Factor Investing

Quantitative investing is based on backtesting. Backtesting is based on data. Traditionally, researches did not question the start dates for most of their data and resulting backtests. The earliest traditional data set for U.S. equities starts in 1925 via the CRSP dataset. However, U.S. and many global equity markets have centuries of data before that. This book lifts the curtain of history beyond the traditional start dates for the most popular factors and with newly extended...
More Description

Quantitative investing is based on backtesting. Backtesting is based on data. Traditionally, researches did not question the start dates for most of their data and resulting backtests. The earliest traditional data set for U.S. equities starts in 1925 via the CRSP dataset. However, U.S. and many global equity markets have centuries of data before that. This book lifts the curtain of history beyond the traditional start dates for the most popular factors and with newly extended histories helps answer important questions about factors. Most importantly, the existential question of whether factors are real in the first place or results of datamining over the short histories in which they were discovered. And if they are real, what should their expected returns and risks look like over the long-run. How often do these factors crash and why? For how long can they be expected to underperform? Answers to these questions help investors avoid the panic and confusion that arises from the recent factor performance deterioration. Longer history puts recent history into perspective by revealing a more complete distribution of outcomes for various factors, that have not been seen in the shorter histories.

In addition to reviewing and replicating some recent factor extension studies of Price Momentum back to 1800 for country equities, bonds, currencies and commodities, this book reports new extended evidence for the Value and Size factors back to 1800 in the U.S. and international stock level data, and global equities, bonds, currencies and commodity futures. In addition, this book works with a unique hand collected data set of commodity futures prices extended back to 1871, the inception date of the first commodity futures contract. Using this data, the book reports Value, Momentum and Basis in commodity futures - which are currently documented back to only 1959.

The main insights of the book are as follows:

* Longer history can teach investors a lot. "History doesn't repeat itself, but it often rhymes", as Mark Twain allegedly said. Many quants and investors tend to ignore longer history deeming it irrelevant to the ever-changing present moment. This book claims that long-history is vitally important and could have warned us about the 2009 momentum crash, and the decade-long dry spell that we are experiencing. Having realistic expectations prevents disappointment and unplanned changes in portfolios.
* Factors are real. Momentum, Value, Trend have worked across centuries and across asset classes. Data-mining arguments about these factors are invalid. Academics, especially the rational-efficient-markets-based group, needs to accept this fact, and develop rational or behavioral theories of why these factors have earned positive premia. Because factors are real, properly applied long term factor investing can be beneficial to overall portfolios.
* Factor returns are smaller over the long-run than during the periods in which they were first discovered. This points to a problematic investor's overestimation of factor's expected returns and portfolio benefits of factors.
* Factors periodically crash. For example, before 2009, in the traditional data from Fama French's website, Momentum crashed only once after the great depression. Most quants ignored that early crash as an anomaly, artifact of early data. But 2009, produced a second large crash, and since then the industry has focused a lot more on momentum crashes. Looking at the complete U.S. equity history back to 1800, one sees several other similar crashes, confirming that large crash risk is indeed a part of Momentum's distribution. Samonov and Gray discuss factor crashes and drawdowns across history and asset classes. This knowledge enhances investor's ability to do proper risk management and factor weighting.