ROBUSTNESS CHECKS AND SENSITIVITY ANALYSIS OF ECONOMETRIC MODELS USING SIMULATED SYNTHETIC DATA: REPLICATING THE STATISTICAL PROPERTIES OF NIGERIAN MACROECONOMIC VARIABLES

Abstract

This study applies a Monte Carlo simulation approach to assess the robustness of macroeconomic growth models in Nigeria from 1990 to 2023 by generating 33 synthetic datasets that mimic the statistical behavior and interrelationships of key variables such as GDP, exchange rate, inflation, interest rate, industrialization, carbon emissions, and trade openness. Using parameters derived from time-series models like VAR, the simulated data preserve essential features including mean, variance, autocorrelation, and long-run relationships, enabling detailed sensitivity analysis. The findings show that log transformation generally stabilizes variance, although GDP, exchange rate, and inflation still exhibit skewness and heavy-tailed distributions, indicating persistent economic volatility. Stationarity tests confirm that all transformed variables are stable at levels. Long-run estimation using FMOLS reveals that inflation and interest rates positively influence GDP, while industrialization negatively affects growth, suggesting structural inefficiencies; other variables show limited long-run impact. Short-run dynamics from ARDL models highlight cyclical GDP adjustments, negative effects of exchange rate depreciation, and unstable trade openness impacts, alongside complex lag interactions among variables. Cointegration tests confirm stable long-run relationships, and VECM results indicate that, over time, industrialization, trade openness, carbon emissions, and inflation promote growth, whereas exchange rate depreciation and high interest rates hinder it, with short-run fluctuations largely driven by monetary and price adjustments.