Dina Sittel

Type of the article: Research Article

Abstract
Market efficiency remains a central issue in financial economics, particularly in increasingly automated financial markets. The Adaptive Market Hypothesis suggests that efficiency evolves as market conditions change, making high-frequency analysis useful for capturing short-lived changes in return behavior. This study examines whether the Moroccan stock market exhibits time-varying short-horizon return dynamics using Moroccan All Shares Index data from August 2016 to May 2025.
The final database contains 1,222,039 unique observations. Because timestamps are irregular, we align prices to fixed intervals within continuous trading before constructing returns. The analysis combines rolling multiple variance-ratio tests with wild-bootstrap inference, a bootstrap-centered measure of random-walk departure, regression analysis, and non-Bayesian time-varying autoregressive modeling.
The results show that departures from the random-walk benchmark vary markedly over time, with similar patterns across 5-, 10-, and 15-minute frequencies. In the main 10-minute specification, the joint random-walk null is rejected in 373 of 2,814 rolling windows (13.3%). Intraday differences are modest, but formal tests do not establish statistically significant differences across the three periods. Realized volatility is positively associated with larger contemporaneous departures, whereas associations with selected calendar and event conditions are mixed and specification-dependent. Formal parameter-constancy and slope-stability tests reject stability (slope-bootstrap p = 0.0091), while estimated linear predictability remains below the pointwise 95% bootstrap upper bound throughout the sample. The findings are consistent with the Adaptive Market Hypothesis and show that changing return dynamics do not necessarily generate persistent predictability.