The COVID-19 pandemic caused a dramatic one-month decline in oil prices in March 2020, reaching negative territory for the first time. The subsequent recovery coincided with expansionary monetary policies implemented globally. This event, along with the varying monetary policy responses of developed (G4) and developing (BRIC) economies, motivated this study to examine the impact of liquidity from these two groups of countries on real oil prices. Previous research has largely focused on demand and supply factors, underestimating the financialization of commodities. This study aims to fill this gap by analyzing the contribution of BRIC and G4 liquidity to real oil prices while controlling for global short-term interest rates, global aggregate demand, and global oil supply. The study uses real M2 as a measure of liquidity, covering the period from January 1999 to December 2020, which includes the COVID-19 outbreak. The use of Markov-switching models allows for the quantification of nonlinear effects of BRIC and G4 liquidity shocks on oil prices, addressing the limitation of previous studies that did not account for structural breaks and regime shifts in oil price dynamics. The study contributes to the existing literature by examining the distinct impact of BRIC and G4 liquidity on oil prices, considering the nonlinear effects during the COVID-19 crisis, and comparing the effects of liquidity shocks with those of demand and supply shocks. The authors posit that excess liquidity may lead to speculative price movements unjustifiable by fundamental supply and demand factors.

Existing literature on oil price dynamics falls into two main streams. The first focuses on demand and supply shocks, explaining price increases before the 2008 crisis as driven by emerging market growth and the subsequent plunge as a result of negative demand and positive supply shocks from COVID-19. The second stream emphasizes the financialization of commodities, arguing that oil prices are determined by financial factors such as the increasing financialization of oil futures markets, highlighting the role of financial liquidity. This study builds on the second perspective by focusing on the impact of monetary aggregates on oil prices. While numerous studies have analyzed the link between global liquidity and commodity prices in general, fewer have focused specifically on oil, and none have differentiated between the sources of liquidity in developed (G4) and developing (BRIC) economies and their impact during the COVID-19 pandemic. Some previous works have suggested that BRIC liquidity positively impacts oil prices, but these lack the ability to incorporate different economic regimes.

This study employs both single-state vector autoregressive (VAR) and Markov-switching vector autoregressive (MSVAR) models to analyze the impact of BRIC and G4 liquidity on real oil prices. The single-state VAR model serves as a baseline, while the MSVAR model accounts for potential regime shifts in the data. Monthly data from January 1999 to December 2020 are used. The liquidity variable is constructed using real M2 (M2 adjusted for inflation) for both the BRIC and G4 economies. Global interest rates (IR), industrial production (IP), and global crude oil production (GP) are included as control variables, constructed using principal component analysis to create global measures. The variables are tested for stationarity using ADF and Phillips-Perron unit root tests. For the MSVAR model, several specifications are considered including Markov-switching models with regime-dependent intercepts (MSIs), regime-dependent intercepts and autoregressive coefficients (MSIAs), regime-dependent intercepts and heteroscedasticity (MSIHs), and regime-dependent intercepts, autoregressive coefficients, and heteroscedasticity (MSIAHs). Model selection is based on information criteria (AIC, HQ, SIC) with a preference for parsimonious models. The chosen model is estimated using maximum likelihood estimation via the expectation-maximization algorithm. Finally, impulse response functions (IRFs) are employed to analyze the dynamic effects of liquidity shocks on oil prices across different regimes identified in the Markov-switching model. Cholesky decomposition is utilized to address identification issues.

The single-state VAR(1) model reveals that only the BRIC liquidity measure has a statistically significant impact on real oil returns. However, the MSVAR model, which allows for regime switching, provides a more nuanced picture. The MSVAR model selection process (using AIC, HQ, and SIC) led to the selection of an MSIH(3,1) model, indicating three regimes and a constant VAR matrix. The three regimes are characterized by varying volatility of real oil returns: a low volatility regime, a high volatility regime, and a crisis regime. The crisis regime is most notably associated with the subprime crisis and the COVID-19 pandemic. The MSVAR model demonstrates that both BRIC and G4 liquidity shocks significantly affect real oil prices, with the impact of G4 liquidity being particularly pronounced during the crisis regime (almost double the impact in normal times). Furthermore, the impact of BRIC liquidity during the crisis is nearly three times larger than in normal periods. The impulse response functions show that the influence of BRIC liquidity on real oil prices is consistently greater than that of G4 liquidity across all regimes. The coefficients of the control variables IP(t-1) and GP(t-1) are generally not statistically significant in the chosen model, indicating that demand and supply shocks play a less dominant role in determining real oil prices than liquidity shocks in this context. The analysis further suggests a transmission mechanism of liquidity from G4 to BRIC countries, but not vice versa, highlighting the predictability of BRIC monetary policy compared to G4.

The findings suggest that the financialization of commodities plays a significant role in shaping oil prices, as excess liquidity contributes to speculative price movements beyond those justified by fundamental supply and demand. The single-state VAR model's inability to capture the effect of G4 liquidity underscores the importance of using models that can accommodate regime changes and nonlinear dynamics. The significant impact of liquidity shocks, particularly during crisis periods, highlights the interconnectedness of global financial markets and the influence of monetary policy on commodity prices. The greater impact of BRIC liquidity, especially during crises, suggests the increasing influence of emerging economies on global commodity markets. The lack of significance for the demand and supply variables suggests the need for future research to further disentangle the specific interactions between financial and fundamental factors in oil price determination.

This study demonstrates the significant and nonlinear impact of both BRIC and G4 liquidity on real oil prices, particularly during crisis periods. The use of Markov-switching VAR models provides a more comprehensive understanding than traditional VAR models, capturing the regime-dependent effects. Future research could explore the specific mechanisms through which liquidity affects oil prices, investigate the role of different types of financial investments in driving oil price volatility, and examine the effectiveness of policy interventions designed to mitigate the impact of liquidity shocks on energy markets. The current findings also provide warnings about the potential for further price increases if expansionary monetary policies continue.

The study's focus on real M2 as a proxy for liquidity might overlook other potentially relevant aspects of financial market conditions. While the study controls for global interest rates, industrial production, and global oil production, there could be other omitted variables affecting oil prices. The model's assumptions, such as the normality of error terms and the Markovian nature of regime shifts, may not perfectly reflect the complexity of real-world dynamics. Future research could explore the use of alternative liquidity measures and incorporate additional variables to enhance the model's explanatory power.