The variability of HLA genes provides a framework for the study of evolutionary processes and the genetic history of populations. We analyse the polymorphism of 19 African populations at 8 HLA loci. We describe the patterns of intra-population and inter-population polymorphism to make inferences on the mechanisms shaping HLA gene profiles, evaluate the efficiency of chip genotyping and imputation in capturing polymorphism compared to PacBio sequencing, and identify biases introduced by the typing method and their effect on population genetic analyses. We find that the density of variants identified by the Illumina Omni5 chip is insufficient to characterize the extent of HLA polymorphism, and that the chip under-estimates heterozygosity due to biases in SNP selection and typing errors. Imputation exacerbates the heterozygosity deficit and under-estimates population differentiation, likely due the bias present in the pre-imputation data and the reference panel rather than an endogenous bias in the imputation process.