Skip to content

Home

Jump to bottom
Arman Hassanniakalager edited this page Jan 5, 2023 · 23 revisions

Welcome to the knockoff_index wiki!

This page needs revising.

What?

This project aims to generate positive alpha by using an econometric tool –knockoff filter as in Barber and Candes (2015). The practitioner shall set a fundamental factor (say book-to-market ratio) to select 50 US stocks as candidates to construct active portfolios. The portfolios are based on analysis of the past 10 years performance and rebalanced annually. The performance of portfolios are measured by excess return (relative to effective funds rate) and are tracked over 1996-2020.

Why?

  • This strategy considers both fundamentals and market performance.
  • This is a low-cost strategy to create a positive return in the stock market.
  • This strategy is reproducible for non-US stock markets. If you have such data and are keen please get in touch.
  • The investment strategy can be linked to certain investment factors or goals. For instance, the same approach could be used to consider firms with the highest ESG scores.

How?

The knockoff filter develops a regression model by controlling for false discoveries. This project uses four US indexes – DJIA, S&P 500, NASDAQ, and Russell 1000 – to pick an optimal subset of qualified stocks. It is up to the practitioner to define the qualified stocks. The backtest results below are based on qualified stocks defined as 50 US stocks with the highest book-to-market. The stocks picked by the knockoff filter are used to build optimised portfolios with minimum portfolio variance. All assets are bought on the first Friday of a calendar year and sold on the last Friday of the same year. Flowchart of analysis

Deployment

See the codes and the accompanying read me file for a step-by-step guide to replicate the results.

Portfolios

  • You can now access 2022 portfolios for Russell 1000 here, S&P500 here, and DJIA here

  • Historical portfolios for calendar years from 1996 to 2021 are accessible in the folder "Portfolios".

  • A guidance for reading portfolios is provided in the folder "Portfolios" and in this ReadMe file

** All portfolios are in CSV format.

Backtest results

The figure below shows the value of a portfolio worth $100 portfolio on 31 December 1995 over backtest period. All portfolios are equal-weight. portfolio backtested vs benchmarks

Clone this wiki locally