Detecting statistically significant changes in connectedness: A bootstrap-based technique

Connectedness quantifies the extent of interlinkages within economies or markets based on a network approach. Connectedness is measured by the Diebold–Yilmaz spillover index, and abrupt increases in this measure are thought to result from major events. However, formal statistical evidence of events causing such increases is scant. We develop a bootstrap-based technique to evaluate the probability that the value of the spillover index changes at a statistically significant level following an exogenously defined event. We further show how our procedure can detect the dates of unknown events endogenously. The results of a simulation exercise support the effectiveness of our method. We revisit the original dataset from Diebold and Yilmaz’s seminal work and obtain statistical support that the spillover index increases quickly in the wake of adverse shocks. Our methodology accounts for small sample bias and is robust with respect to modifications of the pre-event period and forecast horizon. •The Diebold–Yilmaz (DY) spillover index is widely used to measure connectedness.•Formal evidence that spikes in the spillover index are due to adverse shocks is scant.•Our procedure statistically links adverse shocks to spikes in the DY spillover index.•Our method is empirically verified, and its effectiveness is shown via a simulation.•Our technique accounts for small sample bias and detects unknown events endogenously.