Since its inception financial research has been on the hunt for factors that can consistently generate positive returns. Most famously Fama and French’s value factor.
This search has led to a what one author has termed the “factor zoo” – a proliferation of factors – a direct consequence of data mining.
There is also a replication crisis – that factors are not internally (i.e. the results can’t be replicated within the original sample) and externally (i.e. results can’t be replicated out of sample) valid.
This paper (summary here) is a rebuttal of these issue – it uses Bayesian updating from a prior that a factor’s usefulness is zero. Their work finds that no crisis exists.
One idea worth thinking about is that according to the authors the 153 factors explored actually cluster into 13 themes – “possessing a high degree of within-theme return correlation and economic concept similarity, and low across-theme correlation” (as seen in the chart).
