International price volatility transmission and structural change: a market connectivity analysis in the beef sector

The study investigates how international beef prices connect with and transmit to domestic retail beef prices in net-importing countries. Motivated by disruptions in meat value chains (e.g., BSE, COVID-19) and the paucity of cross-border beef price transmission research, the authors address limitations of traditional ECM approaches that cannot capture time-varying linkages. They apply dynamic conditional correlation methods to quantify evolving market connectivity and explicitly account for structural breaks that may bias time-series inference. The purpose is to uncover the direction, timing, and strength of price and volatility spillovers between global and local beef markets, and to inform policy on how to mitigate external shocks to domestic food security.

Prior agricultural price transmission work largely focuses on domestic vertical chains and often uses ECMs, which assume constant relationships and cannot visualize time-varying linkages. DCC-based approaches are common in finance and energy for modeling dynamic comovements and volatility spillovers but are rare in agricultural markets. A few cereal market studies (e.g., Guo and Tanaka 2019; Tanaka and Guo 2020) have applied DCC to world–local linkages; Guo and Tanaka (2020) is among the only works examining international beef price passthroughs. Several beef market studies considered structural breaks within domestic chains, and Cappiello et al. (2006) introduced structural changes into DCC-type models, yet prior DCC applications to agricultural price transmission generally did not test for breaks in time-varying correlations. This paper fills gaps by focusing on cross-border beef markets, employing DCC-GARCH with structural breaks, and developing AR models for DCCs with break dummies.

Data: Monthly international and domestic retail beef prices for nine net beef-importing countries (Azerbaijan, Georgia, Japan, Kazakhstan, Kyrgyzstan, Tajikistan, Tunisia, UK, USA) from January 2006 to May 2020, expressed in USD. Domestic data sources include FAO GIEWS (several countries), Japan’s Agriculture & Livestock Industries Corporation, UK Office for National Statistics, and US Bureau of Labor Statistics; exchange rates from FRED. Prices are seasonally adjusted using X-13-ARIMA. Returns are first differences of log prices. Preliminary analysis: Descriptive statistics document non-normality, negative skewness (mostly), fat tails, and low positive unconditional correlations with the international price. Stationarity is confirmed for returns with ADF and KPSS tests. To account for potential breaks, Zivot–Andrews (1992) and Perron (1997) unit root tests with one endogenous break are also applied, indicating stationarity in first differences and identifying break dates around 2007–2009 and 2014–2016. GARCH-type modeling with structural breaks: For each series, an AR(1)-GARCH-type model is estimated to capture conditional mean and variance with potential asymmetry: standard GARCH(1,1), EGARCH(1,1), and GJR-GARCH(1,1). Model selection uses BIC and residual diagnostics. Following Lamoureux and Lastrapes (1990), structural breaks in mean and volatility are incorporated. Breaks in mean are detected via Bai and Perron’s (2003) sequential test on AR models; residuals are then transformed to identify variance breaks. Corresponding dummy variables for detected break dates enter mean and/or variance equations. Granger causality via CCF: Using standardized residuals and squared residuals from the selected univariate GARCH models, Hong’s (2001) non-uniform weighting cross-correlation function tests for Granger causality-in-mean and causality-in-variance between international and domestic prices over horizons M = 5, 10, and 15 months. Time-varying correlations and their breaks: Bivariate DCC-GARCH models estimate dynamic conditional correlations between international and domestic returns. Four specifications are considered and selected by BIC: standard DCC, asymmetric DCC (A-DCC), generalized DCC (G-DCC), and asymmetric generalized DCC (AG-DCC). Parameters are estimated by Gaussian QMLE with BFGS optimization under the a + b < 1 stationarity constraint. Structural breaks in the estimated DCC series are then identified using Bai–Perron tests. To quantify regime changes in correlation dynamics, AR(1) models with structural break dummies are estimated for each DCC series, assessing how break episodes alter the level of correlation. Diagnostics: Ljung–Box and ARCH-LM tests confirm adequate fit and absence of residual autocorrelation and remaining ARCH effects in the univariate models; Breusch–Godfrey and ARCH tests support the AR(1) specifications for DCC series.

- Structural breaks in returns: Bai–Perron tests indicate structural changes in the mean and/or volatility of domestic beef price returns for several countries, largely after the 2007–2009 crisis period. Breaks are found for Azerbaijan (mean and variance), Georgia (mean), Japan (variance), Kazakhstan (mean), Kyrgyzstan (mean), Tajikistan (mean), and the UK (variance). No breaks are detected for the international price, Tunisia, or the USA in Table 3. - Granger causality-in-mean: Unidirectional causality from the international price (IP) to domestic prices is found for Georgia, Japan, Kazakhstan, Tajikistan, and the USA, typically with short lags (1–5 months). No reverse causality from domestic to international prices is detected. - Granger causality-in-variance: No significant variance causality is found between IP and domestic prices for Azerbaijan and Tunisia. Variance spillovers from IP to domestic prices are detected for Kyrgyzstan and the UK (without mean causality), with longer lags for the UK (10 months) and the USA (15 months). Overall, Georgia, Tajikistan, and the USA show IP→domestic causality in both mean and variance; Japan and Kazakhstan show mean-only spillovers; Kyrgyzstan and the UK show variance-only spillovers. - Dynamic correlations (DCC): Market linkages are generally weak. Mean DCCs range from -0.099 (Japan) to 0.210 (Kyrgyzstan); medians from -0.109 (Japan) to 0.213 (Kyrgyzstan). Extremes include a maximum DCC of 0.999 (USA) and a minimum of -0.755 (Kazakhstan). DCC volatility is highest for Azerbaijan (Std. Dev. 0.258) and most stable for Georgia (Std. Dev. 0.062). Some pairs (Japan, Kazakhstan, UK) exhibit negative DCCs at times, consistent with lagged adjustment. - Crisis-related comovement: DCCs often spike and fluctuate sharply around the 2007–2008 global food/financial crisis and again around 2014–2015, indicating temporary strengthening of global–local connectivity during turmoil. - Breaks in DCCs: Bai–Perron tests identify structural breaks in DCCs for most pairs, clustered in four periods: 2007–2008 (crisis onset), 2009–2012 (post-crisis adjustment), 2014–2015, and 2017–2018. AR(1) models with break dummies show significant positive or negative shifts in correlations at these dates, varying by country. - Policy implication: Given generally weak peacetime connectivity but intensified linkages during global crises, temporary trade measures (e.g., quotas or taxes) could shield domestic markets during emergencies while avoiding efficiency losses in normal times.

The findings address the core question of how international beef prices connect to domestic markets by showing that spillovers are predominantly unidirectional from global to local markets and that these linkages are time-varying and sensitive to structural breaks. Most countries exhibit weak average dynamic correlations, implying limited routine transmission. However, during periods of global stress (2007–2008; 2014–2015), correlations rise and become more volatile, indicating heightened synchronization and faster transmission of shocks. Causality tests refine this picture: some countries (Georgia, Tajikistan, USA) are influenced in both price levels and volatility by the international market; others experience only mean (Japan, Kazakhstan) or only variance (Kyrgyzstan, UK) spillovers; Azerbaijan and Tunisia show no significant links. No domestic market Granger-causes the international price, underscoring the dominance of global pricing in these relationships. The structural break analysis demonstrates that ignoring breaks biases inference about market connectivity. Break dummies significantly shift DCC levels, confirming that regime changes materially affect correlations. Comparisons with wheat suggest beef markets are less tightly connected globally, potentially due to more flexible supply adjustments (e.g., slaughtering) versus climate-dependent crops. The weak routine linkage but crisis-sensitive transmission supports targeted, temporary policy interventions during global disruptions to protect domestic stability.

This paper provides the first application in meat or grain price transmission research of DCC models augmented with structural break dummies in AR specifications, offering a clearer visualization of time-varying global–local linkages. Using seasonally adjusted monthly data (2006–2020) for nine beef-importing countries and a suite of GARCH/DCC models, the study shows: (1) several domestic markets experienced structural breaks in mean and/or variance post-2010; (2) international prices unidirectionally Granger-cause domestic prices—both in mean and variance for Georgia, Tajikistan, and the USA—with no evidence of reverse transmission; and (3) dynamic correlations are generally low but intensify around crisis periods, with identifiable structural shifts. Policy-wise, temporary trade restrictions during global emergencies may buffer domestic markets, whereas such measures are not warranted in normal times. Methodologically, incorporating structural breaks improves estimation reliability for volatility and correlation dynamics. Future research could extend to additional countries, higher-frequency data, alternative distributional assumptions, and structural models linking self-sufficiency, policy regimes, and connectivity.