Measuring Performance Chasing

Sell the winners & buy the losers

June 2024. Reading Time: 10 Minutes. Author: Nicolas Rabener.


  • Performance chasing can be measured via extreme excess returns
  • Abnormal negative returns lead to subsequent outperformance
  • While abnormal positive returns lead to subsequent underperformance


Morningstar recently published a list highlighting the top 10 fund management companies that destroyed the most wealth in the decade ending in 2023, which includes boutique firms like Roundhill Investments and large banks like Barclays. The wealth-destroying products of these asset managers can broadly be categorized into short strategies, Chinese equities, VIX proxies, and thematic investing.

The first three of these categories represent market exposures and calling out ProShares for providing a short product on the Nasdaq seems somewhat unfair as that product is expected to lose when stock markets gain.

However, it is more challenging to overlook the massive wealth destruction of thematic products like the ARK Innovation (ETF), which destroyed $7.1 billion as per Morningstar. In 2021, ARKK represented the typical combination of a seducing narrative, namely investing in disruptive companies, and strong outperformance. The ever-present fund manager, Ms Cathie Wood, marketed the hell out of this product, but is not particularly skilled at investing, so the majority of investors are now sitting on an 80% loss (read An Anatomy of Thematic Investing and Thematic Investing: Thematically Wrong?).

In this research article, we will demonstrate that performance chasing leads to poor returns and how investors can identify such funds.


There are various ways of measuring abnormal returns and some funds like statistical arbitrage funds focus exclusively on identifying securities that exhibit extremely positive or negative returns on a relative basis. We define abnormal returns as when the 12-month excess return is more or less than three times the tracking error from the average 12-month excess return. The tracking error is calculated as the standard deviation of the difference