Reproducibility in Finance - 'When Equity Factors Drop Their Shorts' Paper

Introduction

Welcome to another post on reproducibility in finance literature!

Last year, Pim van Vliet from Robeco‘s quant team tweeted a new research study (joint work with David Blitz and Guido Baltussen) – “When equity factors drop their shorts"¹ – investigating the common equity factors from the perspectives of long and short leg separately and assessing each leg’s contribution to the overall performance of the portfolio.

NEW STUDY | Curious to know what happens ‘when equity factors drop their shorts’? Here you go quants!👇 https://t.co/LRranuQzX3

— Pim van Vliet (@paradoxinvestor) November 27, 2019

I was quite intrigued with the title of the paper, so decided to spend some time reading and then reproducing the results of the paper using Python 🐍 and better understand the methodology followed by the authors. All the code is available on Github.

⚠️ The purpose of this post is just to confirm the reproducibility of the research. Paper is already well summarized in the below links.

Nice discussion of paper ‘when Equity factors drop their shorts’ by @larryswedroe https://t.co/GHZKaacxuv

— Pim van Vliet (@paradoxinvestor) December 10, 2019

Do short legs add value? Summary👇 https://t.co/XUScvLg9Mq

— Pim van Vliet (@paradoxinvestor) December 16, 2019

Data

Most of the data used in the paper is sourced from Kenneth French Data Library. I’d like to thank Matthias Hanauer for assisting with data on 2x3 size-volatility sorted factor portfolios for US and ex-US (global) markets based on his paper “Enhanced Momentum Strategies"²

Results

Table 1: Breakdown of factor premiums

Original Paper Table	Reproduced Table

Table 2: Diversification benefits of factors

Original Paper Table	Reproduced Table

Table 3: Added value long- and short-legs

Original Paper Table	Reproduced Table

Table 4: Small versus large caps

Original Paper Table	Reproduced Table

Original Paper Table	Reproduced Table

Original Paper Table	Reproduced Table

Table 5: Robustness to definition of factor legs

Note: In Panel C and Panel D, for the VOL factor, paper uses cap-weighted version, while I have used equal-weighted version. Hence the discrepancy in the results for VOL factor and All (i.e. combined factor) results.

Original Paper Table	Reproduced Table

Original Paper Table	Reproduced Table

Original Paper Table	Reproduced Table

Original Paper Table	Reproduced Table

Table 6: Long versus short legs based on 5x5 portfolios

Note: Since I do not have access to 5x5 size volatility factor portfolios used by the authors, a similar risk metric of 5x5 size/risk sorted portfolios is used from the Kenneth French data library. Hence the discrepancy in the VOL factor and All (i.e. combined factor) results.

Original Paper Table	Reproduced Table

Table 7: Portfolio optimization results for 5x5 portfolios

Note: This table utilizes the same data used in Table 6. Due to different data source for 5x5 size-volatility factor portfolios (VOL), the below results on maximum sharpe ratio optimization are not exactly the same as mentioned in the paper. However, the conclusions remain indifferent (i.e. major allocation is for the long leg of factors across different sizes) in both separate and additive optimization.

Original Paper Table	Reproduced Table

Table 8: Volatility and long versus short legs

Original Paper Table	Reproduced Table

Table 10: HML and long versus short legs

Original Paper Table	Reproduced Table

Original Paper Table	Reproduced Table

Original Paper Table	Reproduced Table

Original Paper Table	Reproduced Table

Table 11: Volatility and long versus short legs

Original Paper Table	Reproduced Table

Original Paper Table	Reproduced Table

Original Paper Table	Reproduced Table

Original Paper Table	Reproduced Table

Figure 2: The long and short of equity factors

Original Paper Figure	Reproduced Figure

Figure 3: Diversification benefit of equity factors

Original Paper Figure	Reproduced Figure

Figure 4: Subperiod results

Original Paper Figure	Reproduced Figure

Original Paper Figure	Reproduced Figure

Figure 5: Drawdowns of factor portfolios

Original Paper Figure	Reproduced Figure

Table A1: Breakdown of q-factor premiums

Original Paper Table	Reproduced Table

Figure A1: Rolling (10-year) correlation between long leg and short leg

Original Paper Figure	Reproduced Figure

Figure A2: Rolling (10-year) alpha of long versus short legs

Original Paper Figure	Reproduced Figure

References