Robust Bootstrap Quantile Causality under Structural Breaks and Heavy-Tailed Innovations: A Monte Carlo Study

Abstract This paper develops a robust bootstrap quantile-causality based approach for modelling the predictors in time series with dynamical structure, serial dependence and heavy-tails innovations. The proposed model combines candidate structural-break detection, robust residual handling and dependent block bootstrap procedures with quantile-specific causality detection. In an empirical illustration, 312 monthly observations on inflation, exchange rate, and Nigerian Stock Index (NSI) from January 2000 to December 2025 are used. October 2022 is the identified candidate structural break. The evidence of distributional heterogeneity is observed with both inflation and exchange rate predicting exchange rate and inflation respectively at the 0.10, 0.50, and 0.75 quantiles, respectively, the former relationship indicates bi-directional predictability at the 0.10, 0.50 quantiles while the latter reveals no predictability. The stronger predictive influence of these measures of macroeconomic fundamentals over the NSI is detected at particular quantiles. A Monte Carlo simulation of sample size of 120 and M = 5,000 repetitions, with Gaussian and standardised Student-t(3) innovations, with known structural break, no-predictiveness, and predictive power at lower, middle, higher quantiles are implemented to examine the finite-sample performance. The simulation is designed to assess finite-sample size and power at various quantiles. In sum, the findings show notable variations in macro-financial causal relationships across the conditional distribution, while established inferential techniques are appropriate to handle such variations that is conditional on structural break, serial dependence, heavy-tail robust innovations. The paper extends distribution-sensitive causal inference for macro-finance investigation. Communications in Statistics -...