Evidence for disassortative mating based on the human-specific MHC, i.e. HLA, is equivocal1. Initial evidence for disassortative HLA-pairing in the European-descent Hutterites2 has tended not to replicate in other populations. Recent studies, rather, reflect assortative mating associated with cosmopolitan population structure1. Although their configuration is more relevant to the majority of human evolutionary history, few small-scale, endogamous populations have been examined. Here, we test whether the extent of HLA dissimilarity between individuals influences mate choice by contrasting chosen partnerships (both love marriages and informal partnerships) with arranged partnerships among the Himba, an endogamous and formerly natural fertility group living in northern Namibia. With targeted sequencing we typed eight classical HLA genes for n=366 individuals and recorded self-reported partnership types and informal partnerships discovered through shared biological offspring. Across 249 known partnerships, we find no difference in HLA dissimilarity between chosen and arranged partnerships and no effect of the two partnership types that differs from a random, background distribution of partnerships. We also looked directly for fitness benefits of the potential offspring genotypes by testing HLA-based pathogen binding affinities, but also found no difference between arranged and chosen partners and no effects that differed from random partners. It is possible that due to high background HLA diversity in the Himba, most unrelated individuals in the population are sufficiently dissimilar at the HLA that there are few fitness benefits to offspring by maximizing mate dissimilarity. We do find, however, extensive haplotype sharing at the HLA region, suggestive of recent positive selection3,4. Episodes of fluctuating positive selection may be a more important driver in maintenance of HLA polymorphism than disassortative mating.