Multifractality and time-varying market dynamics across sectors: Evidence from Brazil during the global financial crisis and COVID-19

Abstract This study examines the multifractal properties and time-varying scaling of four Brazilian sectoral indices during the Global Financial Crisis (GFC) and the COVID-19 pandemic. Multifractal Detrended Fluctuation Analysis (MF-DFA) is combined with shuffled and phase-randomized series, formal moving-block bootstrap comparisons, rank Gaussianization, finite-sample simulations, matched fractional Gaussian noise (fGn), parameter sensitivity checks, and rolling windows. Point estimates of multifractal intensity are generally larger during the GFC, particularly for the financial sector, but none of the cross-crisis or cross-sector differences is statistically significant after Benjamini–Hochberg adjustment. The transformed-series diagnostics indicate substantial sensitivity to heavy-tailed marginal distributions, especially during the GFC. Relative to matched finite-sample fGn, all four GFC series retain excess spectrum width, whereas during COVID-19 only the real estate sector does so at the 5% level under the $$q\in [-4,4]$$ validation. Rolling estimates vary over time but overlapping 250-day windows generate frequent false excursions under monofractal simulations; they are therefore interpreted as exploratory regime indicators rather than formal break tests. The evidence supports multifractal scaling in Brazilian sectoral returns while showing that finite-sample effects, marginal tails, and statistical uncertainty preclude universal claims of genuine multifractality or statistically established sector and crisis rankings.