Data Dictionary for the French Monthly-Return Study
Identity: French_20261005_revised_history_A10A_v1. The wide table's primary key is month; the research long table's primary key is (snapshot_id, month, industry_id). UTC is used for audit timestamps; month is a calendar interval, not a fabricated trading timestamp. There is no permanent security-level ID.
| Field | Economic meaning / key | Type, unit, frequency | event_time |
available_at |
ingested_at |
revision_at |
as_of |
|---|---|---|---|---|---|---|---|
snapshot_id |
Fixed input identity; part of composite key | string, version | End of snapshot coverage | Earliest release of current version: unknown | Manifest download UTC | Unknown; retain new versions | This candidate's fixed snapshot |
month |
Calendar return month; primary key | YYYYMM string / monthly |
Start and end of that month | Historical month-by-month availability: unknown | Inherits file download | Historical time: unknown | Inherits snapshot |
industry_id |
Long-form ID for a ten-industry column; primary key | enum / monthly | Supplier group effective interval | Member release: unknown | Inherits file | Unknown | Inherits snapshot |
NoDur / Durbl / Manuf / Enrgy / HiTec / Telcm / Shops / Hlth / Utils / Other (each column) |
Value-weighted industry total return, including dividends; the same contract applies to every column | float, decimal simple total return / month | month interval |
Unknown; do not fill with “known at month-end” | Industry file download UTC | Unknown | Fixed snapshot |
Mkt-RF |
Published market excess return | float, decimal monthly spread | month interval |
Unknown; post-hoc attribution only | FF3 download UTC | Unknown | Fixed FF3 |
SMB |
Small-minus-big return spread | float, decimal monthly spread | Same as above | Same as above | Same as above | Same as above | Same as above |
HML |
High-minus-low book-to-market return spread | float, decimal monthly spread | Same as above | Same as above | Same as above | Same as above | Same as above |
RF |
Proxy for the month's risk-free return | float, decimal simple monthly return, not an annual rate | Same as above | Unknown; post-hoc statistics only, with any cash convention stated separately | FF3 download UTC | Unknown | Fixed FF3 |
signal_score |
Compounded return over the prior 11 months, minus 1 | float, dimensionless / monthly | After all inputs are available; historical timing unknown | Computation time in UTC; only in A10-B if authorized | New version if config changes | Input cutoff and snapshot | |
target_weight |
Top 3 each |
float, capital weight / monthly | Month t when allocation takes effect |
Signal cutoff | Future computation time | Protocol version | Corresponding cutoff |
holding_value / cash / nav |
Monetary amount and NAV | float or Decimal, illustrative USD or normalized unit / monthly | Ledger node | When node completes | Future ledger UTC | Do not overwrite prior runs | run_id |
traded_notional / fee |
Traded amount and cost | Amount; fee rate as decimal, bps divided by 10,000 | Rebalance node | When node completes | Future ledger UTC | Run version | Same as above |
gross_return / net_return / label |
Portfolio gross/net return and target-period return | float, decimal / monthly | Holding month t |
After label end and source release/ingestion; historical times unknown | Future computation UTC | Snapshot version | Same as above |
label_start / label_end / label_available |
Label interval and maturity time | UTC timestamps / monthly | Interval boundary | Actual availability unknown; explicitly specified for constructed examples | Future record | New version | Fit cutoff |
cutoff / allocated_at |
Information cutoff and hypothetical allocation-effective time | UTC timestamps | Protocol rule | Rule known in advance | Protocol registration UTC | Revision chain | protocol_id |
Missing fields are rejected or marked unknown; they are not replaced with zero. Subtract RF once from portfolio total return; do not subtract RF again from Mkt-RF / SMB / HML. Do not mix same-named FF5 columns into the FF3 dataset.
Corporate-event applicability
Splits and dividends are already reflected in the supplier's total-return methodology; do not add separate corporate-action cash. There is no security-level event table for an independent reconstruction. Delistings and the stock universe are represented only at the supplier's aggregate level; do not claim to have verified completeness of the underlying whole market. Financial-statement announcements and restatements are not direct feature inputs for the selected price-momentum question and therefore do not apply in this batch. Adding value or quality features would change the question and require new PIT fields; it is not a free option under the current protocol. Historical changes to portfolio classifications are not changes in tradable security IDs. Normalized wealth indices are for ledger demonstrations only and must not enter a real-price execution interface.
This dictionary is for the research plan. The two-asset engineering dictionary covers USD/share, shares, and second-level constructed runs. The interfaces are not interchangeable. The time contract specifies eligibility and label decisions; the protocol gives the full cost and statistical equations; the README gives current acceptance status.