Work

Identifying Market Criticality: Shock Bounds, Propagation and State Selection

Year: 2026
Type: preprint
Abstract: Which feedback networks could have produced a given set of market losses? For a nonnegative linear network, coordinate bounds on the initiating losses yield a sharp interval for the spectral radius. Repeated episodes restrict the network further. In six synthetic episodes, allowing own-market feedback changes the upper bound on the full spectral radius from 0.9214 to 0.9217, while the lower bound on the cross-market cycle falls from 0.8758 to 0.3684. Two networks on opposite sides of 0.90 show how precisely initiating losses must be measured to resolve that threshold. The August 2015 dislocation in two S&P 500 funds illustrates how a traded benchmark can discipline the scalar bound and where a counterfactual assumption enters. For event counts, an exponential Hawkes process and a latent-shock Cox process can match the mean, the covariance at every horizon and the long-window third cumulant. With 250-millisecond windows, a simulated 5% test distinguishes the two specified processes in 92.6% of two-day Hawkes samples. Independent event insertions give a separate structural prediction. These results distinguish what prices and counts identify about propagation from the evidence required to establish that markets adapt toward a stability boundary.
Language: English
Open Access status: green
OpenAlex ID: W7236597629
Cites:
Cited by:
Related to: 10