Who is the "constituency" of USDA-AIPL?   Keep in mind that their industry contacts would come primarily from University Extension, established AI systems, and purebred Breed organizations.    The personnel involved do tend to rotate a bit between these three major employment sectors, so some would say it is an "inbred" industry.   AIPL would by its [government bureau] nature and [Beltsville MD, suburb of Washington DC] location have little substantive contact with individual dairymen.   Thus the established formats for evaluation have a fair amount of "inertial momentum".    If your herd is not on DHIA, your data is not in the AIPL database; if your cattle are not registered, your pedigrees are not in the AIPL database.

Are there problems with population genetics as practiced in dairy?   Most of the problems we identify come as a result of unintended consequences from previous assumptions or decisions.   If you accept the basic premise that a bovine "genetic" generation is four years (nine months gestation + two years rearing + fifteen months sorting the data from all steps in evaluation), there have been ten generations of change in dairy cattle since the 1964 USDA sire summary's use of "Predicted Differences".   On a basic biological basis-- turnover of generations-- we have seen thirteen generations.   In this time period, frozen semen became commercially available, organized AI became ascendant, young sire sampling programs were developed, Holstein type classification was dramatically changed twice, the "true type" models revamped, embryo transfer was developed, protein testing (and SCC recording) started, global exchange of genetics accelerated, and the "Interbull"(MACE) recalculations of sire data for global rankings accepted.    Adoption of each of these changes produced a change in the selection direction in our herds.   Today's dairy animals bear little to no resemblance to the ancestral material of four decades ago.

Neither does today's dairy industry, at least on the production side, have much in common with four decades ago.    Per-cow production is twice the level of four decades ago, in every major breed.  Milk cooperatives have gone national, with milk trucked from coast to coast prior to processing; dairy processing has likewise merged into near-monopoly size corporations.   Milk is as much a "commodity" as is corn or wheat, with pricing affected as much by CME contract trading as by the Federal Milk orders.   Vast amounts of mechanical and chemical technology and new hormonal therapies have been implemented.   Balancing of feed rations have adapted to much heavier grain loads, are computerized and now input a wide array of commodities through total mixed rations.  More dairymen never milk their own cows, doing their work from an office instead of a tractor seat.    At the same time a "parallel reality" is evolving within which we find intensive rotation grazing, organic production, and a renaissance in on-farm specialty marketing of fresh dairy products.   Crossbred cows and heritage breeds are having a new trial, as are some European "dual purpose" breeds, driven by niche markets and the frustrations of dairymen with health, fertility and calving ease issues in our highly-selected breeds.   Shortened cow life is exacerbated by the larger group cow housing and handling systems.   This group of hands-on dairymen question that "more milk yield" automatically means more profit, or even if milk quality has suffered from all the technological "progress" employed to produce it in quantity.

Defining the "genetic" difference

In the beginning, we had "daughter averages".   This was a standardized (305 days, 2x milking, ME) lactation average of daughters on official testing.   The higher the average, the better the bull must be.  Of course, the "management effect" made this a poor predictor-- herds in regions of higher quality feed, and herdsmen with exceptional feeding ability, could generate higher lactation records than an average farmer.   This approach was used from the beginnings of DHI into the late 1940s.

To compensate for this variation between herds, the concept of "daughter vs dam comparisons" was inaugurated in the 1950s.   The 305d 2x ME lactations completed for a bull's daughters would be compared to the average of all 305d 2x ME lactations of the cows who produced these daughters.   The difference was assumed to be the bull's genetic contribution.   Of course, the "management effect" that could play havoc with this measurement involved any changes in management and feeding that gave the newer generation (daughters) an environmental advantage over the older generation (dams).

As a result, in the 1964 sire summary we see those first "predicted difference" calculations, wherein a bull's daughters were compared against herdmates by different sires.   This gave us an in-generation comparison that seemed to neutralize the questions regarding management levels between regions and herdsmanship levels, and management improvement from one generation to the next.