Laser-induced breakdown spectroscopy (LIBS) was applied in this study to quantitatively detect heavy metal lead (Pb) in paint coatings. By optimizing experimental parameters, the plasma spectrum of the paint coating within 308-417 nm was collected using a multi-channel spectrometer. Univariate and multivariate linear regression together with competitive adaptive reweighted sampling-partial least squares (CARS-PLS) analyses were later performed on the obtained Pb data of paint coatings to achieve more accurate quantitative results. Among the three analytical models, CARS-PLS displayed the highest accuracy; the determination coefficient of the calibration set was 0.991, and the relative errors of 4# and 7# of the predicted samples were 1.4% and 1.5%, respectively. Thus, CARS variable screening effectively screened important Pb-related data, and accurately detected Pb in the paint coating when combined with PLS.