not clarify whether a same-day flow occurred before or after the valuation, do not silently invent an intraday ordering. State the convention supplied by the problem, or note that exact return measurement would require more precise timing information.
The following checks help keep a partition consistent:
- Every external-flow time inside the window appears once. Omitting one leaves a subperiod that crosses a change in invested capital.
- No internal portfolio event creates a boundary. Retained income, reinvestments, and trades within the account are not investor contributions or withdrawals.
- Boundary dates are chronological and distinct. Same-time flows do not create an interval between them.
- Intervals cover the full measurement window. Add their elapsed lengths and compare with the full elapsed time.
- Amounts and timing are different tasks. Flow amounts may matter in later calculations, but the partition itself requires the event classification and its time.
Check Your Understanding
For each situation, identify the ordered subperiods. When asked, also give elapsed lengths in days.
- A hypothetical account is measured from January 1 to July 1, 2025. It receives a contribution on February 1 and has a withdrawal on April 1. List the subperiod boundaries. (Answer: January 1, February 1, April 1, July 1; three subperiods.)
- A portfolio is measured from March 10 to May 10, 2025, with one contribution on April 10. Give the subperiod lengths using elapsed calendar days. (Answer: March 10–April 10 is 31 days; April 10–May 10 is 30 days.)
- From January 1 to January 1 of the following year, an account has a retained interest credit on March 1, an external withdrawal on March 15, and an internal sale on October 1. Which dates are internal boundaries? (Answer: March 15 only; the interest credit is retained internally and the sale is an internal transaction.)
- A hypothetical record runs from January 1 through December 31, 2025, with external flows on April 30, July 31, and October 31. State the boundary sequence and number of subperiods. (Answer: January 1, April 30, July 31, October 31, December 31; four subperiods.)
- An account is measured from June 1 to September 1, 2025. It has a $7,000 contribution and a $1,500 withdrawal at the same recorded time on July 1, then a further contribution on August 1. Give the ordered boundaries and the elapsed lengths. (Answer: June 1, July 1, August 1, September 1; 30 days, 31 days, and 31 days.)