Distributional Properties of Nigerian Macroeconomic Indicators, 2008–2024: Less Non-Normal Than Assumed

Abstract: Regression inference and many machine-learning loss functions assume, or at least work best with, errors that are not extremely fat-tailed or asymmetric. Emerging-market series are widely believed to violate this, and the belief is rarely tested carefully. This paper documents the distributional properties of ten Nigerian macroeconomic indicators over 67 quarters from 2008Q2 to 2024Q4, and finds a more mixed picture than the folklore suggests. Three of the ten do not reject normality at the 5% level: FDI itself (Jarque-Bera 2.11, p = 0.348), GDP (0.24, p = 0.885) and oil price (3.63, p = 0.163). Seven do reject, with exchange-rate volatility the most extreme (skewness 3.21, excess kurtosis 10.14). The result for FDI matters most, because FDI is the dependent variable in most Nigerian FDI research and is routinely assumed to be severely non-normal. An earlier version of this analysis, run on a longer sample in which everything before 2008 was interpolated from annual figures, found all ten variables rejecting normality and reported FDI with excess kurtosis of 7.95. On genuine data FDI has excess kurtosis of -0.17. The apparent fat tails were an artefact of splicing a smoothed series onto a noisy one.

Keywords: Non-Normality; Skewness; Kurtosis; Jarque-Bera; Nigerian Macroeconomic Data.