メインコンテンツへ移動

Macro by Mark の Cookie.ログイン、言語、安全性のために必須 Cookie を使用します。許可いただいた場合は、設定の保存、分析、Meta(Facebook)ピクセルによる広告計測のために任意の Cookie も使用します。詳しくはCookie ポリシーをご覧ください。

Macro by MarkMacro by Mark
  • ホーム
読み込み中…
読み込み中…
← モデル
Macro概念政策モデル

Network spillover Model

Network-based decomposition of forecast-error variance into directional spillovers across sectors or markets.

概要グラフ数式参考資料

How much of the forecast uncertainty in one market, sector, or country is caused by shocks originating elsewhere, and how does that connectedness change over time?

Background

Financial crises reveal linkages that appear invisible in calm times. Before 2008, individual bank risk metrics looked fine; the problem was the network of exposures linking them. Diebold and Yilmaz (2009) proposed a framework that measures interconnectedness directly from observable time series, without requiring bilateral exposure data. The idea: estimate a VAR on NNN variables (country equity returns, sector volatilities, sovereign CDS spreads), compute the generalized forecast-error variance decomposition (GFEVD), and read the off-diagonal shares as spillover intensities. The total spillover index aggregates these into a single number that tracks system-wide connectedness over time.

The mechanics work in two steps. First, estimate a VAR(ppp) on NNN stationary series and compute the HHH-step-ahead generalized forecast-error variance decomposition (Pesaran and Shin, 1998). Unlike Cholesky-based decompositions, the generalized version does not depend on variable ordering - critical when there is no theoretical ordering among, say, 20 country equity markets. The GFEVD produces an N×NN \times NN×N matrix θH\theta^HθH where θijH\theta^H_{ij}θijH​ is the share of variable iii's HHH-step forecast-error variance attributable to shocks in variable jjj. Row-normalize so each row sums to 1. The off-diagonal elements are spillovers; the diagonal elements are own-variance shares.

The total spillover index is the average off-diagonal share: S=100⋅N−1∑i≠jθijHS = 100 \cdot N^{-1} \sum_{i \neq j} \theta^H_{ij}S=100⋅N−1∑i=j​θijH​. This single number ranges from 0 (all variables are isolated - own shocks explain everything) to 100 (all forecast uncertainty comes from cross-variable shocks). Rolling-window estimation produces a time series of StS_tSt​ that spikes during crises: the 2008 Global Financial Crisis, the 2010 - 12 Euro-area debt crisis, the 2020 COVID crash. Directional spillovers - how much variable iii transmits to versus receives from others - identify net transmitters (systemically important nodes) and net receivers (vulnerable nodes).

The framework is used at the IMF for financial surveillance (Global Financial Stability Report), at the BIS for systemic-risk monitoring, at the Federal Reserve for tracking cross-asset contagion, and in academic research on international business-cycle transmission, oil-price spillovers, and climate-risk propagation. Barunik and Krehlik (2018) extended it to frequency-domain spillovers, decomposing connectedness by short-run versus long-run horizons. Demirer, Diebold, Liu, and Yilmaz (2018) scaled it to networks of 100+ global banks.

Connections to explore

Math ↗The structure and its conditions, step by step.References ↗Academic sources and guidance for using the model.

How the Parts Fit Together

θijH\theta^H_{ij}θijH​

Pairwise directional spillover

The share of variable iii's HHH-step forecast-error variance attributable to shocks in variable jjj. Row-normalized to sum to 1.

Inputs are NNN stationary time series observed at the same frequency over TTT periods. In the original application: realized volatilities of equity indices for major economies. In other applications: credit spreads, bilateral trade flows (differenced), sector-level output growth, or bank-level CDS spreads. The series must be stationary; log-differencing, realized volatility, or HP-filter deviations are common transformations. The data enter a VAR(ppp) where ppp is selected by AIC/BIC.

The model estimates the VAR yt=c+A1yt−1+⋯+Apyt−p+ety_t = c + A_1 y_{t-1} + \cdots + A_p y_{t-p} + e_tyt​=c+A1​yt−1​+⋯+Ap​yt−p​+et​, then inverts it into its moving-average representation yt=∑h=0∞Φhet−hy_t = \sum_{h=0}^\infty \Phi_h e_{t-h}yt​=∑h=0∞​Φh​et−h​ where Φh\Phi_hΦh​ are N×NN \times NN×N MA coefficient matrices. The generalized FEVD decomposes the HHH-step forecast error yt+H−y^t+H∣ty_{t+H} - \hat{y}_{t+H|t}yt+H​−y^​t+H∣t​ into contributions from each variable's innovations. The (i,j)(i,j)(i,j) element of the unnormalized GFEVD is θ~ijH=σjj−1∑h=0H−1(ei′ΦhΣej)2/∑h=0H−1ei′ΦhΣΦh′ei\tilde{\theta}^H_{ij} = \sigma_{jj}^{-1} \sum_{h=0}^{H-1} (e_i' \Phi_h \Sigma e_j)^2 / \sum_{h=0}^{H-1} e_i' \Phi_h \Sigma \Phi_h' e_iθ~ijH​=σjj−1​∑h=0H−1​(ei​′Φh​Σej​)2/∑h=0H−1​ei​′Φh​ΣΦh​′ei​. Row-normalizing gives θijH=θ~ijH/∑kθ~ikH\theta^H_{ij} = \tilde{\theta}^H_{ij} / \sum_k \tilde{\theta}^H_{ik}θijH​=θ~ijH​/∑k​θ~ikH​.

Rolling-window estimation repeats this for overlapping subsamples of width WWW (typically 100 - 200 observations for daily data, 40 - 60 for quarterly). Each window produces a spillover matrix θtH\theta^H_tθtH​, a total spillover index StS_tSt​, and directional spillover profiles for each variable. The resulting time series of StS_tSt​ is the primary output: a real-time connectedness tracker. Alternative approaches - Bayesian TVP-VARs or DCC-GARCH-based spillovers - avoid the window-width choice but add computational complexity.

Assumptions

Testable

Stationarity

All NNN series are covariance-stationary, or transformations have been applied to make them stationary.

If violated:Non-stationary series produce spurious VAR estimates and meaningless FEVD shares. Unit-root tests (ADF, KPSS) should precede estimation.

Testable

Correct VAR lag order

The VAR has lag order ppp chosen by AIC/BIC, and this correctly captures the dynamic structure.

If violated:Misspecified lag order distorts the MA coefficients Φh\Phi_hΦh​, which are the basis for the GFEVD. Residual autocorrelation tests should be run.

Testable

Gaussian innovations (for GFEVD)

The generalized FEVD is exact under Gaussian innovations. For non-Gaussian data, it provides an approximation.

If violated:With heavy-tailed or skewed innovations (common in financial data), the GFEVD is approximate. Bootstrapped confidence intervals mitigate this.

Maintained

Stable rolling-window width

The window width WWW is large enough to estimate the VAR reliably but short enough to capture time variation.

If violated:Too-short windows produce noisy spillover time series with spurious spikes. Too-long windows smooth out genuine crisis-period jumps. Sensitivity to WWW should be reported.

Testable

No structural breaks within the window

Within each rolling window, the VAR parameters are approximately constant.

If violated:A structural break inside a rolling window (e.g., a crisis onset) mixes pre- and post-break data, producing a blended VAR that represents neither regime accurately.

Maintained

Order-invariant identification

The GFEVD (Pesaran-Shin) does not depend on the ordering of variables in the VAR, unlike the Cholesky FEVD.

If violated:Not a failure per se, but the GFEVD rows do not sum to exactly 1 before normalization. The normalization is a convenience, not a structural result. This means spillover shares are relative, not absolute.

Applications

The IMF's Global Financial Stability Report uses the Diebold-Yilmaz framework to track cross-country financial connectedness. A rolling-window spillover index on sovereign CDS spreads of 20+ countries reveals when contagion risk is elevated. During the 2010 - 12 euro-area crisis, the total spillover index spiked from around 40 to over 70, driven by directional spillovers from Greece, Ireland, and Portugal to Spain and Italy. This time-varying measure informs the IMF's surveillance discussions about systemic-risk buildup.

Central banks monitor cross-asset connectedness for macro-prudential policy. The Federal Reserve Bank of New York computes spillover indices across US equity sectors, Treasury volatility, credit spreads, and currency markets. A rising spillover index signals that asset classes are moving together more than usual - a precursor to financial stress. This complements traditional financial-conditions indices by providing directionality: which markets are transmitting stress and which are receiving it.

Academic research on oil-price spillovers uses the framework to measure how crude-oil shocks propagate to equity markets, exchange rates, and real economic activity across countries. Diebold and Yilmaz (2012) showed that oil-price volatility spillovers to US equities increased dramatically after 2003, coinciding with the financialization of commodity markets. This finding influenced the debate about whether financial speculation in commodities amplifies real-economy volatility.

The framework breaks down when NNN is large relative to TTT (or relative to the rolling window WWW). A VAR with 50 variables and 2 lags has 50×(50×2+1)=505050 \times (50 \times 2 + 1) = 505050×(50×2+1)=5050 parameters. With a 200-observation rolling window, the VAR is severely overparameterized. Solutions include LASSO-VAR (penalized estimation), factor-augmented approaches (compress NNN variables into fewer factors before running the VAR), or the elastic-net VAR of Demirer et al. (2018). The framework also does not identify causal mechanisms: a high spillover from variable iii to jjj means iii's innovations help forecast jjj's future, but this could reflect common exposure to a third factor rather than direct transmission.

Extensions and open questions

Variants

  • Frequency-domain spillovers (Barunik-Krehlik, 2018) - decomposes connectedness into short-run and long-run components
  • TVP-VAR spillovers (Antonakakis et al., 2020) - replaces rolling windows with time-varying parameter VAR to avoid window-width dependence
  • Quantile connectedness (Ando et al., 2022) - measures spillovers at different quantiles of the distribution (tail risk)
  • LASSO-VAR spillovers (Demirer et al., 2018) - regularized VAR for high-dimensional networks
  • Partial spillovers (Greenwood-Nimmo et al., 2021) - conditions on common factors to isolate bilateral linkages from common shocks

Open questions

  • Whether GFEVD-based spillovers measure causal transmission or merely statistical association driven by common factors
  • How to construct valid confidence intervals for the rolling-window spillover index, given the overlapping-sample and estimation-error complications
  • Whether frequency-domain decomposition or quantile-domain decomposition is more informative for financial-stability monitoring

Keep going

Read next

Continue with a related concept, model, or example.

Agent-based model

Supply-Chain / Network ABM

a contrasting take on Network Spillover (Diebold-Yilmaz)

Contrast it →
Theory model

IS-LM Model

a well-connected theoretical model two steps from here

New angle: theoretical model →
Macro by Mark
ヘルプ
プライバシーポリシー利用規約倫理とコンプライアンス

© 2026 Mark Jayson Nation LLC