Pre-1963 Tests of the Classic Factors

Almost every well-known factor study — Fama-French, momentum papers, and most of what feeds into standard factor data libraries — starts its sample around 1926, or in the case of many with cleaner accounting data, 1963. That start date is not chosen because markets or the value and momentum effects began then; it is chosen because the standard US database (CRSP) begins coverage in 1926, and reliable book-value data for the value factor becomes usable only from around 1963 onward. A factor "confirmed" only on data starting in 1963 is really being tested on one continuous sample of roughly sixty years, which is a real constraint on how much confidence a single backtest deserves.

Researchers who care about this have pushed the tests back further using hand-collected or alternative datasets: studies using UK and US data back into the 1800s, and reconstructions of momentum and value effects using newspaper-archived prices from the early twentieth century, generally find the same premia persisting — momentum profits appear in US data back to the 1920s and in some UK reconstructions back further still, and value's outperformance shows up in samples predating the CRSP era. This out-of-sample evidence, entirely disjoint from the period most factor papers were built and tuned on, is one of the stronger pieces of evidence that factors like value and momentum are not statistical artifacts of a single overfit sample window.