McLean and Pontiff: Decay After Publication

McLean and Pontiff's 2016 paper asks a question with a genuinely surprising design: does publishing a trading strategy make it stop working? They took 97 stock-return anomalies from the academic literature, and for each one split its history into three periods — the years the original study actually used, the years between the end of that sample and the paper's publication (a period the authors knew about but the wider market did not yet), and the years after publication, when anyone could read the paper and trade on it. Comparing average returns across the three windows isolates a specific mechanism: if a return premium shrinks specifically after the paper becomes public, that is direct evidence that trading on the published information erased the mispricing.

What they found

Average anomaly returns fell by roughly 26% after publication compared with the in-sample period, and importantly the decline was concentrated in the post-publication window specifically — not spread evenly across the whole post-sample period, which is what you'd expect if the anomaly were simply decaying on its own due to changing market conditions rather than being traded away. The out-of-sample-but-pre-publication period showed some decline too, consistent with information leaking informally through seminars, working papers, and practitioner networks before formal publication, but the drop after publication was sharper.

Worked example. Suppose an anomaly showed an average monthly long-short return of 0.8% in its original sample period, say 1975–2000. In the years between the paper's sample ending and its eventual publication — perhaps 2001–2004, when the working paper was circulating among academics and some practitioners but not yet public — the return averages 0.65%, a modest decline consistent with early information leakage. After formal publication in 2005, the same portfolio construction applied to 2006–2015 data returns an average of 0.55% a month. That total decline from 0.8% to 0.55% is a drop of roughly 31%, and McLean and Pontiff's point is that the sharper leg of that decline lines up with the publication date specifically, not with the passage of time alone — the paper is a plausible cause, not just a correlated event.

Which anomalies decayed most

McLean and Pontiff found the decay was larger for anomalies that were easier to trade — liquid, large-cap stocks with low transaction costs saw sharper post-publication declines than anomalies concentrated in small, illiquid names where trading costs and limits to arbitrage make it expensive for capital to fully arrive and correct the mispricing. That pattern is itself informative: it says the market's response to new information is bounded by how cheaply that information can actually be traded on, not by how quickly everyone reads the paper.

What this means for using published research

The result gives a rough, evidence-based prior: if you are considering trading a signal straight out of an academic paper, expect roughly a quarter to a third of its historical return to already be gone by the time you can act on it, purely from the fact that it is public. This does not make published anomalies worthless — many still carry a genuine premium after decay — but it argues against taking a paper's headline Sharpe ratio at face value for a live strategy, and it is part of why serious quant shops treat academic factor papers as a starting hypothesis to test on fresh, out-of-sample data rather than a ready-made strategy.

In interviews

This paper is the standard answer to "does publishing a strategy destroy it?" — say yes, cite the roughly 26–30% post-publication decline, and explain the three-window design that lets them attribute the decline specifically to publication rather than to time passing. A strong answer adds the liquidity nuance: decay is worse in liquid names because that's where arbitrage capital can actually act, which ties the empirical result back to a mechanism (limits to arbitrage) rather than leaving it as a bare statistic.