Cookies on Macro by Mark.We use essential cookies for sign-in, locale, and security. With your permission, we also use optional cookies for preferences, for analytics, and for advertising measurement with the Meta (Facebook) Pixel. Read our cookie policy.
Network-based decomposition of forecast-error variance into directional spillovers across sectors or markets.
Graph workspace
The total spillover index is the headline output. Plotted over time, it reveals how system-wide connectedness rises sharply during crises and subsides during calm periods. Annotating crisis events (Lehman collapse, euro-area debt, COVID) connects the abstract index to real-world episodes. This is the visualization that policymakers and financial-stability reports use.
Full-sample total spillover
Average connectedness across the entire sample. Baseline against which crisis-period spikes are measured.
23.28%
percent
Peak spillover
Maximum total spillover index reached during the sample. Identifies the period of highest system-wide connectedness.
25.49%
percent
Peak-to-trough range
Amplitude of connectedness variation. Large ranges indicate pronounced crisis/calm regime differences.
3.56 pp
percentage points
Top net transmitter
Variable that is the largest average net transmitter of shocks to the system: the most systemically important node. In this toy V1's shocks feed the others most strongly, so it is the top transmitter whenever there are cross linkages; with no cross linkage every net spillover is zero and the readout is blank.
1.00
variable index
Cross-variable dynamics: 0.1
How strongly one series' past moves feed into the others; higher means tighter links.
Increasing raises the spillover index across the entire sample. The level of shifts upward.
Innovation correlation: 0.25
How closely surprise moves in different series happen at the same time.
Higher increases the spillover index even without dynamic linkages. Separating contemporaneous from dynamic spillovers becomes harder.
Rolling window width: 100
How many past observations go into each estimate, trading smoothness against quick reaction.
Shorter windows make more responsive to regime changes but noisier. Longer windows smooth the series but delay detection of crisis-period spikes.
How many series are in the network; more links to find, but more to estimate.
More variables produce a denser bottom panel (more net-spillover bars/areas). The top-panel level may increase if additional variables are interconnected.
Crisis intensity multiplier: 2
How much stronger the links between series become during the crisis stretch of the sample.
Higher produces a more dramatic spike in during the crisis window. The peak-to-trough range of the spillover index widens.