If a paper finds a pattern predicting which shares will rise, it stands to reason that investors will exploit it. It also stands to reason that the exploitation will eat the pattern itself. Ninety-seven published variables were examined before publication and after. What happened to them?
They lost most of the return. the return fell by 26%, and it fell by 58%. The difference suggests wears away as investors learn the pattern and trade on it. Part of the rest probably came from , producing a rule that looked good in the past and did not survive into the future.
A stock market responds to knowledge. If a paper reveals that a particular characteristic predicts return, buying the matching shares early pushes their price up and narrows the future profit.
The authors gathered 97 predictors published in the literature and divided their time into three: the original sample, years that were not used in the research but preceded publication, and the years after publication.
The 26% decline before publication serves as an upper bound on the effect of overfitting to data. The total 58% decline afterwards leaves a gap consistent with traders having learned from the research.
The decomposition is not a controlled experiment, so the numbers are no absolute division between error and learning. The clear message is that a historical return weakens both when it meets fresh time and when it becomes known.
wat. After 97 return patterns were published, the profit associated with them fell by 58% on average.
wat wat
An investment strategy is not a fixed law of nature: the knowledge of it can move prices, and a historical test can find a pattern that was never there.
How do we know?
The authors split each variable’s return into the research period, a fresh period before publication, and the period after the paper became public.
What is still unknown
The split cannot say precisely how much of the decline came from the market learning and how much from other changes in the economy and in trading.
Learn something?
Where to next
Does a day of falls on the stock market show up in hospital records?markets
How does the list of funds at work change the risk in your savings?markets
What happened to a share after more people searched for it online?markets