The COVID-19 pandemic caused unprecedented swings in financial market returns, challenging the Efficient Market Hypothesis (EMH). EMH posits that security prices are unpredictable based on past information, while the AMH, proposed by Lo (2004), accounts for periods of varying predictability in market efficiency. The AMH incorporates elements of both EMH and behavioral finance, suggesting that market efficiency and volatility change over time due to factors like information technology, macroeconomic conditions, and investor behavior. The COVID-19 pandemic, with its distinct phases (incubation, outbreak, and fever), provides a unique context to test the AMH. This study investigates whether the shocks to commodity markets during the pandemic validate the AMH, focusing on commodities traded on the Chicago Board of Options Exchange (CBOE). The study uses data from December 2, 2019, to December 31, 2020, examining commodity returns from three sectors: agriculture, energy, and precious metals. The analysis is conducted in two parts: first, applying the Generalized Spectral (GS) test to detect adaptive behavior; second, using linear and non-linear econometric models to identify predictability during different COVID-19 phases.

Existing research highlights the significant impact of COVID-19 on global economies and financial markets. Studies have shown negative relationships between COVID-19 death rates and stock market returns (Al-Awadhi et al., 2020; Ashraf, 2020; Liu et al., 2020; Onali, 2020). The pandemic's impact on U.S. financial markets was notably stronger than past pandemics (Baker et al., 2020). Other studies explored the pandemic's effects on various market indices, including equity, bond, and commodity markets (Ali et al., 2020; Lin and Su, 2021; Gunay, 2021; Okorie and Lin, 2021; Al-Refai et al., 2022; Matos et al., 2021). Prior research by Shahid et al. (2020a) supported the AMH in commodity markets during periods of crisis. This study builds on this existing literature by specifically focusing on the impact of COVID-19 on commodity markets and testing the AMH within the context of the pandemic's distinct phases.

The study uses daily data from December 2, 2019, to December 31, 2020, obtained from DataStream, focusing on popular commodities from the agriculture, energy, and precious metals sectors traded on the CBOE. The sample period is divided into sub-samples representing the epidemic, pandemic, endemic, and fatality phases of COVID-19 based on WHO data and existing literature. Commodity returns are calculated using the formula: r_t = [ln(P_t) - ln(P_t-1)] × 100. To analyze the time-varying behavior of commodity returns, the study employs the Generalized Spectral (GS) test with a one-month rolling window. This non-parametric test detects both linear and non-linear dependencies. Linear tests (autocorrelation, runs test, and variance ratio test) are used to assess the predictability of returns, while the BDS test examines non-linear predictability. The linear tests help determine the degree of dependence in the returns series during different COVID-19 phases. The variance ratio test assesses the random walk hypothesis, determining whether the variance of k-period returns is equal to k times the variance of one-period returns. The runs test evaluates the randomness of the returns series. The BDS test, a non-linear method, detects non-linear serial dependencies, considering the embedding dimension (m) and distance (ε). A pre-whitening AR model is applied to remove linearity from the series before applying the BDS test to the residuals. P-values are computed to determine the statistical significance of the test results, with the significance level set at 5%.

The GS test revealed that commodity returns exhibited adaptive behavior during the COVID-19 period, with significant variations in predictability across different phases. Linear tests (autocorrelation, runs test, and variance ratio test) showed varying levels of predictability for different commodities across the epidemic, pandemic, and endemic phases. The results showed that most indices experienced periods of significant predictability and no predictability, supporting the AMH. For example, in the agriculture sector, most indices showed no predictability during the epidemic, high predictability during the pandemic, and low predictability during the endemic period. Similar patterns were observed in the energy and precious metals sectors. The BDS test revealed significant non-linear predictability during the pandemic phase for most commodities but not during the epidemic or endemic phases, also supporting the AMH. The results further suggest that the behavior of commodity returns is influenced by the specific phase of the COVID-19 crisis.

The findings strongly support the AMH, demonstrating that commodity market returns exhibit adaptive behavior during the COVID-19 pandemic. The varying levels of predictability observed across different phases highlight the dynamic nature of market efficiency. The study's use of both linear and non-linear tests enhances the robustness of the findings, providing a comprehensive assessment of return predictability. The results contribute to the understanding of financial market dynamics during periods of significant uncertainty and crisis, emphasizing the role of investor behavior and adaptive strategies. The study's focus on commodity markets adds to the existing literature that predominantly concentrates on equity markets.

This study provides robust evidence supporting the AMH in commodity markets during the COVID-19 pandemic. The findings demonstrate time-varying predictability in commodity returns across different phases of the pandemic, emphasizing the importance of considering adaptive behavior in financial market analysis. Further research could extend this analysis to other asset classes such as debt and equity markets, providing a broader perspective on the AMH's applicability across different market segments.

The study's focus on a specific period (December 2019-December 2020) may limit the generalizability of the findings to other periods. The chosen commodities may not fully represent the entire commodity market. The model may not capture all factors influencing commodity prices during this complex period.