SSRN, Working Papers and Preprint Risk

SSRN (Social Science Research Network) and arXiv host "working papers" — drafts researchers post before, or instead of, formal peer review. A huge share of quant finance's most-cited factor and anomaly papers spent years as working papers, so ignoring them means missing real information. But nothing has checked the paper's data, code, or statistics except the authors themselves. No referee has forced them to show a robustness table, defend a sample-selection choice, or admit a result died out-of-sample.

Preprint risk is the extra chance that a working paper's result is an artifact of the specific sample, specific parameter choices, or specific backtest the authors happened to run — dressed up with a clean abstract and a strong-looking t-stat. The paper may never get published because the effect doesn't survive scrutiny, or it may get published years later in a much-weakened form after referees strip out the overfit pieces.

A practical read of a working paper checks three things before trusting a number: does the sample period end suspiciously close to when the anomaly was "discovered" (a sign the strategy was mined, not tested)? Is there an out-of-sample or later-period test the authors ran themselves, and how much does the effect shrink there? And does the paper cite its own limitations, or read like a sales pitch? A paper that only shows one backtest window and one specification is worth far less than one showing the result survives several.