Linkage disequilibrium (the correlations between neighboring genetic variants) enable the identification of broad genomic loci associated with disease but complicate the identification of individual causal variants. Genetic risk scores rely on these correlations to predict genetic risk without knowing the causal variants. Unfortunately, variation across populations in linkage disequilibrium mean that risk scores do not always translate well across populations. We will discuss statistical models of linkage disequilibrium across populations, consequences on the design of association studies, and possible solutions to overcome the pervasive sampling inequality in genomic studies.