- Equity markets have recovered some of their losses from earlier in the year, which raises the specter of crowding in certain segments of the market.
- We tested three approaches for a discretionary fund manager to curb crowding risk, using the MSCI Global Equity Factor Model to assess a stock’s level of crowding.
- We were able to reduce the level of crowding in each approach, but with varying degrees of outperformance and tracking error when compared to the original fund.
Stock markets around the world have heated up this summer. The MSCI ACWI Investable Market Index (IMI) has gained 5% in the third quarter, a partial thaw from its chill when it fell 20% through the first half of the year. One market segment leading the current rally is high-volatility stocks, which, last year, we highlighted as crowded and at risk of a sharp reversal — a scenario that subsequently came to fruition.
With talk of crowds returning, how can investors, particularly discretionary fund managers, manage crowding risk?
Trying to avoid crowds
Before we dive into that, let’s revisit what stock crowding is and why managing it has mattered. We use the stock-crowding measure in the MSCI Global Equity Model to assess crowding. This employs several components to evaluate a stock relative to its own history.
For example, a stock that is heavily shorted, experiencing unusually strong price momentum, volatility or trading activity or is relatively expensive compared to its normal valuation, could be at risk of crowding. An increase in any of these measures against the stock’s five-year historical median would increase the degree of crowding.
Crowded stocks have not been rewarded historically
Stock-crowding factor returns are from the MSCI Global Equity Trading Factor Model and are shown from January 1995 through July 2022.
As shown above, managing this risk has mattered because crowded stocks have historically trailed the market. Crowded stocks, as measured by their factor return, have fallen by almost half over the past 25-plus years (left plot above). In fact, by this measure, the COVID-19-related rally of 2021 was the only favorable year for crowded stocks over the full period (right plot above).
Three potential approaches for crowd control
Given that backdrop, we simulated three approaches to adjust a stock fund’s level of crowding. We use the example of a popular U.S. large-cap growth-oriented fund to show how this could potentially work, as summarized below.
A description of the simulated approaches
|Screen||Exclude crowded stocks and proportionately reweight|
|Replace||Replace crowded stocks with very similar, but uncrowded stocks|
|Control||Reduce fund's overall crowding level by reweighting all stocks|
The first approach (“Screen”) is the simplest but can potentially remove a large share of the fund manager’s stock picks. On average, the fund historically had about 12% of its weight in moderately to very crowded stocks.
The second (“Replace”) is like the first, but swaps out a crowded position with an alternative, less crowded peer. While an actual fund manager might select from a universe of like firms for each of their positions, we use MSCI Peer Similarity Scores to systematically identify peers.1
Lastly, the “Control” approach attempts to reduce crowding risk while preserving all the fund’s other factor exposures, such as its growth and large-cap styles. Existing positions are reweighted, and no new stocks are added. This is the closest to a so-called “quantamental” investment process that blends methods such as optimization and traditional stock picking.2
Varying degrees of success
The results for each approach are shown below, simulated from 2011 through 2022. All three approaches would have reduced crowding risk and generated improved return as compared to the original fund, but produced varying degrees of tracking error.
All three approaches to managing crowding risk improved returns
Active return and tracking error are annualized and relative to the original fund. Data from March 2011 through July 2022. Hypothetical funds are rebalanced quarterly. MSCI Peer Similarity Score history begins in March 2011.
While both the Screen and Control approaches outperformed the original fund by a similar amount, the latter resulted in the smallest deviations across other style factors (turquoise bars shown below), resulting in a significantly lower tracking error. This approach also resulted in the smallest deviations to the original fund industry- and stock-specific views.
Control approach best preserved fund’s exposures
Values are the average differences in exposure against the original fund from March 2011 through July 2022 using the MSCI Global Equity Trading and FaCS models.
Conversely, we found that the Replace approach gave back some of its gains by introducing unintended biases, notably to higher volatility, smaller companies. This also increased tracking error. To further refine this approach, we could, for example, impose more stringent control on other attributes (such as a stock’s capitalization) during peer selection.
Avoiding the summer crowd
Our sample methods of mitigating crowding risk are by no means exhaustive. Additionally, our example fund had a large portfolio with many positions, which provides more flexibility when excluding or reweighting any one of them. More concentrated funds with few positions may face an added challenge. Nonetheless, our results indicated that avoiding the crowds could have yielded benefits for fund managers without making significant changes to the original portfolio.
1MSCI Peer Similarity Scores quantify firm similarities using return correlations, analyst coverage overlaps, news co-mentions, financial fundamentals and 10-K business descriptions. We use a threshold of 75% similarity for this exercise.
2We use an exposure one z-score as a threshold to identify crowded stocks in the Screen and Replace approaches, targeting a portfolio-level exposure of 0.25 less than the original fund for the Control approach.