Quantifying nonlinear effects of BRIC and G4 liquidity on oil prices

The paper examines how monetary liquidity from major developed (G4: Eurozone, Japan, UK, US) and emerging (BRIC: Brazil, Russia, India, China) economies influences real crude oil prices, particularly amid significant events such as the 2007–2009 Great Recession and the COVID-19 pandemic. Motivated by unprecedented global monetary expansions and differing policy responses across countries, the study focuses on the financialization channel of commodities, hypothesizing that liquidity shocks—beyond traditional demand and supply fundamentals—materially affect oil price dynamics. The authors aim to quantify and compare the impacts of BRIC and G4 liquidity on real oil prices, allowing for structural breaks and regime changes that may alter these relationships.

Two primary strands of prior work are discussed. (1) Fundamentals-based studies attribute oil price movements to demand and supply shocks, linking price increases to emerging market growth and highlighting the effects of negative-demand and positive-supply shocks, particularly around crises. (2) Financialization-based studies argue that oil prices are increasingly driven by financial factors and global liquidity rather than fundamentals, emphasizing portfolio flows and monetary conditions. Empirically, past work examined global liquidity’s impact on broad commodities and oil but rarely differentiated liquidity sources between developed and developing economies or accounted for regime shifts. Notably, Ratti and Vespignani (2013, 2015) found BRIC liquidity shocks significantly increase real oil prices, with limited effects from developed-country liquidity, but they did not model regime dependence. The present study advances this literature by employing Markov-switching VAR models to capture nonlinear, regime-dependent effects, including during the COVID-19 period.

Data: Monthly data from January 1999 to December 2020. Real crude oil prices are WTI nominal prices deflated by U.S. CPI (base year 1999). Liquidity proxies are real M2 aggregates: G4 liquidity sums Eurozone, Japan, UK, and U.S. M2; BRIC liquidity sums Brazil, Russia, India (L2 proxy), and China M2, all converted to USD using IMF IFS exchange rates and then deflated by U.S. CPI. All series (real oil prices and real M2) are seasonally adjusted via X-13-ARIMA. Additional variables include global short-term interest rates (IR), global industrial production (IP), and global oil production (GP). Global IR and IP are constructed as first principal components of country-specific central bank discount rates and IP growth rates for the eight economies (G4+BRIC). GP is the average monthly growth in global crude oil production (EIA). Returns and liquidity growth are first log differences: oil return R_t = Δln(P_t); G4 Liq_t = Δln(G4 M2_t); BRIC Liq_t = Δln(BRIC M2_t). The VAR state vector is y_t = [Δln(P_t), IR_t, IP_t, GP_t, Δln(G4 M2_t), Δln(BRIC M2_t)]. Stationarity is confirmed using ADF and Phillips-Perron tests. Models: (1) Single-state VAR(p) estimated via OLS; lag length selected by information criteria (SIC favors p=1). (2) Markov-switching VAR models with varying regime dependence in intercepts, autoregressive coefficients, and heteroscedasticity (MSI, MSIA, MSIH, MSIAH). Model selection uses LR tests and AIC/HQ/SIC; MSIH(3,1) is chosen by parsimony (HQ/SIC): three regimes with regime-dependent intercepts and covariance matrices, but a time-invariant VAR(1) coefficient matrix. Estimation uses maximum likelihood via the EM algorithm (Hamilton filter and smoother) to obtain smoothed regime probabilities and parameters. Identification for impulse responses uses regime-specific Cholesky decompositions of covariance matrices. Impulse response functions (IRFs) are computed state-dependently over 30 months with 95% confidence intervals via Monte Carlo (10,000 draws).

- Single-state VAR(1): The lagged BRIC liquidity shock significantly increases real oil returns; G4 liquidity is not significant. Table 4 shows BRIC Liq (t-1) coefficient on OIL is 1.685 (SE 0.643, p<0.01), while G4 Liq (t-1) is insignificant. - Markov-switching MSIH(3,1): Three regimes are identified by volatility—Regime 1 (low volatility), Regime 2 (high volatility), Regime 3 (crisis). The coefficient matrix is constant across regimes, but intercepts and covariances differ. In this framework, both BRIC Liq (t-1) and G4 Liq (t-1) significantly raise real oil returns (Table 6: G4 Liq (t-1) coefficient 0.680, SE 0.396, p<0.10; BRIC Liq (t-1) coefficient 1.232, SE 0.489, p<0.05). IP(t-1) and GP(t-1) are generally not significant determinants of real oil returns, suggesting fundamentals are not primary drivers in this setup. - Regime characteristics (Table 7): Regime 1 stayer probability 0.917 (avg duration ~12.1 months), characterizing post-summer 2009 recovery and post-June 2020 recovery; Regime 2 stayer probability 0.919, capturing pre-subprime periods; Regime 3 is a crisis state with extreme volatility, covering the 2008–2009 financial crisis and March 2020 COVID-19 shock (with additional spikes in Sep 2011 and Aug 2015). Volatility of OIL increases markedly from Regime 1 to Regime 3 (e.g., diagonal variance elements ~40.8, 64.6, 609.8 respectively). - Nonlinear IRFs (MSVAR): G4 liquidity shock effects on real oil returns are state-dependent—approximately 0.648% (low volatility), 0.812% (high volatility), and 1.290% (crisis), indicating nearly double the impact during crisis vs. normal periods. BRIC liquidity shocks have even stronger effects, with the crisis-regime impact roughly three times larger than in the other regimes. Across regimes, BRIC liquidity shocks generate larger oil price responses than G4 liquidity shocks. - Spillovers: Evidence of liquidity transmission from G4 to BRIC (lagged G4 Liq significantly affects BRIC Liq), but not vice versa, consistent with advanced-to-emerging liquidity spillovers. - Correlation patterns: During high-volatility regimes, real oil returns are positively correlated with both BRIC and G4 liquidity; real oil returns are negatively correlated with global oil production. - Overall, Markov-switching models uncover significant, nonlinear, regime-dependent impacts of both BRIC and G4 liquidity on real oil prices, which a single-state VAR misses for G4.

The findings support the hypothesis that monetary liquidity plays a significant, nonlinear role in determining real oil prices, reinforcing the financialization perspective. While traditional demand (IP) and supply (GP) proxies are not primary drivers in the preferred model, liquidity shocks from both BRIC and G4 significantly increase oil prices, with magnified effects during crisis regimes. The identification of three regimes—low volatility, high volatility, and crisis—demonstrates that structural breaks and macro-financial stress materially alter the sensitivity of oil prices to liquidity. Policy-wise, simultaneous expansionary policies in developed and emerging markets (e.g., during COVID-19) can channel large liquidity shocks into energy price inflation, highlighting the relevance of monitoring global monetary aggregates for commodity market stability. The stronger and more persistent effects of BRIC liquidity and the larger crisis-regime impacts emphasize the role of emerging-market monetary expansions and global spillovers in commodity price dynamics.

The study shows that BRIC and G4 liquidity shocks exert significant, nonlinear effects on real oil prices. A single-state VAR indicates BRIC liquidity increases oil prices, while MSVAR reveals significant impacts from both BRIC and G4 liquidity once regime switches are allowed. Three regimes—low volatility, high volatility, and crisis—are identified, with crisis periods exhibiting amplified sensitivities: G4 liquidity effects nearly double relative to normal times, and BRIC liquidity effects are roughly triple. BRIC shocks generally have larger impacts than G4 shocks. These results contribute to understanding the financialization of commodities and the transmission of global monetary conditions to energy prices, especially during crises. Potential future research directions include forecasting frameworks incorporating regime-switching liquidity effects, exploration of alternative liquidity measures and financial channels, and disaggregated analyses across commodity classes and regional oil benchmarks.

The analysis does not provide forecasts of energy price dynamics, as explicitly noted by the authors. The preferred MSIH(3,1) specification, chosen for parsimony, imposes time-invariant autoregressive coefficients across regimes with regime-dependent intercepts and covariances; identification relies on Cholesky decomposition. Results are based on monthly data, specific liquidity proxies (real M2 aggregates), and principal-component-based global factors, which may affect generalizability to alternative measures or higher-frequency dynamics.