Chapters

Chapters

Lesson 03 / 05

In this lesson

Reconstructing the original opportunity set
Closures, mergers and weighting
Membership at the historical date
Observation, release and revision dates
Remaining evidence and decision limits

Count the investments that disappeared

Reconstruct the choices and information available then, including investments that later disappeared and data released only afterward.

Restore the fund missing from the presentation.

Four fictional funds start with €10,000 each. Complete one-year total returns include distributions and fund expenses, before investor tax. No extra flows or personal costs. D closes at year end and returns €6,000.

Common scale: €0–€12,000 · dashed line = each €10,000 start

A€12,000 · +20%

B€11,000 · +10%

C€9,000 · −10%

D · closed€6,000 · −40% · paid on closure

A + B + C only
€32,000 / €30,000
Mean ≈ +6.67%

A + B + C + D
€38,000 / €40,000
Mean −5%

Removing D from the display does not remove its loss. Closure itself establishes neither underperformance nor total loss: this complete record shows €6,000 returned.

Reconcile all four original holdings and closure proceeds

Equal starting amounts make these equal-weight means. Survivors: (20% + 10% − 10%) ÷ 3 ≈ 6.67%. Original set: (20% + 10% − 10% − 40%) ÷ 4 = −5%.

SPIVA includes liquidated/merged funds in their original opportunity set. The closure label alone cannot choose the recorded return.

How were exits, missing observations and weights handled?

A mid-study closure needs pre-exit returns, proceeds and treatment of the remaining period. Dropping the record, assigning every missing value zero or assuming immediate recovery changes the answer.

Equal-weighted and asset-weighted averages ask different questions. Specify weight dates; neither reconstructs an individual’s dated cash flows. Unrecoverable omissions remain a limitation.

Mergers and multiple share classes can change the apparent sample

A merged fund’s later record does not automatically replace its predecessors. Multiple share classes can make one strategy look like independent choices. SPIVA distinguishes average-fund and larger-fund-weighted views; the same opportunity-set issues extend to companies, assets and markets.

Today’s constituents were not all known then.

Current index membership reflects the additions, removals and survival that occurred after the simulated decision date. Historical prices for today’s names cannot alone reconstruct the eligible choices. Use historical members, corporate events and effective dates.

Put information on its availability date.

Fictional March figure. Events in order; spacing does not measure elapsed days.

1 April · decisionMarch has ended. Neither version available.
15 April · first release+0.2% enters the public record.
1 May · decisionCan use +0.2%. June revision unavailable.
15 June · revision−0.4% becomes available only now.

The observation month is not a publication date. Later releases/revisions give a simulation information an investor lacked, even when describing an earlier period.

Vintage data and formation timelines

ALFRED retains vintages rather than only today’s revised estimate. The French Data Library specifies formation dates and prior accounting periods: evidence to assess, not a personal factor-selection routine to reproduce.

Correct arithmetic cannot restore unavailable evidence.

Survivorship removes observations; look-ahead uses unavailable information. Both can arise without deliberate deception. Better evidence still leaves implementation and future uncertainty.

Check your understanding

Try each question, then check your answer.

Choose questions from another chapter

Reconstructing an index portfolio

A simulation buys today's index members at prices from twenty years ago. What is missing from the historical-investor perspective?

The membership and eligible choices that were actually known at each earlier date.
A current sector breakdown to apply unchanged across all earlier dates.
Historical prices for today's members, without earlier membership records.

Show the explanation

Correct answer: The membership and eligible choices that were actually known at each earlier date.

Current membership incorporates later changes, including which companies survived or qualified. A historical test needs the opportunity set and rules available then, not future knowledge of today's constituents. It must also handle companies that disappeared or entered later.

Return to the explanation

A December figure

A figure describes December but is first released on 15 February. A simulation uses it for a 3 January decision. What is wrong?

Nothing, because December had ended before 3 January.
It uses information that was not yet public at the simulated decision date.
It is acceptable if a later revision confirms the figure, because that validates the earlier decision input.

Show the explanation

Correct answer: It uses information that was not yet public at the simulated decision date.

The figure describes an earlier period, but investors could not use that release before 15 February. Use the appropriate vintage and availability date. Substituting a later revision can introduce another information advantage rather than repair the test.