RATES OF IMPROVEMENT BY PROGENY TESTING IN DAIRY HERDS OF VARIOUS SIZES By Lawrence VI. Specht AN ABSTRACT Submitted to the School of Advanced Graduate Studie Michigan State University of Agriculture and Applied Science in partial fulfillment of the requirements for the degree of DOCTOR OF PHILOSOPHY Department of Dairy Year 1907 cr Approved __ -■>.Crr V 'i , AY " i - \ /-a c y jy ' U - / a U a j y . p i ■ 1 ABSTRACT LAWRENCE W. SPECHT The production level of Michigan Holsteins and the contri­ bution of A.I. sires to the dairy herds of the state under the present system of sire selection were measured. Rates of gene­ tic gain thought possible and practical with progeny testing were compared for individual herds and for A.I. populations of varying sizes. The study utilized information from 34,073 Holstein cows with 65,392 DHIA records. Of this number, 18,327 records were collected from the mechanical computing system employed by the Michigan DHIA since October, 1954. The average annual pro­ duction of Holsteins In Michigan DHIA herds during the 19451955 period Increased at the rate of 0.9 par cent of the average annual yield. All records averaged 12,237 pounds of milk and 435 pounds of fat on a 2X-305 day-M.E. basis. The 5,098 daughters of 187 A.I. sires averaged 12,305 pounds of milk and 444 pounds of fat. The evidence suggests that the bulls used by the A.I. studs have not substantially increased the milk producing ability of the tested Holstein population. The data did not show an advantage for using the conven­ tional paired dam and daughter natural proof instead of the simple daughter average for predicting the production of the A.I. progeny of a sire. Correlation coefficients between the difference in production of the daughters minus their dams and the A.I. daughter average for 98 Holstein sires were 0.35 for milk and 0.33 for fat. The comparable regression coefficients wer^ 0.18 for milk and 0.19 for fat. Correlations between the ABSTRACT LAWRENCE W. SPECHT simple daughter average from the first available natural proof and the sire's A.I. duaghter average were 0.37 for milk and 0.30 for fat. The regressions were 0.18 for milk and O .16 for fat. Repeatabilities were 0 A S for milk and 0.40 for fat from the 18,327 records that were computed mechanically. An analysis of 2,631 records of A.I. progeny gave regressions of — ---n + 12 for milk and __I L . . . f o r fat for predicting the production n + 16 of future A.I. daughters from the production of n tested A.I. daughters. Heritabilities estimated from the paternal half- sib correlations found for the data were 0.31 for milk and 0.23 for fat. Statistics for generation interval, percentage of cows re­ moved annually, cow removals by lactation, reasons for removals, and heifer mortality were obtained from the mechanically computed data. Most of the estimates agreed well with previously published data. In herds of less than 100 cows, progeny testing was less efficient than selection of young sires based on the production of their dams. Progeny testing in herds of 100 and 200 cows had a slight advantage over the alternative method. progeny testing In A.I. populations of 2,000 and 10,000 production tested cows gave evidence of making annual genetic progress of from 1.5 to 2.3 per cent of the average annual yield. For a practical situation involving the current oper­ ating procedures of the stud breeding 150,000 cows In Michigan (10,000 of which are production tested), maximum progress from 3 ABSTRACT LAWRENCE W. SPECHT a progeny testing scheme with young sires was estimated at 1.7 to 1.8 per cent of the average annual yield. Further improve­ ment appears possible if the number of cows on test is increased and if more cows are bred per A.I. sire. The careful selection of young sires for progeny testing under A.I. conditions and the saving of the best of the progeny tested young sires for extensive A.I. use seem to offer a more reliable method of Improving the genetic merit of the dairy population than does the system of selecting A.I. sires on the basis of their natural proofs. RATES OF IMPROVEMENT BY PROGENY TESTING IN DAIRY HERDS OF VARIOUS SIZES By Lawrence W. Specht A THESIS Submitted to the School of Advanced Graduate Studies of Michigan State University of Agriculture and Applied Science in partial fulfillment of the requirements for the degree of DOCTOR OF PHILOSOPHY Department of Dairy 1957 ProQuest Number: 10008587 Alt rights reserved INFORMATION TO ALL USERS The quality of this reproduction is dependent upon the quality o f the copy submitted. In the unlikely event that the author did not send a complete manuscript and there are missing pages, these will be noted. Also, if material had to be removed, a note will indicate the deletion. uest. ProQuest 10008587 Published by ProQuest LLC (2016). Copyright of the Dissertation is held by the Author. All rights reserved. This work is protected against unauthorized copying under Title 17, United States Code Microform Edition © ProQuest LLC. ProQuest LLC. 789 East Eisenhower Parkway P.O. Box 1346 Ann Arbor, Ml 4 8 1 0 6 - 1346 ACKNOWLEDGMENTS The writer wishes to thank Dr. L. D. McGilliard for his friendly counsel and many helpful suggestions during the in­ vestigation of the problem and the writing of the thesis. He is indebted to D r . N. P. Ralston, Head of the Dairy Department, for financial assistance in the form of graduate research and teaching assistantships during the program of study. Grateful acknowledgment is due Dr. D. E. Madden for his helpful suggestions and words of encouragement. Thanks are extended to Mr. R. P. Witte, research assistant, and Mr. A. J. Thelan, office supervisor of the Michigan DHIAIBM tabulating center, for their assistance with the machine procedures and their help in obtaining certain portions of the d ata . Special mention Is also due to the following men and their organizations for help given the writer in obtaining the needed Information on the A.I. sires used in Michigan: Mr. John Hecker of the Michigan Artificial Breeding Cooperative, East Lansing, Michigan; Mr. Jack Van Hoven, representative of the American Breeders Service, Jenison, Michigan; Mr. Elmer Hansen, Director of Public Relations, Curtiss Improved Stud Service, Inc., Cary, Illinois; and Mr. Alton Dale Block, Badger Breeders Cooperative, Shawano, Wisconsin. ii TABLE OP CONTENTS Page INTRODUCTION ............................................. 1 REVIEW OF LITERATURE .................................... 3 METHODS AND R E S U L T S ...................... . .............. 10 Source of D a t a ....................................... 10 Lactation d a t a ...................................10 Sire d a t a ....................................... 10 Sire p r o o f s ..................................... 10 .......................... Recent Production Trends 11 Daughter Averages for A.I. Sires................... 16 Methods of Selecting Sires for A.I. Service . . . . 17 Repeatability ...................................... 21 Heritability........................................... 22 Paternal Sib Correlations .......................... 23 Predicting future daughters' production from n tested d a u g h t e r s ......................28 Means ancT standard deviations....................29 Non-genetic Parameters ............................ 30 Generation interval .......................... 31 Percentage of cows removed annually............. 31 Cows removed during each lactation ........... 33 Number of available herd replacements........... 3§ Minimum number of tested bulls needed........... 4l Useful life span of tested bulls in A.I. . . . 42 ........ 42 Survival percentage for young sires Calculating Genetic Gain in the Individual Herd . . 43 Genetic Gain with Progeny Testing ................. 48 Results in Individual Herds 52 ....................... Twenty-five cow h e r d ............................ 57 Fifty cow h e r d ...................................57 One hundred cow h e r d ............................ 58 Two hundred cow herd ...................... 58 iii Page Results in A.I. P o p u l a t i o n s ........................... 59 Small A.I. p o p u l a t i o n s ........................... 59 Large A.I.populations ..........................6b .......................... 65 A practical example E c o n o m i c s ....................................... 68 SUMMARY...................................................... 70 LITERATURE C I T E D ........ ' ................................ 72 iv LIST OP TABLES TABLE Page 1. Michigan DHIA Herd Averages for Holsteins . . . 11 2. Yearly Holstein Lactation Averages from DHIA-IBM Records ............................ 13 3. 5. 6. 78. 9. 10. Daughter Averages for Natural Service Sire G r o u p s ............................. 14 Daughter Averages for A.I. S i r e s ......... 16 Production Averages and Correlation and Regression Coefficients for the Naturally Sired and A.I. Daughters of Stud Sires . . . 18 Components of Variance for Milk and Pat P r o d u c t i o n ...................... Components of Variance for Milk and Pat Production for Naturally Sired Cows 21 . . . . 2b Components of Variance for Milk and Fat Production for A.I. P r o g e n y ........... Distribution of Variance in Individual Years. 25 . Variance Components for A.I. Daughters from DHIA-IBM D a t a ...................... 11. Means and Standard Deviations of Milk and 12. Annual Cow R e m o v a l ......................... 13- Reasons for Cow Removal in 269 Michigan Holstein Herds. ................... 26 27 Pat . 30 32 33 lb. Cows Removed during Each L a c t a t i o n ........ 15. Effect of Lactation Number on the Percentage of Voluntary and Involuntary Culling....... 35 16 . Distribution of Cows by Age at First C a l v i n g ........................ v 34- 38 LIST OF FIGURES FIGURE Page 1. Genetic Improvement with Progeny Testing in a 25 Cow Herd, Saving One of n Sires Sampled............53 2. Genetic Improvement with Progeny Testing in a 50 Cow Herd, Saving One of n Sires Sampled............54 3. Genetic Improvement with Progeny Testing in a 100 Cow Herd, Saving One of n Sires Sampled............55 4. Genetic Improvement with Progeny Testing in a 200 Cow Herd, Saving One of n Sires Sampled............55 5. Genetic Improvement with Progeny Testing in an A.I. Population of 2,000 Tested Cows, Saving One of n Sires Sampled............. 60 6. Genetic Improvement with Progeny Testing In an A.I. Population of 10,000 Tested Cows, Saving One of n_ Sires Sampled...................... 6l 7. Genetic Improvement with Progeny Testing in an A.I. Population of 10,000 Tested Cows, Saving Two of n Sires Sampled...................... 62 8. Genetic Improvement with Progeny Testing in an A.I. Population of 150,000 Cows (10,000 Tested), Saving Five of n Sires Sampled . . . . vi 63 INTRODUCTION The major problem in appraising the rate of genetic gain in dairy herds lies with the difficulty in distinguishing b e ­ tween the relative contributions of heredity and environment to the annual change in the average production. Most of the actual and the theoretical estimates of gain have indicated that the annual improvement in the genetic merit of dairy herds which can be attained does not exceed 1 per cent of the average annual yield. Recent estimates of genetic gain possible have been published by Dickerson and Hazel Robertson (19^*0 and by Rendel and (1950) for single herds, and by Robertson and Rendel (1950) for artificially inseminated Robertson and Rendel (A.I.) populations. (1950) have suggested that the judi­ cious use of progeny testing, coupled with an A.I. program, could lead to marked improvement In the rate of genetic gain In dairy cattle. Artificial insemination has markedly extended the influence of a single sire by making possible the breeding of a large number of females to a few bulls. Continued im­ provement In semen processing and storing and in Insemination techniques will reduce still further the number of bulls needed by A.I. units. In addition, the expansion of the A.I. program could persist for many years. These factors will permit the dairy cattle breeder to Increase the Intensity of selecting males for extensive breeding. Although considerable information Is available in the literature concerning repeatability, herltability, phenotypic 1 2 and genotypic variances of milk and fat production, many of the statistics needed for calculating genetic gain are nongenetic in nature and are primarily affected by the management practices in dairy herds. Values assumed for such parameters have been based on scanty evidence because of the difficulty in collecting data in a routine manner from a sizable number of herds. Adapting the Michigan dairy records program to mechanized computing equipment has greatly reduced this problem. The information accumulated from the Michigan DHIA-IBM program and lactation records completed in standard DHIA afforded an opportunity to obtain some of the statistics needed. The problem under consideration involves measuring the production level of Holsteins in Michigan and the contribution of A.I. sires to that population under the present system of sire selection, and comparing rates of genetic gain possible and practical with varying degrees of emphasis on progeny testing In dairy herds of various sizes. REVIEW OP LITERATURE The progeny testing of dairy bulls under the auspices of the Mproved" sire program has been emphasized in the educational programs of the state and federal dairy extension agencies since 1935* Such a test can be the most accurate single source of information about a sire’s breeding value, since more in­ formation will usually be available in the form of records from a number of female offspring than can be obtained from the bull's pedigree. However, progeny testing tends to increase the generation interval and may actually decrease the genetic progress made per year. Dickerson and Hazel (19^) compared the amount of genetic progress expected from a progeny testing plan with the improvement possible from mass selection (using each c o w ’s own production to determine which will be allowed to produce the offspring of the next generation) and reported for a herd of 120 cows that the additional improvement possible with progeny testing was more than offset by the longer gener­ ation interval. Progeny testing for herds of this size was less efficient than mass selection. Robertson and Rendel (1950), using a slightly different procedure, estimated that progeny testing would have a slight advantage over mass selection for lactation yield in a herd of 120 cows. They also pointed out that in smaller herds other problems arise in the use of the progeny test. For example, when young sires have been bred to enough cows to progeny test them adequately, there are few cows left to breed to the tested 3 4 bulls. Also, the number of bulls that can be tested in small herds is small, and, therefore, the amount of selection among those tested vdll necessarily be small. These difficulties are partly the result of the few calves conceived which even­ tually complete one lactation. Lush (19^9) estimated that 20 cows must be in calf to a single sire if the dairyman is to have a fair chance that at least five of the bull's daughters will finish their first lactations. Lush's estimate that one out of every four cows in calf will produce a daughter with a milk record of her own is more conservative than the one in three, based on New Zealand data and reported by Robertson and Rendel (1950). However, Robertson (195^) amended the earlier estimate and suggested that one in six was more accurate. A recent publication from the New Zealand Dairy Board (1956) stated that for tested herds utilizing A.I. service, 68 per cent of the identified A.I. heifers born in 1953 were reported on test as two-year-olds in 1955. A comparable sur­ vey for naturally sired calves born in 19^2 showed that 63 per cent of the Identified calves entered the same herd as two-year-olds. Under New Zealand conditions, for every 100 cows In calf, 3^ heifers will be reared. No data were given concerning the fate of the other 16 females expected with a normal sex ratio. same herd. Twenty-three of the 3^ will freshen In the These data are in good agreement with Lush's esti­ mate that 4 cows must be in calf to realize one milking heifer for sire testing purposes. Theoretical estimates of the amount of genetic progress possible in single herds are of the order of one per cent of 5 the average yield per year (Dickerson and Hazel, 1944, and Rendel and Robertson, 1950). are scarce. Estimates from actual situations Rendel and Robertson (1950) analyzed the records of a single herd operated for nearly fifty years and calculated that the probable improvement achieved by selection was about 0.7 per cent of the average yield per year. The generation interval for the herd was extremely short, just over 3 years, compared to a mean generation length for the Shorthorn and Friesian breeds in England of approximately 4 l/2 years. The authors stated that the longer generation interval would reduce the improvement expected annually to about 0.6 per cent of the average yield. The herd involved averaged about 800 gal­ lons of milk per cow per year. In terms of the English gallon, this is roughly 8,240 pounds of milk per cow and would mean an average annual genetic increase of about 50 pounds per year. Harvey (1953) reported an average yearly increase for 25 years of 8.0 pounds of fat for Holsteins and 5*3 pounds for Jerseys after adjusting for environmental effects, in the University of Idaho herds. Harvey Indicated the method of analysis employed may have overestimated the genetic progress. The adjusted herd average for the Holstein herd was 486 pounds of fat and for the Jersey herd, 396 pounds of fat. Estimated rates of genetic improvement are 1.6 per cent and 1.3 per cent of the mean yield. Dillon et a l . (1955) utilized the same method as Harvey and found that the average real producing ability in the University of Illinois Holstein herd changed very little over a 54 year period. The increase in average real producing 6 ability per year was 0.68 + 14.00 pounds of fat-corrected milk:. The environmental change per year was 53*54 + 14.44 pounds of fat-corrected milk:. Plum and Rumery (1956) estimated that the genetic im­ provement for the Holstein herd at the North Platte Experiment Station amounted to approximately 7 pounds of butterfat per generation. Average generation lengths of 5*2 years for cows and 4.2 years for bulls were reported. The study covered a period from 1913 to 1954. Laben and Herman (1950) studied the production in the University of Missouri Holstein herd from 1902 to 1950. From information presented by these authors, Plum and Rumery (1956) concluded that the improvement per 6 year generation interval equalled 9 pounds of butterfat. No attempt was made to differ­ entiate between the effects of heredity and management. McGilliard (1952) reviewed the question and calculated the annual improvement in a closed herd of Holsteins at Iowa State College. The herd average production increased at the rate of 5-0 pounds of fat per year, while the producing ability increased 2.4 pounds per year. The latter value measures the change in hereditary merit of the herd plus changes in the permanent environment, while the remainder is associated with the changes in management and other temporary environmental influences. Such a division provides a sorely needed estimate of the relative contribution of heredity and environment to changes in the annual production averages of dairy cattle populations. 7 Earlier attempts to estimate genetic gain are discussed by M e G i l H a r d (1952), by Rendel and Robertson (1950), and by Rendel et a l . (1951)* Most of the methods reviewed are criti­ cized on the grounds that the increase with age of the suc­ cessive records of a cow is not independent of the effects of management. Robertson and Rendel (1950) have shown that the use of progeny testing in A.I. units could lead to an annual genetic improvement of from 1.5 to 1.7 per cent of the average pro­ duction for studs breeding 2,000 cows. where the number of cows bred per stud In the United States is much larger, recorded populations of 10,000 or more cows for a single breed are not uncommon, even when only 5 to 10 per cent of the serviced herds are on test.According to these authors' calculations, the annual improvement could exceed 2 per cent of the average yield for a population of 10,000 cows. Several groups of workers have suggested ways to use the progeny test to appraise sires destined for use in A.I. studs: Robertson and Rendel (1950); Robertson (195^); Henderson and Dunbar (1952); Henderson (195^); Legates (195^). Since advanced age rules out the progeny testing of bulls originally brought into the stud with "proofs" in natural service, such plans would be feasible only with young sires. Information from New York (Albrectsen, 1953, and Carter, 195^) a^d Oregon (Oregon Dairy Breeders' Newsletter, 1957) suggests that the A.I. daughters of young bulls selected on the basis of pedigree carefully produce as well as the 8 A.I. daughters of bulls selected on the basis of a natural "proof." New Zealand Dairy Board (1956) data indicated that 20 naturally proven Jersey bulls had an average difference from "Expectancy" of +18 on the basis of the A.I. daughter performance. The expectancy value Is designed to take Into account the level of production for the herd(s) in which the bull's daughters made their records. Bulls with plus expectancy figures are considered above average in breeding value, and those with a minus value are considered below average. The comparable value for the 20 sires computed from their naturally sired daughters was +3^-* Sixteen unproven Jersey bulls averaged +8 from "Expectancy," and only A of these had minus expectancy values. While the New Zealand results are not as favorable to the young sires as those from the United States, the report concluded that the results are encouraging and not likely to be detrimental to the New Zealand A.I. program. If the in­ formation available represents the true situation, an additional gain from selection should be possible when the pedigree selected young sires are progeny tested under A.I. conditions, and only the best saved for extensive use in A.I. This procedure should provide sires that could be used with considerable confidence. The previously mentioned proponents of progeny testing for young sires envisage a scheme whereby the young bulls sampled are a result of mating the best progeny tested sires In the unit to the outstanding cows in the area. As Robertson (195^) suggested, under such a plan, the A.I. units can play a con­ structive role in dairy cattle breeding, rather than having their rate of genetic progress tied to that of the bullbreeding herds of the country. METHODS AND RESULTS Source of Data Lactation data. The Holstein production data utilized In the study were obtained from (1) the Michigan DHIA-IBM program and (2) the standard DHIA completed lactation cards (hereafter referred to as "7l8n cards) returned by the Dairy Husbandry Research Branch, Washington, D.C. Only completed records be­ tween 180 and 305 days In length were studied. were corrected for age by standard DHIA factors All records (National Cooperative Dairy Herd Improvement Newsletter, January, 1955). Records from cows milked 3 times per day and from cows with Incomplete lactations were omitted. Various portions of one or both groups of records were utilized to obtain the desired statistics. Sire data. The four editions of the Michigan Artificial Breeders Cooperative sire book provided Information on the life in service of the Holstein bulls In the stud. Similar infor­ mation was obtained by correspondence with the other A.I. units serving Michigan. With this Information, it was possible to determine the span of time during which a particular sire's A.I. progeny would be born. The A.I. daughters of stud sires were then selected from the data on the basis of their birth dat es. Sire proofs. The sire lists and monthly newsletters of the National Cooperative Dairy Herd Improvement Program provided 10 11 the bulk of the first available proofs on the stud sires. An HIR proof or an unofficial proof compiled by the stud was utilized if the sire did not have a DHIA proof. Recent Production Trends The average production for Michigan Holstein herds in the Dairy Herd Improvement Program has risen appreciably since 19^5* Table 1 gives the average milk and butterfat production for the ten years from 1945 to 1955> as tabulated by the Dairy Husbandry Research Branch, Washington, D.C. TABLE 1 MICHIGAN DHIA HERD AVERAGES FOR HOLSTEINS Year No. Herds Milk(lb.) Fat (lb.) % 1945-45 - 10,098 359 3.56 1946-4-7 - 10,233 360 3.52 1947-48 - 10,423 373 3.58 1948-49 - 10,253 365 3.56 1949-50 - 10,278 367 3 .57 1950-51 - 10,762 365 3 .58 1951-52 - 10,907 390 3 .60 1952-53 730 10,818 390 3.61 1953-54 1,128 10,906 395 3.62 1954-55 1,102 10,919 393 3.60 Not available prior to 1952. Table 1 shows an Increase of 34 pounds of butterfat for the ten year period, an average gain of 3.4 pounds per year, 12 which is equal to 0.9 per cent of the average yield. The question arises as to how much of the improvement is due to improved feeding and management and how much is due to an in­ crease in genetic merit of the cow population. There are three possibilities with respect to the level of feeding and manage­ ment. Either it has improved, it has declined, or it is rela­ tively unchanged from what it was ten years ago. It is hoped that the management level of the large number of herds involved h a s n ’t declined, (although this cannot be known precisely) in view of the efforts put forth by the educational and commercial groups interested in dairying. If the management level hasn't declined, but has remained unchanged, the genetic gain could be as large as 0.9 P^r cent of the average yield. It Is also possible that all of the gain is due to improved feeding and management, and that the genetic worth of the herds has changed scarcely at a l l . Such a situation was indicated by Dillon et a l . (1955)> although only for a single herd. Probably neither extreme is correct, and the situation is as McGIlliard (1 9 5 2 ) found, both factors are making some contribution to the overall yearly Improvement. Whether each factor contributes almost equally, as in McGilliard’s example, Is not known. If the environment has merely maintained the level it had ten years ago, the largest possible genetic gain this popula­ tion has obtained would fall short of the values predicted In theoretical approaches to the problem. More likely, the genetic gain in the population has been only a portion of the total gain. Regardless of the Importance of the environmental effect, It would be worthwhile to know whether or not the present rate 13 of improvement could be increased by the widespread adoption of progeny testing in Michigan herds. The average production for various groups of both naturally sired and A.I. cows was calculated in order to make comparisons with the information in Table 1. Tables 2, 3, and k. The data are presented in The higher average production in these tables is the result of using mature equivalent records, while Table 1 is in terms of actual production. Yearly production averages were not available, since the Individual lactation records of a cow were averaged in order to reduce the number of cards handled in subsequent computations. Averages for the years 1953, 195^, &nd 1955 were obtained from DHIA-IBM records utilized in the repeatability analysis. The lactation averages for records begun In the 1953-1955 period are given in Table 2. TABLE 2 YEARLY HOLSTEIN LACTATION AVERAGES PROM DHIA-IBM RECORDS Year of Calving No. of Records 2X-305-M •E. Milk(lb.) Pat (ib.) % 1953 8 59 12,526 453 3.62 1954 7,324 12,014 433 3.6 0 1955 10,144 12,202 440 3.61 18,327 12,142 438 3.61 Average Since yearly production averages were not calculated from the bulk of the data, an estimate of the trend in the production level for Michigan Holsteins was obtained by summarizing the daughter averages by groups of sire registration numbers. 14 Table 3 presents the average milk and butterfat yields from the daughters of various groups of sires used only In natural service. TABLE 3 DAUGHTER AVERAGES FOR NATURAL SERVICE SIRE GROUPS Sire Identification No. of Sires No. of Daus. Canadian Sires 1949 2892 5432 11,782 404 3.43 U. S. Sires Grades 1287 4426 7500 11,783 421 3.57 Purebred s-*< 700,000 146 297 588 12,041 423 3-51 700,000 - 749,999 302 721 1519 12,421 428 3.45 7 5 0 , 0 0 0 - 799,999 475 1447 3154 12,200 423 3.47 8 0 0 , 0 0 0 - 849,999 558 2448 5641 12,140 426 3-51 8 5 0 , 0 0 0 - 899,999 6 30 2877 6660 11,909 420 3-53 9 0 0 , 0 0 0 - 949.999 642 3241 7212 12,340 439 3.56 9 5 0 , 0 0 0 - 999,999 543 2592 5191 12,205 433 3.55 1,000,000-1,049,999 510 2243 3930 12,487 447 3.58 1,050,000-1,099,999 483 2118 3166 12,552 452 3.60 1 ,1 0 0 , 0 0 0 478 1398 1672 12,608 453 3-59 Total or average purebreds 4767 19,382 38,733 12,280 435 3.54 No. of Records 2X-305-M. E. Pat Milk ft 1Grouped by sire registration number. The complete data involved 34,073 daughters of 7>550 sires. Only 187 of the sires had been in A.I. service. The daughters of all sires had a total of 6 5 , 3 9 2 records, which averaged 12,237 pounds of milk and 435 pounds of fat with a 3.55 per cent butterfat test. 15 Table 3 shows that the daughter average for Canadian sires is below that of the native sires because of a lower average butterfat test. The table also indicates that the number of "718" lactation reports from grade cows must be small in rela­ tion to the number of grades tested since the DHIA-IBM data contained practically equal numbers of purebred and grade records. About one-half of the grades reported on the latter program had no sire number indicated and, therefore, could not be included in the daughter averages. The small number of grades reported on the "718" forms is probably due to the difficulty in identifying the sire of a grade animal in herds where the records are poorly kept and less interest on the part of the herdowner and the supervisor in reporting grade records. Production for the grade cows that were reported is surprisingly good in view of the popular notion that purebred cattle are much superior to grades. There is a noticeable time trend in the average fat pro­ duction of the purebred daughters in Table 3. It seems to be due principally to a small but steady increase in the butterfat percentage, rather than to an improvement in the average milk yield. Further, there seems to be a definite jump in the average fat production in the middle of the table. This can not be ascribed to any given year but probably coincides with the increase shown in Table 1 between 1950 and 1951. 16 Daughter Averages for A.I. Sires The average production of the naturally sired daughters and the A.I. daughters of bulls used in stud service in Michigan is given in Table 4. Knowing the length of A.I. service for each sire made It possible to divide his female progeny on the basis of their birth dates into naturally sired A.I. daughters. (N.S.) and A few naturally sired daughters resulted from bulls being returned to natural service because of an unsatis­ factory breeding record In A.I. These were not Included. TABLE 4 DAUGHTER AVERAGES FOR A.I. SIRES No. of Sires 2X-■305-M .E . Fat Milk i No. of Daus. No. of Records 129 1917 4369 13,334 486 3.64 A.I. 139 2678 ^938 12,537 451 3.60 Grade daus. N.S. 62 267 485 12,357 445 3.60 A.I. 117 2420 3753 12,049 437 3.63 Purebred d aus . N.S. The table illustrates the expected drop in a sire's daughter average from his naturally sired to his A .1 .daughters. The decrease is more noticeable for the purebred daughters than for the grades. The 5,098 A.I. daughters averaged 12,305 pounds of milk; an(j 444 pounds of butterfat with a 3.6l per cent test. This would indicate some slight superiority of the A.I. daughters to the rest of the population. However, 40 per cent of the 17 A.I. daughters completed records during the 1953-1955 period. These years have the highest annual production averages of the 10 year period shown in Table 1. If the upward trend in the annual average reflects important contributions from improved feeding and management practices, some of the superiority of A.I. daughters could be due to environment rather than to heredity. This conclusion is supported by a study of A.I. progeny and their naturally sired contemporaries by Wadell (1957) which showed no significant difference in the production of the two groups. The evidence suggests that the bulls used in the A.I. units are not substantially Increasing the pro­ ductive level of the tested Holstein herds in Michigan. Methods of Selecting Sires for A.I. Service Three methods of measuring a sire’s breeding value prior to stud service were examined to learn the degree of association between these indicators and a sire’s A.I. daughter average. For the methods examined, the A.I. daughter average, compiled from the population of records studied, served as the dependent variable. The independent variables used were: (l) the average production of all N.S. daughters of a sire available from the data studied; (2) the daughter average given in a sire's first available natural proof; and (3) the amount that the naturally sired daughters’ production average was above or below that of the dams in the first available natural proof. The lactation records for all daughters of the A.I. sires were assembled and split into N.S. and A.I. groups, as discussed earlier. Sixty-seven sires with a minimum of 5 naturally sired 18 daughters and 5 A.I. daughters were available for the first comparison. Ninety-eight sires had a natural proof reported and had at least 5 A.I. daughters. second and third comparisons. These sires were used in the For each sire, the first avail­ able natural proof was used, since it was the one most likely to have been used as the basis for purchasing a sire for A.I. service. The sire proofs were almost always of DHIA origin. Occasionally, it was necessary to use an HIR proof or an unof­ ficial proof compiled by the stud. The pertinent data and the correlation and regression coefficients are shown in Table 5. TABLE 5 PRODUCTION AVERAGES AND CORRELATION AND REGRESSION COEFFICIENTS FOR THE NATURALLY SIRED AND A.I. DAUGHTERS OF STUD SIRES Method A.I. 12,330 446 1275 N.S. 13,220 479 5005 A.I. 12,260 445 963 N.S. 13,800 501 5005 A.I. 12,260 445 795 N.S. 98 l,090a r b r b .43 .30 • 4150 Fat Milk .37 .18 .30 .16 .35 .1 8 .33 .19 CM (3) 67 98 2X-305 -M.E. Milk Fat Q\ (2) No. of Daus. 00 • (i) Type Daus. No. of Sires 56a aRefers to average amount daughters exceeded their dams on the 98 natural service proofs. Table 5 indicates that there is some advantage in having information on all the naturally sired daughters before placing the bull in A.I. service. In practice, however, sire selection 19 committees cannot wait that long, and a bull is usually pur­ chased as soon as the first proof is available. These data do not demonstrate any ventional dam-daughter comparison type of the daughters. Lush and McGilliard advantage for the con­ of proof over the average (1955) discussed this point briefly and suggested that to use only daughters from tested dams wastes the information contained in the records of daughters from untested dams, and assuming that the untested dams averaged the same as the tested ones would not be far wrong. This review pointed out that Lush (1944), Rice (1944), and Robertson and Rendel (1954), among others, do not believe that using the d a m ’s records corrects a proof enough to be worthwhile. Legates (1954) remarked that in artificial insemi­ nation only about one-half of the dams have records, and this limits the it usefulness of the measure. In view of the above, appears that bulls can be chosen as accurately on the basis oftheir daughter average as on the conventional type of proof. In the discussion that follows, comparisons are made be­ tween the correlation and regression coefficients found in this study and those from previous work. Reports of the regression of future A.I. daughters’ production on that from either n tested natural service daughters or on the first n tested A.I. daughters are also compared. Schaefer (1953) reported a correlation between the daughter average of the "original" proof and the A.I. daughter average of 0.09 for 90 Holstein bulls used vania studs. in the New York and Pennsyl­ Sendelbach et a l . (1957) reported 0.6l for the regression of future A.I. progeny on the first <^5 A.I. daughters. 20 These authors stated that the results indicated that 20 to 30 A.I. daughters are sufficient to estimate future A.I. daughters with reasonable accuracy. The regression of future A.I. daugh­ ters on 25 N.S. daughters of a sire was approximately 0.30 for butterfat yield. For 10 naturally sired daughters, the regres­ sion coefficient was approximately 0.10. Henderson and Carter (1957) presented information to indicate that the regression of the future A.I. average on the present progeny average is roughly 0.10 for any number of naturally sired daughters between 5 and 100. When the records were adjusted for herd, year, and season effects, the value rose to approximately 0.20 for any number of natural service daughters above 10. Legates (195^) indicated that the regression of future daughters in artificial insemination on the first tested daughters, where the first ones were all in the same herd, was between 0.10 and 0.20, for any number of first tested daughters between 5 and 50. The regression value for future A.I. daughters on the first 25 daughters tested In many herds approximately 0.60. (as under A.I. conditions) was Even as few as 10 A.I. daughters gave regression coefficients ranging from 0.30 to 0.^5 In studies by Legates (195^)> Sendelbach et a l . (1957)* and Henderson and Carter (1957)* There is much evidence that the A.I. daughter average for a sire has a distinct advantage over the conventional dam and daughter proof (where the daughters in the proof are in 1 or 2 herds) for estimating the production of future A.I. daughters. Such a conclusion suggests that additional emphasis should be given to A.I. daughter averages when evaluating the breeding 21 values of stud sires. As previously indicated by other workers, a young sire testing program in A.I. would seem to offer the greatest possibility of steadily improving the genetic merit of dairy herds. Repeatability The 18,327 DHIA-IBM records were utilized to obtain an estimate of the repeatability among the records of an indivi­ dual cow. An estimate of repeatability was thought useful since it represents the upper limit of heritability. by Beal It was also used (1957) to correct the estimates of heritability to a single record basis. The estimate was calculated by analysis of variance components by Henderson’s Method I (Henderson, 1953). The components of variance are shown in Table 6. Milk was expressed in hundredweights to reduce the number of digits carried. TABLE 6 COMPONENTS OP VARIANCE FOR MILK AND PAT PRODUCTION Variance Component Value Fraction of Total Milk Fat Milk Fat Herds 246 34l4 0.34 0.36 Years 3 33 0.00 0.00 29 412 0.04 0.04 Cows within herds 202 2247 0.28 0.24 Residual 238 3339 0.33 0.35 718 9445 Herds X Years Total 22 The repeatability of milk and fat within a herd may be calculated from the ratio —— 5— C + E where C is the variation be- tween cows in the same herd and E represents the unanalyzed portion of the total variance. The repeatabilities were 0.46 for milk yield and 0.40 for butterfat production. The repeata­ bilities show excellent agreement with those reported in the literature. Legates and Lush (1954) reported that the herd component accounted for 39 per cent of the total variance, which was slightly larger than the estimates of 35 and 30 per cent pro­ vided by Plum (1935) and Lush and Straus (1942), respectively. The herd component found in this study agrees best with the one reported by Plum. There is general agreement from these studies that the differences among herds account for about one-third of the variation found in lactation records. Heritability Beal (1957) calculated the heritabilities for both the naturally sired and the A.I. populations from the data utilized in the present study. The estimates were determined from the intra-sire regression of daughters on dams as outlined by Lush (1940). The values were obtained from cow averages and corrected to a single record basis. The estimates of herita­ bility varied considerably between grade and purebred popula­ tions of A.I. sired daughters and their dams. The heritability figures for both milk and fat were of the order of 0.17 for the purebred A.I. daughters, while the corresponding values 23 for grade A.I. animals were 0.23 for milk and The number of records for grades was small. The bulk of the Information population. The heritability of was from the 0 . 0 5 for fat. non-A.I. sired milk yield was approximately 0.20 for grades and 0.28 for purebreds, while the comparable values for fat were 0.25 and O. 2 7 . Pooling the data gave heri­ tability figures of 0.264 ± 0 . 0 1 6 for milk and 0 . 2 6 2 ± 0 . 0 1 8 for f a t . The commonly accepted values for the heritability of milk and butterfat range between 0.2 and 0.3* There has been some question recently as to whether heritability Is the same for herds with different levels of production. Legates (1957) reported heritabilities for the Holstein, Jersey, and Guernsey breeds from herds at five different levels of production. None of the individual heritabilities were significantly different from the pooled value for the breed. The pooled values were 0.21 -t 0.06 for Guernseys, 0.22 ± 0.04 for Holsteins, and 0.25 ± 0.05 for Jerseys. Mason and Robertson (1956) reported significant differences in heritability of milk yield among high, medium, and low producing herds. These authors stressed that their results differed from most other published work. Paternal Sib Correlations The paternal sib correlations were computed for several portions of the data using Henderson's Method I (1953) in an analysis of variance components. Milk production was expressed in hundredweights, and the third digit was dropped from the fat production figures for ease of computation. Three estimates were made as follows: (1) An analysis of 26,700 naturally sired cows with 5 1 , 6 5 6 records is shown in Table 7 . TABLE 7 COMPONENTS OF VARIANCE FOR MILK AND FAT _____________ PRODUCTION FOR NATURALLY SIRED COWS Variance Component Milk Value_______ Fat Herds 202 28 0.31 0.33 Sires 90 12 0.14 0.14 356 45 0.55 0.53 Residual Total 648 85 Fraction of Total Milk Fat - The paternal sib correlation within herds would be equal to , where S is the variation among sires and E equals S + E the remaining unanalyzed variance. The paternal sib correlation for these data was 0.20 for milk and 0.21 for fat. The values were computed using cow averages rather than single records. Correction to a single record basis yields a paternal sib correlation of 0.16 for both milk and fat. Multiplying the paternal sib correlation by 4 gives an estimate for heritability of 0.64. The environmental correlation between paternal half- sibs In the same herd is likely causing an inflated estimate of heritability. Such a correlation could arise since all daugh ters of a sire are usually reared and milked during the same period of time. A change In environmental conditions from the time one bull's daughters are producing until the period when another sire's daughters are In milk could cause the correlation 25 (2) The paternal sib correlation was determined for 5,098 A.I. daughters with 8 , 6 9 1 records made between 1944 and 1955, Cow averages rather than single records were used. The ana­ lysis is given in Table 8 . TABLE 8 COMPONENTS OF VARIANCE FOR MILK AND FAT PRODUCTION FOR A.I. PROGENY Variance Component Value Fraction of Total Fat Milk Milk Fat Herds 211 30 0.35 0.39 Sires 29 2 0.05 0.03 Herds X Sires 42 5 0.07 0.07 313 40 0.53 0.52 595 77 Residual Total - - The paternal sib correlation is calculated from ________ S______ 9 where S and E are as previously defined and S + HS + H + E H and HS refer to the portions of the variance among herds and that due to the interaction of herd by sire groups, respec­ tively. The The values were 0.05 fa** milk and 0.03 for fat. appearance of a positive herd by sire component is unusual, for most published data Indicates that such a com­ ponent for interaction is zero or nearly so (Legates et a l ., 1956; Mason and Robertson, 1956). Bratton A report by Henderson and (1950) stated that the effect of the herd-sire Inter­ action was a more important factor in differences in production among A.I. daughters than were differences in the average transmitting abilities of the different sires. Later reports 26 by the same group of workers have not verified this contention. Wadell (1957), using first lactation records of A.I. progeny and their naturally sired contemporaries after the manner of Robertson and Rendel (1954), found no slre-herd interaction for Holstein data taken from the DHIA-IBM program. Since the pre­ sent study involved data collected over a ten year period, it is possible that management trends have been a complicating factor. (3) To circumvent possible time trends, an analysis was carried out on single records for the separate years of 1 9 5 3 , 1954, and 1955 from 2,631 records by A.I. progeny on the IBM testing program. Each record was assigned to the year in which the record began. Where cows started two records In a single year, one was randomly eliminated. There were 134, 1,045, and 1,452 single records for the respective years. A comparison of the individual years is shown In Table 9 . TABLE 9 DISTRIBUTION OF VARIANCE IN INDIVIDUAL YEARS 1953 Milk 1954 1955 Herds 0 .29 0.40 Sires 1953 Fat 1954 1955 0.38 0.26 0.39 0.4l 0.04 0.08 0.05 0.02 1 0 0 t—1 Source of Variation 0.10 Herds X Sires 0.03 0.02 0.00 -0.09 0.09 -0.07 Residual 0.68 0.48 O .5 8 0.75 0.46 0.63 The wide yearly differences observed might be expected with such small numbers of records, particularly for the 1953 sample. Disregarding the first y e a r ’s data, it appears that differences 27 between herds within years account for roughly 40 per cent of the total variation, while the residual component contained most of the remaining variance. The sire component was positive for both milk and fat for both years, although it varied con­ siderably in size. The herd-sire interaction term was most inconsistant, being small or nil for milk and being strongly positive for fat one year and just the opposite the following year. The data from the individual years were combined and reanalyzed. The results are given in Table 10. TABLE 10 VARIANCE COMPONENTS FOR A.I. DAUGHTERS FROM DHIA-IBM DATA Variance Component Value Fraction of Total Milk Fat Milk Fat Herds 244 33 0.36 0.38 Sires 52 5 0.08 0.06 Years 5 1 0.01 0.01 6l 8 0 .09 0.09 309 4o 0A 6 0.46 671 87 Herds X Sires Residual Total - - The paternal sib correlations were 0.08 for milk and 0.06 for fat. These data suggest a herd-sire interaction exists that is equal to or slightly larger than the sire effect. The possible causes of the different results for the herdsire component obtained by Wadell should be considered. (1957) and the present study Wadell's information was taken from re­ cords made In herds that had both A.I. and naturally sired 28 daughters. Where cows had more than one record, the additional records were randomly eliminated. There were 1,497 records studied by Wadell and 2,631 in the present Investigation. Neither sample is large and using a single record for each cow, rather than all records, might yield a different result. This is illustrated in Table 9 by the difference between the HS component from the 1,045 records in 1954 and that from 1,452 records in 1955* The results in the present study were obtained using mature equivalent production figures while Wadell's were determined from the differences between the A.I. daughters' production and that of their naturally sired contemporaries. The difference in method might account for the difference in the results. The problem needs further consideration when more data from A.I. daughters become available. An annual analysis could be conducted to determine if the interaction term changes as much as the present study indicated. The size of the com­ ponent relative to the sire effect needs to be clarified. Predicting future daughters' production from daughters. % tested Prom the data given in Table 8, the regression of future daughters on n tested daughters, each daughter in a different herd, may be estimated. Use of a sire's first n tested A.I. daughters to predict the production of his future A.I. progeny wouldapproximate the requirement that daughters be ina different herd. each of the The regressioncoefficient in terms of the variance components can be written as _______ 2_________ ^ , H + HS + E n + s From the data in Table 8, the regression is 29 -ft-■ for milk and n + 19.5 --- — ---- for fat. n + 37.5 Table 10 give the regressions of The values in n for milk and n + 11.9 _______ n + 16.4 for f a t . The regression coefficients obtained from the nearly 5,100 A.I. daughters whose records were expressed as averages (Table 8 ) are not in good agreement with other published results. Legates et a l . (1956) values for Holsteins were — —------ for n + 16.0 milk and — —— — — for fat. n + 13.8 Robertson and Rendel (1950) showed 7 that if heritability equals 0 .2 5 , the regression of m future daughters on n tested daughters is — ^ . Carter (1956) indicated that, with one record per cow, the value was —----- from the New Zealand data. 1 .In + 14 . 9 Carter also cited that Henderson of New York was using a factor of — ---- . n + 12 The values from Table 10 agree reasonably well with those published. Multiplying the paternal sib correlations for the single record A.I. data (Table 10) by 4 suggests that the heri­ tability values are 0.31 for milk and 0.23 for fat. Means and standard deviations. Table 11 summarizes the means and phenotypic standard deviations obtained for milk and fat from the three sets of data. These values are similar to others found In the literature and agree particularly well with those contained in the paper by Henderson and Carter (1957)* where the phenotypic standard deviation equaled 2,583 pounds for milk and 93 pounds for fat. Reasonable estimates of the phenotypic standard deviation of milk and butterfat production for the population studied are 2 , 6 0 0 pounds of milk 30 TABLE 11 MEANS AND STANDARD DEVIATIONS OF MILK AND FAT Population Data Used No. Cows Naturally sired daus. Cow Av. A.I. daus. A.I. d a u s . (IBM data only) St. D e v . Milk Fat No. Records Mean Fat Mf Ik 26,700 51,656 12,095 425 2546 92 Cow Av. 5,098 8,691 12,255 440 2440 88 Single Records 2,0 58 2,631 12,330 445 2590 93 and 90 pounds of butterfat. If heritability is 0.25, the genetic standard deviations are 1,300 pounds of milk and 45 pounds of butterfat. Non-genetic parameters Estimating the genetic gain possible in a dairy population requires a knowledge of the generation interval, the culling practices involved, the age distribution for the population, and the number of replacements available. In addition, for the males, the minimum number of progeny tested bulls required, the average life span of the tested bull, and the proportion of young bulls that survive the "waiting’' period after sampling must be specified. DHIA-IBM provided an opportunity to survey a large number of records that contain information related to some of these factors. Two hundred and sixty-nine Holstein herds, with an average of 2 7 . 3 cows per herd, had been on the IBM program since October, 1954- These herds were considered representative of all Holstein herds on the DHIA program in Michigan, and certain 31 of the following estimates were from these herds. Generation Interval. The generation interval for females was calculated from the average age at calving of the IBM re­ corded population in a given year. The age at calving was re­ ported on 1 7 , 6 6 2 completed lactations and was 5 2 .1 , 51.9, and 53.7 months for 1953* 1954, and 1955* respectively. The average age in months at calving for all lactations was 5 2 .9 , or 4 . 4 years. This estimate agrees with that made by Rendel and Robert­ son (1950) for the Shorthorn and Friesian breeds from British herdbooks, which averaged 4.6 and 4.8 years, respectively. Lush (1945) estimated the mean generation interval in dairy cattle at 4 to 4.5 years. Gannon and Hansen (1939) found the average age at freshening of cows in Iowa cow-testing associa­ tions to be 4.7 years. Percentage of cows removed annually. The percentage of cows removed annually is of interest since the more cows lost for involuntary reasons, the less freedom the dairyman has to cull on the basis of the animal's production and pedigree. Losses from the Holstein herds that had been on the IBM recording system since its start In October 1954 were tabu­ lated to estimate the number of cows removed each year. These herds had an average of 2 7 . 3 cows per herd, which is nearly the same as the 1956 state average (2 6 .6 ) for all herds in DHIA. Table 12 presents the annual removal of cows from herds for all reasons. 3^ TABLE 12 ANNUAL COW REMOVALS No. of Herds Qows Removed Av. No. Removed per Herd 1954-55 205a 1 ,5 6 7 7.6 2 8.0 1955-56 269 1 ,6 9 8 6.3 23.1 195S-57b 269 1 ,0 1 8 3 .8 ±7.7° Test Year % Removed aCow removal figures not available on the other 64 herds ^First six months of test year only cObtained by doubling the figure for the first 6 months The annual turnover for these herds was 26.3 per cent. The sample of herds was considered representative of all Hol­ stein herds on test, and the result strengthens the wellaccepted notion that a dairyman replaces one-fourth of his herd each year. The percentage of cows removed annually for all reasons from these herds falls in the range of the earlier esti­ mates of Seath (1940) and Asdell (1951). Seath found annual cow removals of 28.6 and 30.9 per cent for Iowa and Kansas herds, although It was suggested that severe drouth may have been a contributing factor. 17 states year. Asdell reported survey information from that showed an average of 21.9 per cent removals per Some of the Information was listed by states in Asdell's report. The comparable Michigan figure was 2^.4 per cent, calculated over an 18 year period. All figures quoted included the animals sold for dairy purposes. In the present study, a complete classification of the reasons for disposal was possible, and It is presented in Table 13. 33 TABLE 13 REASONS FOR COW REMOVAL IN 269 MICHIGAN HOLSTEIN HERDS Reason 1954-55a 1955-56 1956-57 Mean Dairy purposes 15.1 15.4 16.8 15.6 Low production 35.4 36.2 39.1 36.6 Physical injury 10.0 9.8 11.2 10.2 Mastiti s 8.2 6.7 7.3 7.4 Brucellosis 3.1 2.5 2.2 2.7 Tuberculosi s 3.5 4.0 1.8 3.3 Hard milker 2.8 2.6 2.0 2.5 16.0 16.4 14.0 15.7 5.9 6.4 5.7 6 .0 Sterility Died a 2 0 5 herds All cows removed from herds for reasons other than dairy purposes are presumed lost to the population. Although about one cow in six is sold for dairy purposes, many will be sold to non-tested herds and will be lost to the testing program. Some will prove to be such unsatisfactory producers will soon be resold for low production. that they For the above reasons, and for ease of calculation, it is assumed that cows removed from tested herds do not re-enter the tested population. Cow removals during each lactation. Information on the number of cows removed with each additional lactation has seldom been reported. Such information is needed when calculating the involuntary culling losses by the method of Rendel and Robert­ son (1950) which is presented later. Table l4 presents data from ll,4l6 Holstein lactation records reported by the DHIA- 3^ IBM program between October 1, 1956, and July 1 , 1957. Com­ parable Information from studies by Seath (1940), Asdell (1951), and Rendel e t a l . (1 9 5 1 ) are included. TABLE 14 COWS REMOVED DURING EACH LACTATION Age Lact. Present Study Rendel et a l . 2-3 1 2 5 .2 S 2 7 .7 a 2 8 .ia 31.6a 1 2 .6 a 3-4 2 27.2 35.4 31.3 30.4 15-5 17.2 4-5 3 26.2 35.2 23.2 27.2 23.1 26.4 5-6 4 28.5 32.2 25.0 26.8 29.2 35.0 6-7 5 31.9 36.7 27.4 26.1 30.1 34.0 7-8 6 36.0 ^9.4 26.3 29.6 21.6 35.2 8-9 7 40.9 47.1 31.8 32.9 18.0 36.5 9-10 8 54.2 50.6 34.2 42.2 14.3 10-11 9 49.5 74.3 38.4 44.4 Seath Iowa Kansas Asdell Kansas N.Y. 0 .0 - - a Percentage of cows removed each lactation of those surviving previous cullings Seath's and Asdell's reports were computed on a yearly basis, while those from the present study and from Rendel were figured on a lactation basis. The two systems could be con­ sidered equivalent, if one assumes that cows 2 to 3 years of age are In their first lactation, those 3 to 4 years of age are in their second lactation, and so on. The values reported in Table 14 include all cows removed from the herd, regardless of the reason for the cow being culled. In the present study, cows removed because of low production, hard milking, and cows sold for dairy purposes were considered 35 cases of voluntary culling. The remaining animals removed are examples of involuntary culling. category are mastitis, The major reasons in this sterility, and injuries. There are dif­ ferences of opinion as to what causes of removal should be classed as voluntary or involuntary; yet the division used here seems practical. Table 15 gives the proportions of animals removed for voluntary and involuntary reasons. by Asdell Similar data reported (1951) for Kansas and New York herds are included. TABLE 15 EFFECT OF LACTATION NUMBER ON THE PERCENTAGE OF VOLUNTARY AND INVOLUNTARY CULLING Pres. Study V o T T " " Invol. Kansas Vol. Invol. Age Lact. 2-3 1 18.6 6.6 3-4 2 18.4 8.9 11.3 4-5 3 16.3 10.0 5-6 4 15.5 6-7 5 7-8 4.1 2.0 4.0 11.0 6.1 15.7 7.3 15.4 11 .0 13.0 16.9 12.2 17.0 17-9 l 6 .1 15.8 16.8 13.2 19.1 14.9 6 14.8 21.2 10.8 10.8 16.8 18.7 8-9 7 19.2 21.7 9.0 9-7 15.1 21 A 9-10 8 23.2 31.0 5.8 8.5 - - 10-11 9 23.4 26.1 - - - - 4.5a 1.8a New York Vol. Invol aA discrepancy exists in Asdell's data. The total removals are reported as 12.6 per cent, yet a summation of the data into voluntary and involuntary classes yields the above fi­ gures, which total only 6.3 per cent. The New York figures, except for the 2-3 year old class, agree fairly well with the present study. The lower values reported for Kansas and New York at this age would suggest that 36 age and lactation number are not strictly equivalent. The generally lower values reported for Kansas can't be explained, except to suggest that the Kansas dairymen did not cull as severely as the Michigan dairymen at the earlier ages. Rendel and Robertson (i960) assumed a value of one-sixth for the natural mortality of dairy cows for the first three lactations, and a figure of one-third for all lactations be­ yond the third. Exactly what the authors meant by "natural mortality" was not indicated. "natural It seems likely that the term mortality" is identical with the definition previously stated for involuntary culling. If this is true, the assump­ tions made by the English workers are too extreme for our conditions. In the present study (Table 15 )» an arbitrary grouping of the first four lactations gives a mean "involuntary culling” percentage of 9 * 6 per cent, and the remaining lactations have a mean "involuntary culling" level of 23.2. Thus, it would seem more correct to use a value of one-tenth for the first four lactations, and a value of one-fourth for subsequent lactations. Number of available herd replacements. Another figure for which little published evidence is available is the morta 1 ity of females from birth to first calving. The 26 9 Holstein herds utilized previously were tabulated to determine the num­ ber of heifer calves born that survived to begin a milk record of their own. Only heifers born during 195^ were checked. In DHIA-IBM, each time a cow calves, the sex and identity of the calf is reported. Female calves are assigned a control number by the DHIA supervisor. It is intended that the control 37 number be permanent identification. When the heifer calves for the first time, she is identified by the control number assigned her at birth. With such an arrangement, the control numbers for the heifer calves reported during a given period of time can be matched with the control numbers reported for cows when they enter the milking herds to estimate loss of heifers from birth to calving. The number of heifer control numbers with no matching control numbers of cows in milk was taken to estimate the "mortalityM for heifers from birth to first freshening. "Mortality" is used here to mean any condition that prevents a heifer from calving in the herd that reported her birth. There were 2,553 heifers for which control numbers were reported during 1954. In April, 1957* 920 of these had been reported as having freshened. per cent of the heifers. This would account for 36.0 Of the 920 heifers which had calved, 825 were still in the herds and 95 (1 0 . 3 per cent of those recovered) were already dead or sold. Heifers born during 1954 were from 28 to 39 months of age at the time the study was made. Since the age at first calving covers a wide range, notall of the heifers would opportunity to calve and bereported. have had an The actual percentage recovered was adjusted to allow for this situation in the following manner. The distribution of age at first calving was determined from the DHIA-IBM data. Of 5*055 first lactation records, 5,037 had a range in age ofcalving from 16 to 48 months. Eighteen reported an age at months. first calving in excess of 48 The distribution is shown in Table 1 6 . 38 TABLE 16 DISTRIBUTION OP COWS BY AGE AT FIRST CALVING Age Mos . No. Cows 16 6 17 18 19 20 21 22 23 24 5 15 13 33 44 70 204 334 402 368 514 476 448 25 26 27 28 29 30 31 32 4n 318 285 % Total Cum. % 0.12 0.1 0 0 .30 0.26 0.65 0.87 1.38 4.04 6.61 7-95 7.28 10.17 9.42 8.86 8.13 5.46 5.64 Age Mos . No . Cows 0.1 0 .2 33 34 226 202 0.5 0 .8 1.4 2.3 3.7 7-7 14.3 22.3 29.6 35 36 177 139 95 71 61 39.7 49.1 58.0 66.1 78.1 78.1 37 38 39 40 4l 42 43 44 ^5 46 47 48 49+ 37 21 22 15 9 5 4 5 2 18 Cum, % % Total 4 .47 4.00 3.50 2.75 1.88 1.40 1.21 0.73 0.42 0.44 0.30 0.18 0.10 0.08 0.10 0.04 0.36 82.5 86.5 90.0 92 .8 94.7 96.1 97-3 98.0 98.4 98.8 99.2 99.3 99-4 99-5 99-6 99.6 100.0 By taking an actual count of the number of hei fers born in each month of 1954 and considering it equal to the cumulative percentage found in Table 16 for the corresponding age at first calving, it was possible to adjust the data to allow for the number of heifers born in 195^- and still in the herds that had not yet calved. For example, heifers born in January, 1954, were 39 months of age and had an actual recovery figure of 53.2 per cent. Table 16 shows that 97-3 per cent of all first calvers are fresh by the time they are 39 months of age. Dividing the percentage reported to have calved for January 1954 birth dates by 97*3 per cent will give the number even­ tually expected to freshen. This procedure was repeated for each month, and a summation of the twelve adjusted monthly 39 values gave a figure of 46.8 per cent (1 , 1 9 6 animals), which should eventually be reported. If the sample Is representative, it appears that for every two heifers born in tested herds, only one will survive to first calving. Frick and Henry (1956) reported studies on three experiment station herds that indicated from 66 to 83 per cent of all females born entered the milking herd. Even if it is granted that such figures are too optimistic for commercial herds, the difference between the present study and the reported values is still large. of three factors: The discrepancy is probably the result (l) heifer mortality, in terms of actual deaths, may be higher in commercial herds, (2 ) some heifers are sold to untested herds by the commercial tested herds, while institutions as a rule do not sell heifers, and (3 ) iden­ tification in institutional herds is reasonably positive from birth to death, while it is likely in the present study that the DHIA supervisors are improperly reporting control numbers on an Important percentage of the heifers. There is no reasonable check that can be made on the first two factors. However, a check of all animals entering the 2 6 9 herds for the first time during a five week period in April and May of 1957 was made to ascertain the relative accuracy of supervisor reporting of heifer control numbers. One hundred and eighty-four new cows in milk were reported during this period for the herds under observation. These were divided into four categories: (1) Sixty heifers were reported fresh with the proper control number, that is, the control number assigned to the 40 animal at birth. (2 ) Sixty-three animals calving for the first time had never been assigned a control number, either because the animal was purchased from a herd not on the IBM recording system or because the supervisor did not report heifer control numbers. (3) Twenty-eight animals were born before the herds joined the DHIA-IBM program. Again, these could have been purchased cows from untested herds or heifers calving for the first time that were born prior to October 1, 1954 and not given a control number. (4) Thirty-three heifers fresh for the first time were reported with a control number that differed from the control number assigned to them in 1954 as calves or were reported with a control number that corresponded to one assigned to another member of the herd. Therefore, a minimum of one-sixth of the animals checked were Incorrectly reported. If the animals that were never assigned control numbers are omitted, there were 33 out of 93 animals Incorrectly reported. In any event, most of the incorrectly reported animals would be missed In a system of matching heifer and cow control numbers such as was used originally to estimate heifer mortality. If, for every two heifers reported correctly, an additional one Is incorrectly reported, the original estimate of 47 per cent recovery could be increased by half (2 3 . 5 per cent). This would give an estimate of about 70 per cent of all heifers born that ultimately enter the milking herd. Such a figure would agree with the lower values reported by Prick and Henry (1956). 41 Rendel and Robertson (1950) stated that the number of ma­ ture heifers produced per cow in the herd per year lies between 0.35 and 0.40. With the normal sex ratio, this value could equal 0.50 only in the extreme case where all heifers born survived until they calved. Therefore, Rendel and Robertson are implying that 70 to 80 per cent of all the heifer calves born will calve in the same herd. The actual results from the present study (47 per cent of all identified heifer calves entered the same herd as first calvers) agree of the New Zealand Dairy Board closely with the findings (1958), since the figure reported by the latter study (23 heifers will freshen in the same herd for each 100 cows in calf) must be doubled to account for the normal sex ratio. However, since New Zealand conditions affecting the number of heifer calves selected to be raised until mature may differ from conditions in Michigan, the estimates may not be inter­ changeable. In view of other published data and the possibi­ lities that could have affected the estimate from the present study, it seems best to assume that 70 per cent of all heifer calves born are available as herd replacements. Minimum number of tested bulls needed. Publications by Bratton et a l . (1954) and by Foote et a l . (1958) reported that it Is possible to obtain between 25,000 and 50,000 progeny per sire per year at reasonable levels of fertility, using only 50 per cent of the semen produced. The use of frozen semen could increase still further the number of progeny possible from a single sire. Figures for 1958, published in the DHIA newsletter (March, 1957), showed that the average number of cows bred per 42 A.I. sire was 2 ,2 5 7 . No figures were given for the individual breeds, but it is unlikely that any stud is currently breeding an average of 10,000 cows per bull for even their most popular breed , Useful life span of tested bulls in A.I. and Robertson and Rendel Legates (1954) (1950) considered that the service span of a bull in A.I. is approximately three years. published data available, Becker and Arnold In the only (1953) reported 1 89 bulls were used in artificial service for an average of 2.68 years. Twenty-six of these were less than 6 years of age, 121 were from 6 to 10 years old, and 42 were more than 10 years old when they started stud life. the same authors A later report by (Becker and Arnold, 1957) listed 684 bulls with an average age at entry of 7-66 years and an average tenure of 2 . 8 7 years. Seventy-six Holstein bulls listed in the Michigan Arti­ ficial Breeders Cooperative Sire Books had an average age of entry into the stud of 6.4 years. Ninety-one Holstein bulls in service at the stud had an average tenure of 3.05 years. Therefore, for bulls that enter studs at an age equal to that when they would be naturally proven, three years of tenure is a suitable estimate. Becker and Arnold (1957) listed 70 bulls that entered service at 5.5 years (the probable age of sires selected for extensive A.I. use after an A.I. progeny test). The sires averaged 3-7 years tenure, indicating longer service for young sires in a testing program. Survival percentage for young sires. are available on this point. Legates No specific data (1954) and Robertson 43 and Rendel (1950) assumed a mortality figure of 20 per cent from the time the young sire is sampled until the decision is made as to what bulls are to be selected for extensive A.I. use. Carter (195^) also stated that 80 pef cent of the sires selected as young sires in the New York program were alive when their A.I. proof was obtained, while only 30 per cent of the sires selected on the basis of their natural service proof were alive when re-proved in artificial insemination. Calculating Genetic Gain In the Individual Herd Rendel and Robertson (1950) reported a method for calcu­ lating genetic gain In the Individual dairy herd. bertson and Rendel Later, R o ­ (1950) expanded the procedure to Include populations bred by A.I. bulls. Since the methods seemed suited to measuring improvement, possible genetic gains for dif­ ferent sized dairy herds were calculated by replacing as many assumptions as possible with estimates from Michigan. Rendel and Robertson visualize the improvement of genetic worth In dairy cattle as the sum of the individual gains ob­ tained from selecting the female and male parents of the next generation in each of the four possible ways that the parents can pass genes to their offspring; namely, from cows to breed female herd replacements; from cows to breed future sires; from bulls to breed future sires; and from bulls to breed female herd replacements. and I-gQ. These gains are denoted as Iq q , ^CB* ^BB* The genetic gain is the sum of the I's, divided by the sum of the generation intervals which are denoted as L q q , l CB' l BB* and l BC* Thus, the formula for genetic gain can be written as: £>G = -s-=s I» The actual genetic improvement possible with selection in a population is a function of the phenotypic selection differ­ ential and the heritability of the trait. In a theoretical approach, the assumption is made that the trait has a normal distribution. The genetic gain possible from each of the 4 genetic pathways visualized by Rendel and Robertson is given by the product of The symbol i represents the superiority in standard deviations of the proportion of the number available saved to be parents of the next generation and can be determined from Pearson's Tables (1931). The term, h^, represents the correlation between the genotypes and phenotypes of the population for the trait concerned. The (TG represents the genetic standard deviation for the trait and is easily calculated if the heritability and the phenotypic variance are known. In calculating genetic gain for a closed herd of 100 cows, Rendel and Robertson (1950) assumed that no progeny testing was done and that all replacements, both male and female, were bred from the herd. IBB and * B C * Without progeny testing, there is no gain from Therefore, the selection differential needed to be calculated only in selecting cows to breed cows (Iq q ) and in selecting cows to breed bulls (Iq b )* For the former, the fol lowing assumptions were made by Rendel and Robertson. (1) The natural mortality rates were considered to be onesixth for the first three lactations and one-third thereafter. (2) Selection was based on first lactation only, and all calves were rejected that were born from cows themselves culled 45 (3) The number of mature heifers produced per cow per year was estimated to be between O . 3 5 and 0.40. Therefore, from 70 to 80 per cent of all females born survived to become actual herd replacements. Rendel and Robertson calculated that the amount of culling possible would be either 30 or 39 par cent, depending upon the value used in (3) above. Assumptions were also made that the phenotypic standard deviation was 20 per cent of the mean yield, and that the genetic standard deviation was 10 per cent of the mean y i e l d . The expected genetic superiority of the selected cows above their contemporaries (Iq q ) was calculated as outlined previously. If the percentage saved is 70, i from Pearson's Table Is approxi­ mately 0.50. Thus, Iqq was 0.024 Y, where Y is the mean yield. When 6 l per cent were saved, Iqq was 0.032 Y. In calculating I0 3 , Rendel and Robertson cited Lush (1945) when they assumed that bulls were bred from the top 5 per cent 0.135 Y of the herd. I w a s 0.103 Y. Therefore, A G was *■£ 3 — . With the shortest generation interval possible, £L equaled 13 years, and A G was or heifers born entered the milking herd. when 80 per cent of the Or A G was 0.009 Y if only 70 per cent of the females entered the milking herd. If the preceding procedure Is followed, and the values obtained from Michigan herds are substituted when they differ from the English assumptions, the result changes little. A new culling percentage must be established, since the Michigan involuntary culling rates are of the order of onetenth for the first four lactations and one-fourth thereafter. The lower of the two values (70 per cent) for heifer replacements 46 used by Rendel and Robertson seems appropriate in view of the present study and other published data. The percentage that can be culled without reducing herd size was calculated by the method of Rendel and Robertson The number of mature (1950). heifers produced per cow per year for the herd in this case is 0.35. The replacements available (f-jJ will be equal to 0.35 (?2 + f 2 + f 3 + ^4 •* *) > the first f2 being calves born to first lactation cows that are subsequently accepted into the herd. The second f2 , £ 3 , fjp etc., refer to the proportion of the herd in their second, third, fourth, etc. lactations. f 3 + f . ..) . The f2 's can be added and then f-^= 0.35(2f 2 + However, after involuntary cow removals by lacta­ tion stabilize for a particular herd, f^ will equal 9 / 1 0 f2 , = 9 / 1 0 f 3 , fg = 9 / 1 0 fij., fg = 3/4- £5 , etc., so that the later proportions can be expressed in terms of f2 , giving f-j_ = 0.35 (6 .6 l f2 ) . With a 10 per cent loss from involuntary culling in the first lactation, the latter figure becomes 9/10 (0.35 x 6 .6 l f2 ). From this number, the cows kept for a second lactation are selected. 0.9 (6 .6 1 ) (0.35) - 1 = x 0.9 (6 .6 1 ) (0.35) The percentage culled is then 1________ 5-959 (0.35) Thus, the percentage saved equals 0.48 and = 1 0.168 0.35 I is = 0 52 equal to 0.83. Assuming the same heritability figure (0.25) and the same (j"G of 0.10 Y, ICq would equal 0.042 Y. would remain un­ changed, and the sum of the I's would be 0.145 Y. present data, Lqq is equal to 4.4 years, With the is at least that much and possibly more, if bulls are not kept from first caivers. Rendel and Robertson included Lgg and IrgQ in the formula 47 when calculating genetic gain in a closed herd. LgB and LgC would each equal an average age of 2 to 2 . 5 years. Thus, the sum of the generation intervals in the example would be 14 to 15 years. AG If the latter value for generation interval was chosen, would equal 0.010 of the average yield. plausible generation interval only increase to 0.011 Y. If the shortest (13 years) was used, £ G would Therefore, in herds that are selecting male and female replacements solely on the basis of the milk: records of their dams, the annual genetic improvement expected is about one per cent of the average yield. The Michigan data yielded almost identical values with those reported by the English, with the exception of the natural ’’mortality" figures for cows of different ages. However, use of the Michigan "mortality” values had little effect on the out­ come of the calculation. The increased amount of voluntary culling that could be done did not result in any appreciable improvement In the rate of genetic gain over that obtained in the English study. The gain that can be made by selecting the dams of herd sires Is considerably more Important and over­ shadows the contribution made by selecting cows to breed future female replacements. Rendel and Robertson also pointed out that nearly all herds bring in new bulls whose breeding value is less well known than if they were home-bred, and, consequent­ ly, the rate of genetic improvement would be reduced. situation exists for Michigan herds. The same 48 Genetic Gain with Progeny Testing If selection In dairy herds that are not progeny testing results in an increase in average yield of about 1 . 0 per cent per year, it is worthwhile to consider whether progeny testing would Improve their rate of genetic gain. A general outline of the proposed progeny testing plan for young sires in various sized herds is given below. In individual herds of conventional size, young bulls would be saved for sampling from the highest ranking cows. Chance would be an important factor In small herds since bull calves might not be produced by the top cows in the herd for lengthy periods of time. A suitable method of ranking the cows in the population may involve the use of devices to correct cows’ per­ formances for non-genetic influences, the choice and use of various characteristics in some selection Index, and the use of information from close relatives. Enough young bulls would be saved for sampling to allow further selection of the bulls based on results of the test. The young bulls would be sampled when approximately one year of age. All healthy female calves would be raised and the culling of individuals delayed until a heifer had been in milk for 60 to 90 days. The examples that follow assume that all culling is done on the basis of the first lactation production. However, in practice this is not the usual procedure, for culling of producers of doubtful ability is often delayed for additional lactations. In view of the relative contributions from the selection of cows to produce female replacements and from the 49 selection of the dams of herd sires, exactly when the culling among the former group is done may not change genetic gain mater­ ially . The young sires, after sampling, would be leased or main­ tained by the owner until production records were available on the daughters. When the sampled sires were about 5 years old, the decisions could be made as to which sire(s) would be returned to service to breed those cows not needed for testing young bulls, and particularly, to be mated to the best females avail­ able in order to obtain future young sires for sampling. The principles would apply equally well to A.I. populations. A.I. units could contract with herd owners to breed the top cows to the best of the tested bulls in the stud. The program could be handled on a state or regional basis for large populations of production tested cattle using mechanized computing equipment to routinely rank the outstanding males and females in the population. In the present study, rates of genetic improvement ex­ pected from various plans of progeny testing were calculated for individual herds of 2 5 , 5 0 , 100, and 200 cows. Plans were also determined for A.I. populations of 2,000 and 10,000 cows. Since the study is concerned with the maximum genetic gain expected with progeny testing, Figures 1 through 7 were plotted using statistics which would favor progeny testing. For example, the estimate that one milking heifer would result from every 3 cows In calf was used • Also, — 12 was used for the ren + 12 gression of the production of future daughters on the production of n tested daughters of a sire. The involuntary culling rates found from the Michigan data were used. to modify the generation intervals or the and Robertson. There was no reason CTG assumed by Rendel In the practical situation for a specific stud (Figure 8 ), the lower estimate of one milking female for every 4 cows in calf was used. The general approach is the same as that utilized for herds not progeny testing. In addition, the contributions from IBB and IB G > as well as the generation length Lgg and LgG are added. The assumptions made and the steps followed are outlined. IGC will not be affected by the sire testing plan and, therefore, will be the same (0.042 Y) as for selection in a closed herd. The gain from cows to breed bulls, Iq b j is deter­ mined in the same manner as before, although the percentage of bulls saved now depends upon the number of tested bulls needed simultaneously, their length of service life, and the facilities available for progeny testing. With present-day A.I. techniques, a single sire is capable of breeding at least as many cows as are in the sizes of popu­ lations studied. assumption. Figures 1 through 6 have been drawn on this Actually, the tested sire saved for extensive use would need to breed only that portion of the population not mated to young sires. This could vary from a situation where the tested sire would be used only to breed future young sires, to the case where lack of confidence in young bulls might cause 90 per cent or more of the cows to be mated to tested bulls. In line with B e a l ’s (1957) findings and for ease of cal­ culation, a heritability of 0 . 2 5 was used for calculating IC B . 51 Although it was assumed that all young bulls are sired by adequately tested bulls, it is recognized that such a condition would not likely occur in individual herds, and could not exist in an A.I. population for at least 5 years. IBB can be cal­ culated from the selection intensity (proportion of bulls chosen for extensive use from those selected for sampling), and the accuracy of the progeny test, based on the number of tested daughters sired by each young bull. The paternal sib data provided — — _ to predict the production of a sire's future n + 12 daughters from n tested daughters. An estimate of — 2--” n + 15 was employed by Robertson and Rendel (1950). The difference between the two was found to be of little consequence for when varying numbers of daughters were checked. The n depends upon the number of bulls being sampled and the number of cows that must be bred in order to produce a mature heifer with a milk record of her own. The present study and published reports indicate that from 3 to 4 cows must be in calf in order to obtain one tested heifer. In all examples except the last, one heifer for three cows was used. I i s the percentage of the population mated to tested bulls times IBB since once the stud has converted entirely to the plan, only tested bulls are used to breed the cows not needed for testing young sires. The tested bulls had been selected as the best of former groups of young sires and would be gene­ tically superior to other sires by the amount, lB B . The genetic standard deviation (0.10 Y) and the individual generation intervals (Lq q , etc.) were not changed, since the Michigan data led to estimates almost exactly the same as those 52 assumed by Robertson and Rendel. The improvement per year with progeny testing is shown by the following formula: = 00 + CB * LCC + l CB + BB + + BC * Results in Individual Herds The series of figures which follow depict the annual gain expected from progeny testing in different sized cow populations. Figures 1 through 4 illustrate the values for herds of 2 5 , 50, 100, and 200 cows. A reference line is drawn on all graphs to mark: the level of genetic progress expected without progeny testing but with selection of cows and young bulls. Varying numbers of sires were sampled, depending upon the size of the herd. The percentage of the herd bred to young sires was also varied. If all cows are bred to young sires (the point on the graph where the percentage bred to young sires equals 1 0 0 ), there are no cows left to breed to tested bulls. Therefore, there is no gain from progeny testing, and IBB and IBG are zero. Since ICG is unaffected, only ICB can be manipulated by sampling different numbers the best ofyoung bulls. 2 bullcalves For example, to save only from a population will give agreater selection intensity (hence a larger i value) than to save more than 2 from the same population. IGB will be the largest in the former case, and the annual rate of improvement will decline when larger numbers of bulls are saved for sampling, since the selection differential decreases. 53 Figure 1. Genetic improvement with progeny testing in a 25 cow herd, saving one of n sires sampled. 1,7r 1 .6- 1.4 1.3 1.1 1.0 Level of annual progress without progeny testing 0 .9 0 .8 per year in cent 1.2 per of average yield 1.5- improvement 0.6 Genetic 0 .7 0.4 0.5' 0.3 0.2 No. of young sires sampled 2 0.1 20 J 40 60 80 100 Percent of cows bred to young sires cent of average yield Figure 2 . Genetic improvement with progeny testin'? in a 30 cow herd, saving one of n sires sampled. Genetic improvement per year in per Level of annual progress without progeny testing 0.4 — 0.2 No. of young sires sampled 20 40 80 Percent of cows bred to young sires 100 55 cent of average yield Figure 3. Genetic improvement with progeny testing in a 100 cow herd, saving one of n sires sampled. 0.8 0.6 Genetic improvement per year in per Level of annual progress without progeny testing No. of young sires sampled 0.2 20 60 80 40 Percent of cows bred to young sires 100 56 cent of average yield Figure 4. Qenetic improvement with progeny testing in a 200 cow herd, saving one of n sires sampled. 0.7 Qenetic Improvement per year in per Level of annual progress without progeny testing No. of young sires sampled 20 40 60 80 100 Percent of cows bred to young sires 57 Twenty-five cow h e r d . For a herd of 25 cows, progress with progeny testing is largest when all cows are mated to 2 young bulls, and the better of the two is saved. If progeny testing is to be used, nearly all of the herd should be bred to young sires. Although situations are presented in which 2, 3, and 4 bulls are sampled each year, it is not practical to suggest that more than 2 sires could be progeny tested in a single year in a 25 cow herd. However, since all points plotted fall below the reference line, progeny testing would not be the most efficient plan for a herd of this size. The dairy­ man then has a choice of saving sons of his best producing cow(s) and turning them over rapidly to keep the generation In­ terval as short as possible, or to utilize such discretion as he can in picking A.I. sires to breed his cows. Unless the herd owner is certain his cattle are genetically superior to the stud sires, the latter seems to be the more practical sug­ gestion, especially in view of the possible detrimental effects of Inbreeding in so small a h e r d . Fifty cow herd. A similar situation exists for the fifty cow herd, although the calculated gain from progeny testing rises above the reference line on Figure 2, when at least 60 per cent of the herd are bred to young sires. The optimum number of bulls to sample Is 3, if one is saved for future use, although the advantage over sampling 2 or 4 Is slight. The additional gain beyond what selection without progeny testing could produce is so small, that It is doubtful that it Is worth the labor and expense of keeping an older sire. Therefore, the suggestions made for the 25 cow herd would apply also to the fifty cow herd. One hundred cow herd . The 100 cow herd could exist either as an individual unit or as several smaller herds desiring to follow the same breeding program. Figure 3 shows that progress with progeny testing exceeds that of using only young untested bulls by as much as 20 per cent when 60 per cent or more of the females are mated to young sires, and one tested sire is selected from the 3 or ^ sampled. In all cases, progress increases as the percentage of cows bred to young sires increases. Sampling 3 or 4 sires has an advantage over sampling 2 or 8 sires. Sampling 8 likely would be ruled out for economic reasons, but the dairyman must decide whether he would prefer to sample only 2 bulls per year and take slightly less annual progress or to sample 3 or 4 bulls. If several small herds are involved, the number of bulls sampled might depend upon the number of herds that desired to breed their cows naturally. Since progress from progeny testing offers only a slight gain over selection without progeny testing, factors other than the absolute annual genetic gain may influence the dairyman's breeding plan. Two hundred cow h e r d . The range in improvement possible with a given set of conditions widens as the larger herds are explored. In a herd of 200 cows, nearly all values for progeny testing are above the 1 per cent reference point. The few that are not involve testing more bulls than Is economically feasible. Figure 4 Indicates that peak progress is realized, when 4 young sires are sampled, and one Is saved for further service. Progress declines when multiples of 4 bulls are sampled. However, sampling 3 young sires is almost as fruitful as.sampling 4. Whether 3 or 4 should be sampled might depend 59 upon the number of herds involved In a small breeding ring and upon the ease with which lease agreements can be made with other dairymen for the young sires during their '’waiting" period. In either case, genetic progress Is from one-fourth to one-third above that possible without progeny testing, as long as at least 30 per cent of the cows are mated to young sires. Dairymen concerned with cow populations of this size should seriously consider the use of progeny testing, unless they are satisfied that the A.I. units are doing as well In improving the genetic merit of their bulls. Results in A.I. Populations Small A.I. populations. Figure 5 illustrates the progress that can be made for a small A.I. population of 2,000 cows. All cows are production tested. All levels of genetic gain are well above the reference point, and the majority of the points fall between the 1.6 and 1.9 per cent levels. With large popu­ lations, it is more efficient to sample larger numbers of young sires. Although the optimum is realized when 20 young sires are sampled and only 1 Is saved, nearly equal progress occurs in other instances, when from 50 to 80 per cent of the cows are used to test young sires. Little would be lost if one were to sample 10 young sires on 40 per cent of the cows. However, it is doubtful that a commercial stud could be persuaded to accept any plan except the one involving the lowest bull sampling ratio (3 sampled and 1 saved). progress of about 1 . 6 Figure 5 indicates that per cent can be made when mating from 30 to 70 per cent of the population to young sires. 60 Figure 5. Genetic improvement with progeny testing in an A.I. population of 2,000 tested cows, saving one of n 2.4*-sires sampled. & " 2 .34- 2.1 2.0 1.8 1.7 ft 1.5* 1.4- Genetic per 1.6 improvement year in cent 1.9 per of average yield 2.2 ( 1.31.2 — 1.1 Level of annual progress without progeny testing 1.0 No. of young sires sampled -------3 --------5 0.9 - 0 .8 - t 20 - 25 - X 20 X ^0 x 60 X 80 Percent of cows bred to young sires 4 100 61 Figure 6 . Genetic improvement with progeny testing in an A.I. population of 10,000 tested cows, saving one of n 2.4feires sampled. 2.3 r of average cent 1.8 1.7 1 .6 - 1.9 1-5 Genetic improvement per year 2.0 per 2.1 in yield 2.2 1.4 1.31.2 1.1 l.d Level of annual progress without progeny testing No. of young sires sampled -------3 o. 10 20 o.a t ___________ I___________ 2? 20 40 60 _____ 100 80 Percent of cows bred to young sires Figure 7. Genetic improvement with, progeny testing in an A. X. population of 10,000 tested cows, saving 2 of n sires sampled. — Genetic improvement per year in per cent of average yield 2.2 Level of annual progress without progeny testing N o . of young sires sampled 10 20 40 20 100 Percent of cows bred to young sires 63 Figure 8. Genetic Improvement with progeny testing in an A.I. population of 150,000 cows (10,000 tested), saving of n sires sampled. 2.0 year In per cent of average yield 2.2 '/ 1.2 Genetic improvement per // Level of annual progress without progeny testing No. of young sires sampled 50 100 0.8 125 100 20 Percent of cows bred to young sires 64 This example is the first of those described to show a clear-cut advantage for progeny testing, if studs could be persuaded to undertake such a scheme, even at the lowest level of sampling, individual herds could well afford to put their bull breeding problems in the hands of the stud and dismiss the notion of progeny testing sires for their own use. Unfor­ tunately, there are few young sire programs of this type operating at present. Large A.I. populations. The values in Figure 6 illustrate the maximum progress possible under the proposed plan for an A.I. population of 10,000 production tested cows. A single tested sire is assumed capable of breeding all cows not needed for testing young bulls. Using the same bull sampling ratios as in Figure 5> it is evident that much additional improvement results when the effective cow population is increased. The only deviation from the conditions existing in Figure 5 is that five times as many heifers are available for testing young sires. Therefore, the accuracy of the progeny test increases for each bull sampled. If many bulls are to be tested, peak efficiency in the scheme occurs when 30 to 50 per cent of the population are mated to young sires. The peak moves towards the left as fewer young bulls are sampled. At the more practical levels of sire sampling, Figure 6 substantiates the estimates of Le­ gates (1954) and Henderson (195*0 for the percentage of the population that needs to be mated to young sires for maximum genetic improvement. Figure 7 is Inserted to show the effect of increasing the number of tested bulls needed. All conditions are the same for 65 Figures 6 and 7, except that the best 2 bulls are saved for future use in the latter case. This alteration reduces the genetic progress possible from 10 to 20 percent. Neither of the estimates approach the actual situation, since complete re­ cording of production data is specified in both instances, and actually many untested cows are also bred by A.I. bulls. The majority of the dairy states in this country have from 5 to 15 per cent of their cows on test. Only this portion of the population is available for testing young sires, yet A.I. studs must keep sufficient bulls to breed the unrecorded cows as Consequently, more than 1 or 2 tested sires would be re­ well. quired for large studs. The number of tested bulls needed seems to be the primary factor in establishing the limits of genetic gain with progeny testing. The number of cows to be bred determines the number of tested bulls needed, and the proportion on test determines the number of daughters available for progeny testing young sires. A practical example. Figure 8 represents the progress thought possible for a practical situation. The following ex­ ample conforms to the present situation for the stud doing the largest volume of business in Michigan. It is assumed that the stud Is breeding 150,000 Holstein cows annually. About 6 2/3 per cent of all cows in Michigan are in the DHIA testing program, and there is evidence that the percentage is no higher for A.I. herds. Therefore, the effective population is 10,000 cows, but sufficient bulls must be kept to breed all 1 5 0 , 0 0 0 cows. Assuming each bull can get 10,000 cows in calf per year, 66 a minimum of 15 tested bulls are needed. If the service life of an A.I. bull is 3 years, an average of 5 bulls must be re­ placed each year. In order to compare with the results from earlier examples, the calculations were made for situations Involving comparable bull sampling ratios. Hence, It would be necessary to sample 15* 25, 5 0 , 1 0 0 , and 125 young sires each year. Other statistics which were different than those used in Figures 1 through 7 are (l) a lower estimate that 4 cows in calf to a sire will yield 1 tested daughter was adopted, and (2 ) the generation interval for young sires was set at 2 . 5 years. Values obtained under these conditions are appreciably below those reported for the optimum situation (Figure 6 ), if a large number of bulls are sampled. However, when one of 3 bulls sampled is saved, the difference between the optimum conditions for Figures 6 and 8 is only 0,13 per cent of the mean annual yield. The optimum procedure under the conditions described Is to breed 40 per cent of the cows to young sires and to sample 10 young bulls for each one saved for further use. Such a pro­ cedure would yield an average of 20 tested heifers per bull. If 5 bulls are needed each year, the stud would need to sample 50 young sires annually. Thus, the optimum for maximum progress Is not economically practical from the stud's point of view. Sampling larger numbers of young bulls (for the ratios calculated) gives slightly less progress than with the 10 to 1 ratio. Ruling out the testing of large numbers of bulls brings one to the question of balancing the testing of fewer bulls against the additional gain when 50 circ tested. Testing 3 or 5 young bulls for each one saved is less than optimum, but it is closer to the practical situation. If 5 young bulls are sampled for each one saved, the op­ timum gain occurs when 40 per cent of the cows are bred to young sires. The optimum point shifts to 30 per cent, when only 3 young bulls are sampled per bull saved. The latter situ­ ation Is the most realistic compromise between economy and genetic gain. 0.2 This yields a maximum rate of Improvement that is about per cent of the average annual yield below that possible when 10 young bulls are sampled for each one saved. It Is un­ likely that the additional gain resulting from sampling the larger number of bulls would balance the expense of caring for them. Even if the bull sampling ratio is reduced to 3 to 1, the annual increase In genetic merit of the population would equal about 1.6 per cent. In terms of the present actual production average, this would amount to approximetely 175 pounds of milk and 6 pounds of fat per year. Estimates of progress in actual herds fall short of this amount, except for the one reported by Harvey (1953)• The recommended practical procedure would require that 3 , 0 0 0 cows (30 per cent of the total number tested) be bred to young sires. This number of females should produce 750 tested heifers for sire testing purposes. If 15 young bulls are sampled annually, each would have 50 milking daughters with records. Such a number would be sufficient to satisfy those who require upwards of 30 to 35 A.I. daughters before expressing 68 confidence in an A.I. proof. There are two major possibilities that would allow an additional increase in the rate of genetic gain, above that postulated for this particular situation. Increasing the per­ centage of cows on test would provide more tested daughters per young sire sampled. The accuracy of the progeny test would Increase, and both Igg and Igc would Increase. Second, any improvement in A.I. techniques, which permits a sire to breed more cows per year than is now possible, would reduce the number of tested bulls needed. This will increase the I value and, hence, Iq ^, since fewer bull calves need to be saved for sampling. If fewer young sires of the total number available are sampled and other conditions remain the same, more tested heifers will be available per young sire, and the accuracy of the progeny test used in calculating I-gg will increase, and, therefore, so will IgB and i^ . Bratton et a l . (195^) have suggested that it is possible to obtain from 25,000 to 50,000 progeny from an A.I. sire in one year. Foote et a l . (1 9 5 6 ) reported that recent semen production studies by Almquist and co-workers of Pennsylvania indicated that this potential may be even greater. With these rates of insemination, the stud under study would need to maintain a battery of from 3 to 6 tested sires, and would have to select only 1 or 2 replacements each year. The gains possible under such conditions are the same as those from the examples pre­ sented in Figures 6 and 7Economics. The economics involved In the proposed program have been ignored except with regard to the cost to the stud of maintaining large numbers of young bulls. While the studs are interested in breeding better cattle, their available facilities for progeny testing and the expense Involved force them to compromise between the improvement expected from sampling larger numbers of bulls and the additional costs. Considerable improvement in genetic gain is possible when sam­ pling ratios of 1 to 3, ^ , or 5 are employed. Sampling a larg number than this for each tested bull needed will benefit comm cial studs only if dairymen are enlightened as to the value of A.I. progeny tests and as this results in an Increased volume of business. 70 SUMMARY The 65,392 records from 34,073 Michigan DHIA Holstein cows averaged 12,237 pounds of milk and 435 pounds of fat on a 2X305 day-M.E. basis. The 5,098 daughters of 187 A.I. sires aver­ aged 12,305 pounds of milk and 444 pounds of fat. The evidence suggests that the bulls used by the A.I. studs have not substan­ tially increased the milk producing ability of the tested Hol­ stein population. The yearly increase in the average production for the 19451955 period for Michigan DHIA Holsteins with the present methods of sire selection averaged 0 . 9 per cent of the mean annual yield. There was no advantage in using the conventional paired dam and daughter natural proof instead of the simple daughter average for predicting the production of the A.I. progeny of stud sires. Correlation coefficients between the difference In production of the daughters minus their dams and daughter average for 0.33 for fat. 98 Holstein sires were the A.I. 0.35formilk and The regressions were 0.18 and 0.19 for milk and fat, respectively. Correlations between the simple daughter average from the first available natural proof and the sire's A.I. daughter average were 0.37 for milk and 0.30 for fat. The regressions were 0.18 for milk and 0.l6 for fat. Repeatabilities were 0.46 for milk and 0.40forfat from the 18,327 mechanically computed records. Estimates of heritabillty obtained from the paternal halfsib correlations found for 2,631 single records of 2,058 A.I. 71 progeny were 0.31 for milk and 0.23 for fat. From the same data, the regressions of a sir efs future A.I. daughters’ production average on the average production of his first n tested A.I. daughters were — for milk and — -— _ for fat. n + 12 n + 16 Statistics for generation interval, percentage of cows removed annually, cow removals by lactation, reasons for re­ movals, and heifer mortality agreed well with previously published estimates . In herds of less than 100 cows, progeny testing was less efficient In improving the rate of genetic gain than the use of young sires selected on the basis of their dams’ production. Progeny testing in herds of 100 and 200 cows had a slight advan­ tage over the latter method. Progeny testing in A.I. populations of 2,000 and 10,000 production tested cows gave evidence of making annual genetic progress of from 1.5 to 2.3 per cent of the average annual yield. For a practical situation Involving the current operating techniques of the stud breeding the largest number of cows in Michigan and the percentage of cows on test, progress of from 1.7 to 1.8 per cent of the average annual yield is postulated. Further improvement seems possible if a larger percentage of the cow population were tested and if more efficient use were made of the bulls In A.I. service. The use of a progeny testing plan with young sires in an artificial insemination unit seems to offer a more reliable method of improving the genetic merit of the dairy population than does the present system of selecting sires for A.I. use on the basis of their natural proofs. LITERATURE CITED Albrectsen, Raymond. 1953Genetic analysis of young sires. Proc . of 6th Ann. Conv. of National Assoc, of Artificial Breeders, 1 2 9 . Asdell, S. A. 1951. Variations In amount of culling in D.H.I.A. herds. J. Dairy S c i . 34:529. Beal, V . C ., J r . 1957* Comparisons of mates of sires used artificially and naturally and the heritability of Michigan DHIA Holstein production records. Unpublished M.S. thesis, Michigan State University Library, East Lansing. Becker, R. B., and p. T. D. Arnold. 1953Tenure and turnover of desirable dairy bulls In artificial studs. J. Dairy Sci. 36:579________ a n d ________ . 1957* Tenure of dairy bulls in artificial service. Dairy Sci. 40:622. J. Bratton, R. W., R. H. Foote, and C. R. Henderson. 1954. Semen production and fertility of mature dairy bulls ejaculated either once or twice at eight-day intervals. J. Dairy Sci. 37'* 1444. Cannon, C. Y., and E. N. Hansen. 1939. Expectation of life in dairy cows. 2 2 :1025. J. Dairy Sci. Carter, A. H. 1956. Some aspects of dairy sire selection. Zealand S o c . of Animal Prod., 15:77* Proc. of New Carter, H. W. 1954. Improving the genetic merit of our dairy cattle. M i m e o . Cornell Univ. 9 PP* Dickerson, G. E., and L. N. Hazel. 1944. Effectiveness of selection on progeny performance as a supplement to earlier culling In livestock. J. A g r . Research, 69: 459* Dillon, W. M., Jr., W W. Yapp, and R. W. Touchberry. 1955. Estimating changes In the environment and average real producting ability In a Holstein herd from 1901 through 1954. J. Dairy Sci. 38 :616 . 72 Foote, R. H., C. R. Henderson, and R. w. Bratton. 1955. Testing bulls in artificial insemination centres for lethals, type and production. Proc. of 3rd Inter­ national Congress on Animal Reproduction, Cambridge. Frick, G. E ., and W F . Henry. 1955. Production efficiency on New England dairy farms. V. Adjustments in obtaining dairy herd replacements. New Hampshire A g r . Exp't. Sta. Bull. 430. Harvey, W. R. 1953. Genetic and environmental changes in the University of Idaho Holstein and Jersey herds. Proc. 34th Ann. Meeting, Western Division of Amer. Dairy Sci. Assoc. Henderson, C . R. Estimation of variance and covariance components. 1953. Biometrics. 9:226. 195 ^. Selecting and sampling young sires. Proc. of 7th Ann. Conv. of National Assoc, of Artificial Breeders, 93. Henderson, C. R., and R. W. Bratton. Annual report of the Dean. 1950. Sta. p. 101. Cornell Univ. Agr. E x p ’t. Henderson, C. R., and H. W. Carter. Improvement of progeny tests by adjusting for herd, 1957. year, and season of freshening. J. Dairy Sci. 40:638. Henderson, C.R., and R. S. Dunbar. Improving dairy herds by "sampling” bulls. 1952. Research. 18:3* Farm Laben, R. C ., and H . A . Herman. Genetic factors affecting milk production in a selected 1950. Hoi stein-Friesian herd. Mo. Agr. E x p ’t. Sta. Research Bull. 459* Legates, 1954. 1957'.' Legates, 195^. . E. New tools as aids in sire selection. Mimeo. pre­ sented at 49th Ann. Meeting of Amer. Dairy Sci. Assoc. State College, P a . Heritability of fat yields in herds with different production levels. J. Dairy Sci. 40:631* . E ., and J . L . L u s h . A selection index for fat production in dairy cattle, utilizing the fat yields of the cow and her close relatives. J. Dairy Sci. 37:744. 74 Legates, J. E., F. J. Verlinden, and J. F. Kendrick. 195o. Sire by herd interaction in production traits in dairy cattle. J. Dairy Sci. 39:1055. Lush, J. L. 1940. Intra-sire correlations or regressions of offspring on dam as a method of estimating heritability of characteristics. Proc. 33rd Ann. Meeting Amer. S o c . of Animal Prod. 2 9 3 . 19447 The optimum emphasis on d a m s 1 record when proving dairy sires. J. Dairy Sci. 2 7 :9 3 7 . 1945- Animal Breeding Plans. Iowa State College Press, Ames, Iowa. 443 pj7 1949- Estimating the breeding value of young bulls. Mi meo . presented at 44th Ann. Meeting of Amer. Dairy Sci. Assoc., St. Paul, Minn. Lush, J. L., and L. D. McGilliard. 1955* Proving dairy sires and dams. J. Dairy Sci. 33:163. Lush, J. L., and F. S. Straus. 1942. The heritability of butterfat production in dairy cattle. J. Dairy Sci. 25:975. Mason, I. L-, and A. Robertson. 1956. The progeny testing of dairy bulls at different levels of production. J. Agr. Sci. 47:367McGilliard, L. D. 1952. Usefulness of the herd average In estimating breeding values of dairy cattle. Unpublished Ph. D. thesis, Iowa State College Library, Ames. National Cooperative Dairy Herd Improvement Program Newsletter. 1955. Agricultural Research Service, U.S.D. A., January issue. 1957. Agricultural Research Service, U.S.D.A., March issue. New Zealand Dairy Board . 1956. 32nd Ann. Report. Oregon Dairy Breeders Association Newsletter. 1957Proven vs. select young sires, April 5th Issue. Pearson, K. ^ J . _ 3 +■ tt 1931. Tables for Statisticians and Blometrlcians, Part 1 1 . Cambridge University Press. 75 Plum. M. 1935* Causes of differences in butterfat production of cows in Iowa testing associations. J. Dairv Sci 18 :8 1 1 . J Plum, M., and M, G. A. Rumery. 1956. Forty years of dairy cattle breeding at the North Platte experiment station. Nebraska Research Bull. 182 . Rendel, J. M., and A. Robertson. 1950. Estimation of genetic gain in milk yield by selection in a closed herd of dairy cattle. J. Genetics, 50:1 Rendel, J. M., A. Robertson, and K. A. Alim. 1951. The extent of selection for milk yield indairy cattle. Emp. J. of E x p ’t. Agr. 19:295. Rice, V .A . 1944. A new method for indexing dairy bulls. Sci. 27:921. J. D-iry Robertson, A. 1954• Artificial insemination and livestock improvement. Advances in Genetics. 6:4-51 • Robertson, A., and J. M. Rendel. 1950. The use of progeny testing with artificial in dairy cattle. J. Genetics. 50:21. Insemination . and 1954. The performance of heifers got by artificial insem­ ination. J. Agr. Sci. 4-4:184-. Schaefer, W m . 1953. Comparison of original proof, AB daughters and apparent transmitting ability of 200 bulls used in New York and Penn, breeding cooperatives. Proc. of 6th Ann. Conv. of National Assoc, of Artificial Breeders, 132. Seath, D. M. 1940. The intensity and kind of selection actually practiced in dairy herds. J. Dai^y Sci. 23:931* Sendelbach, A. G., E. L. Corley, and E. E. Heizer. 1957. The repeatability of natural service (NS) and arti­ ficially sired (AB) daughters on subsequent AB performance as measured by regressions of daughter averages. J. Dairy Sci. 40:620. Wadell, L. H. 1957. Influence of artificial breeding on production in selected Michigan dairy herds. Unpublished M.S. thesis, Michigan State University Library, East Lansing.