Are there too many traits in our selection indices?

Abstract

Selective breeding has been practiced since domestication, but early breeders commonly selected on appearance (e.g., coat color) rather than performance traits (e.g., milk yield). A breeding index converts information about several traits into a single number, typically representing total economic merit, that can be used for selection or to predict an animal's own performance. The first USDA selection index released in 1971 included only milk and fat yield, whereas the 2025 revision of the Lifetime Net Merit (NM$) index includes 17 traits and composites (weighted averages of other traits). Many dairy farmers, semen salespeople, sire analysts, and genetics consultants (expressed in personal communications) are concerned that, as the number of traits in the index grows, it becomes less effective because selection response is "diluted" (the efficiency of the index decreases). What actually happens is that the response in individual traits changes, but the overall index response (total economic merit) does not. Correlated response to selection for 24 traits and composites currently evaluated by the Council on Dairy Cattle Breeding (Bowie, MD) and Holstein Association USA (Brattleboro, VT) was calculated for 12 USDA selection indices and their changes over time were examined. Indices including only yield (milk, fat, and protein) traits had the highest heritabilities (0.272-0.290), whereas NM$ had heritabilities ranging from 0.188 in 2010 to 0.243 in 2025. Rates of genetic gain have changed over time but are in favorable directions for most traits. Although selection response has slowed for some traits (e.g., production) as new traits were added to the index, they remain favorable (positive). Other traits, most notably fertility, now have favorable trends when they had unfavorable (negative) trends in earlier versions of the selection index. There is some reranking within birth year cohorts for both genetic merit and reliabilities because older animals were selected using different criteria than contemporary animals. Rank correlations are high across indices, ranging from 0.8746 to 0.9994, meaning that sires rank similarly despite the increasing complexity of the index. Selection gains for fat, protein, productive life, and BW composite have the greatest economic value regardless of how responses are calculated. These results show that properly constructed selection indices produce desirable rates and directions of gain for many traits at once.