One measure of the crowdedness of trades in a portfolio is centrality. Centrality is a concept of network analysis that measures how similar one institution’s portfolio is to its peers by assessing its importance as a network node. Empirical analysis suggests that [1] the centrality of individual portfolios is negatively related to future returns, [2] mutual fund holdings become more similar when volatility is high, and [3] the centrality of portfolios seems to reflect lack of information advantage. This evidence cautions against exposure to crowded trades that rely upon others’ information leadership or are motivated by widely publicized persuasive views.

Mitali, Shema (2019), “Common Holdings and Mutual Fund Performance.”

The below are excerpts from the paper. Headings, bracketed/cursive text and emphasis have been added.
The post ties in with the SRSV summary on endogenous market risk.

What is centrality?

“In a network where funds are connected through portfolio overlap, degree centrality of each fund represents the level of similarity with peers.”

“With mutual funds holdings data, I use network analysis to obtain cross-sectional variations in holdings similarity. More specifically, I use the concept of centrality that graph theory defines as the most important nodes in a network. In social network analysis, central nodes are often associated to influential agents. Since the network analysis based on mutual fund holdings does not rely on social connections, a central fund will be defined as a mutual fund that is highly similar, in terms of holdings, to its peers.”

“The starting point of this analysis is an adjacency matrix. This square matrix is the representation of a given graph [as shown below].”

“The centrality measure will identify mutual funds that have high or low holdings similarity with peers. It measures a node’s importance in the network. A highly central fund has many positions overlapping with its peers and hence is less likely to have an informational advantage about particular firms according to the hypothesis tested in this paper. In our network, the (weighted) degree centrality of a fund will be the sum of its overlaps with each neighboring node.”

“[The figure below] shows, as an example, the first quarter of the sample where red nodes represent mutual funds connected through portfolio overlap. The bigger the red node is, the more common holdings it has.”

Evidence of the negative relative relation between centrality and fund returns

“Using quarterly mutual fund holdings data from 1980:Q1 to 2016:Q4, I implement network analysis and connect actively managed U.S. equity mutual funds through their portfolio overlap (i.e., the number of stocks in common). The degree centrality of each mutual fund in the network at each quarter is measuring its holdings similarity with its peers.”

“Mutual fund degree centrality has a negative and statistically significant effect on mutual fund performance…Holdings similarity leads to lower abnormal fund returns…This main result holds across alternative holdings similarity measure (e.g., eigenvector centrality, distance) or when computed within style cluster networks.”

A portfolio based on stocks owned by low vs. high degree centrality funds yields abnormal returns of 7% per year…A one standard deviation increase in degree centrality leads to lower abnormal fund performance in the next quarter of roughly 20 bps… This main result cannot be explained by other fund characteristics [such as] size, flows, or other measures of activeness relative to a benchmark such as active share or tracking error.”

“I compute network degree centrality for each mutual fund within style clusters. I find a negative relationship of degree centrality on fund performance with the same magnitude as in the main results.”

Centrality rises in turbulent times

“The negative association between holdings similarity and fund performance widens in volatile markets. In uncertain times, mutual funds move towards their benchmark due to asset management constraints.”

“I study how holdings similarity changes in periods of uncertainty. I find that degree centrality increases in a volatile phase, defined by quarters when the Volatility Index (VIX) is in the top decile. This suggests that managers respond…by moving closer to their peers when prices are volatile. Moreover, the negative relationship between degree centrality and fund performance is amplified in these phases. When facing volatility, mutual fund managers create negative price pressure on common assets. This creates spillovers on connected mutual funds. Thus, mutual funds with high degree centrality exhibit lower cumulative abnormal returns.”

“This…contributes to the growing [evidence] of the adverse effects of benchmarking. Due to agency frictions, contracts link compensation to performance relative to a benchmark. With volatile prices, fund managers are exposed to greater risks when being mismatched to the benchmark. Thus, they move towards the index in uncertain times and exacerbate price distortions.”

The informational advantage hypothesis of funds performance

“This paper provides new evidence of the informational advantage hypothesis as a driver of fund performance.”

“To investigate if degree centrality relates to informational advantage, I use a change in regulation as a quasi-natural experiment. In May 2004, the Securities and Exchange Commission (SEC) required more frequent portfolio disclosure from mutual funds… Results show that mutual funds with low degree centrality had their performance decreased by -6.7% in the year 2004. This finding is consistent with a coordination in information acquisition and asset purchases. Fund managers’ initial informational advantage disappears as other fund managers can learn about the same assets owned by the best performing managers, once their holdings are disclosed more frequently.”

SHARE
Previous articleTreasury basis and dollar overshooting
Ralph Sueppel is founder and director of SRSV, a project dedicated to socially responsible macro trading strategies. He has worked in economics and finance for over 25 years for investment banks, the European Central Bank and leading hedge funds. At present, he is head of research and quantitative strategies at Macrosynergy Partners.