Robust Statistical Methods for Analysing Skewed Income Data Among Informal Sector Workers in Lagos

Income data among informal sector workers in Lagos State are characterised by extreme right skewness and outlier sensitivity that undermine the validity of conventional mean-based statistical inference, necessitating robust and distribution-robust alternative methods. This study applies robust statistical methods to income data from 500 informal sector workers in Lagos Island, Surulere, and Alimosho LGAs of Lagos State. Participants were recruited from three market and trade association clusters. Monthly income was self-reported and corroborated by weekly transaction records where available. Trimmed means, winsorised means, M-estimators, and median regression were applied and compared to OLS regression for income determinant analysis. Bootstrap confidence intervals were constructed for all estimators. Mean income was N62,400 per month but the median was N38,200, indicating substantial right skew (skewness = 3.8). Winsorised mean (5 percent both tails) was N47,600. Huber M-estimator provided estimates with 34 percent lower standard error than OLS for income level estimation. Quantile regression at the 25th, 50th, and 75th percentiles revealed heterogeneous effects of education and business type across the income distribution, effects invisible to OLS regression. Male workers earned statistically significantly more at the 75th percentile (median regression coefficient N12,400, p < 0.001) but not at the 25th percentile. The study demonstrates the inferential gains from robust methods for Nigerian informal sector income analysis and recommends their adoption in NBS income survey analysis protocols. Keywords: robust statistics, income distribution, informal sector, Lagos, quantile regression