Abstract: The objective of this study was to compare the efficiency of microsatellites and single nucleotide polymorphism (SNP) markers in paternity identification Microsatellites or SNP genotypes of 10 000 pseudo-parent-child relationship individuals were randomly simulated, exclusion approach was employed in paternity identification, and this procedure was repeated 100 times. The results showed that exclusion probability was mostly influenced by the polymorphism of markers. To achieve the same exclusion probability, the needed number of low polymorphic markers was greatly higher than that of high polymorphic markers. When the genotypes of the actual mother were available, the efficiency of paternity inference could be greatly improved. No matter to high or low polymorphic markers, more SNPs were required than microsatellites in each inferential scenario, to achieve the same exclusion probability. When the average heterozygosity of SNPs and microsatellites were 0.484 and 0.875, the needed number of SNPs was 5～6 times more than that of microsatellites. When the average heterozygosity of SNPs and microsatellites were 0.363 and 0.566, the needed number of SNPs was about 2 times more than that of microsatellites. When at least 2 conflicts were required in inference, 0.45 times more markers were necessary, to achieve the same exclusion probability (> 0.99), comparing to that required at least 1 conflict, for both two kinds of markers. With the dropping of the genotyping cost, and other characteristics of SNPs, such as lower genotyping error rate, higher repeatability and amenable to high-throughput automated analysis, it is very promising that SNPs would replace microsatellites for dairy paternity identification in the future.