Identifying fragility for the stock market: Perspective from the portfolio overlaps network

Abstract

In a financial system, the interconnectedness among entities from investing in common assets (portfolio overlaps) is considered an important channel for the propagation of systemic risk because this interconnectedness can facilitate the contagion of fire sales and lead to widespread sales as well as spiral devaluation of assets. This mechanism also applies to the stock market. Particularly, it is responsible for the presence of fragility, which is defined as the potential global instability of stock prices when confronted by a local negative shock. Thus, we propose to quantitatively explore the temporal pattern of the topological features of the portfolio overlaps network (PON) to identify fragility, with the aim of either capturing a precursor for imminent market crashes or confirming further downward falls during market turmoil. Here, PON is the projected monopartite network used to characterise the mutual overlaps between funds pairs, which is induced from the bipartite network of stocks-funds ownerships. Further, we adopt a solid statistical procedure to construct the validated PON by filtering out mildly risky connections that are not statistically robust. Based on an empirical investigation of the Chinese stock market from 2005 to 2018 that witnessed collective unquenchable price slumps, we found that the topology of the validated PON can indeed provide some sensible indicators for identifying market fragility. The results suggest that (i) the average closeness centrality and the critical value of effective infection rate could serve as leading indicators of impending market turmoil, and (ii) the simple degree-based measure for nodes and relative frequency measures for links are actual synchronous indicators that could also explain the source of fragility.