Modifications to Calving Ease Evaluations – Fall, 2002
by Peter Sullivan
Calving ease genetic evaluations are derived with a complex genetic model that combines
information on calf birth weight and size of the dam, with observed frequencies of
calving assistance. Calving ease records of unassisted (U), easy pull (E), hard pull (H) or
surgical delivery (S) are converted to calving scores, based on probabilities of easy
calving in each specific situation. The scores account for the effects of sex of calf, age of
cow and relative size difference between the parents on the probability of easy calving.
Genetic evaluations for “calving score” (CS) are therefore adjusted for the effects of age
of dam, birth weight, etc. Genetic evaluations for “calving ease” (CE) are predictions
that consider all of the important factors affecting ease of calving, not just CS. Concerns
have been raised about the stability of CE evaluations over time and about the
reasonability of the evaluations for certain bulls. Research was therefore conducted to
determine if the current model for calving ease could be improved.
Materials and Methods
Calving ease records from the fall 2001 evaluation were divided to create one data set for
genetic evaluation of calving score and growth traits, and a second data set for
development of calving ease prediction equations. The first data set (D1) included
records made between 1986 and 1991. The second data set (D2) included records made
between 1992 and 2001. See Table 1 for more details about the two data sets. The fall
2001 genetic evaluation system was used with D1 to compute genetic evaluations for
calving score and for the growth traits. Some of the parents evaluated had progeny in D1
and D2, while others had progeny in D2 only. The genetic evaluations for parents that
had progeny in D2 only were pedigree evaluations computed automatically by the
system, and included predicted breed differences for individuals from different breeds.
Table 1. Numbers of observations used in the analyses (after all data quality edits).
Data Set Calving Ease Birth Wt Weaning Wt Yearling Wt Gain on Test
D1 531832 307002 698643 166175 22082
D2 parity 1 58100
D2 parity 2 59044
D2 parity 3 287394
Observed frequencies of calving assistance in data set D2 were regressed against the
genetic evaluations from D1 to derive the best prediction equations for future frequencies
of unassisted calving. Direct and maternal genetic evaluations for calving score and for
all of the growth traits were used with backwards elimination (P<.01) to derive prediction
equations by age of dam category (1st, 2nd, 3rd or higher parity).
Results and Discussion
The most important factors across all age of dam (parity) categories included direct and
maternal calving score, direct birth weight and maternal weaning gain. Calving difficulty
is more common as calf birth weight increases, and as cow size decreases. Both the
direct and maternal components of growth traits have moderate to high genetic
correlations with mature size. The maternal components have relatively lower genetic
correlations with birth weight direct, however, and were therefore better additive factors
in combination with birth weight direct. Maternal weaning gain was the most important
maternal size effect across the three parities.
The final prediction equations are in Table 2 and have been implemented for the fall 2002
evaluation. The mean rate of unassisted calving increased with increasing parity. Heifer
calves were unassisted more frequently than bull calves in all parities. Regression
estimates decreased in magnitude with increasing parity, for all ABC traits as expected.
Table 2. Regression estimates to predict %unassisted future calvings.
Effect 1 2 3
Mean 75.01 89.98 93.55
Sex of calf (bull – heifer) 12.22 5.27 3.04
Calving Score direct 1.368 1.177 0.617
Calving Score maternal 3.009 1.097 0.546
Birth weight direct -0.926 -0.459 -0.370
Weaning gain maternal 0.242 0.119 0.0418
The relative importance of each ABC trait is difficult to determine from the regression
estimates, because the variation of ABCs is different for each trait. The regression
estimates for each trait were therefore multiplied by the standard deviation of the
corresponding ABC to determine the relative importance of each trait for predicting the
percentage of future progeny to be born without assistance.
Relative importance of all four ABC traits is in Figure 1. In first parity, the birth weight
direct ABC was the most important factor affecting the % unassisted calving, followed by
the maternal calving score evaluation of the dam. In second and third parity, the direct
evaluations for both calving score and birth weight were the most important factors.
Figure 1. Relative importance to predict % U
1 2 3
Calving score direct and maternal were combined, and the size traits of birth weight
direct and weaning gain maternal were combined, to determine the importance of calving
score relative to size evaluations (Figure 2). In all three parities, calving score was
similar, but slightly less important than size.
Figure 2. Relative importance to predict % U
1 2 3
Direct components of calving score and size were combined, and maternal components of
calving score and size were combined, to determine the relative importance of total direct
and total maternal effects (Figure 3). Direct and maternal effects were equally important
in parity 1, but the relative importance of maternal effects decreased in higher parities.
Figure 3. Relative importance to predict % U
1 2 3
Previous predictions of % unassisted put more emphasis on calving score than on size
evaluations. The prediction equations were also more complex, involving first a non-
linear prediction of average calving score and then a non-linear conversion of average
calving score to % unassisted. The current approach was to predict % unassisted directly
and all effects were linear. The relationships between the new and previous evaluations
for % unassisted relative to birth weight direct evaluations are shown in Figure 4.
Figure 4. Distribution of %U in parity 1 with new and old
-10 -5 0 5 10
The new evaluations (ABCs) were much less variable both across and within birth weight
categories. The new range across all animals was approximately –20 to 20, compared
with the previous range of –45 to 25. The maximum range within a birth weight category
was reduced from about 70 to 25. The average prediction for %U also changed across
birth weight categories, most notably for very low and very high birth weight bulls. For
example, the average %U evaluation for bulls that were -10 for birth weight ABC
changed from 4.4 to 10.2. The range for these bulls changed from [-18.9, 23.6] to [3.5,
16.5] (Table 3). With the new evaluations, bulls that were negative for birth weight ABC
were more consistently positive for %U. There were still bulls that were negative for
both birth weight and calving ease, but fewer than before and almost none that were
extremely negative for both.
Table 3. Percentile ranks for calving ease ABC (%unassisted in parity 1) within birth
weight (BW) ABC (lbs) categories.
Calving Ease Percentile Ranks with Previous System
BW ABC 0 (min) 10 50 (median) 90 100 (max)
-10 -18.9 -0.6 4.3 10.5 23.6
-5 -39.5 -0.1 4.8 9.7 26.3
0 -43.1 -4.2 1.3 6.9 26.5
5 -47.5 -15.6 -6.0 0.7 21.8
10 -38.8 -29.6 -18.4 -11.1 4.5
Calving Ease Percentile Ranks with New System
BW ABC 0 (min) 10 50 (median) 90 100 (max)
-10 3.5 8.8 10.2 11.8 16.5
-5 -9.3 3.7 5.1 6.3 13.6
0 -15.9 -1.4 0.1 1.5 9.3
5 -21.5 -7.8 -5.0 -3.1 3.2
10 -18.6 -14.8 -10.8 -8.5 -4.2
Correlations between birth weight ABCs and the old and new ABCs for %U from heifers
(parity 1) are in Table 4. The new %U evaluations were much more strongly related to
birth weight ABC, with correlations increasing in magnitude from -.487 to -.909. Re-
rankings of bulls within a birth weight category were minimal, however, with correlations
ranging from .969 to .997. Substantial re-ranking will be seen, as reflected by the overall
correlation of .800 between old and new %U, but this will almost exclusively be due to
the increase in agreement between rankings for %U and birth weight.
Table 4. ABC correlations. ABCs for %U are for heifers (parity 1).
Animals included ABC trait %U-new Birth weight
All animals %U-new -.909
All animals %U-old .800 -.487
Birth weight = -10 %U-old .969
Birth weight = -5 %U-old .974
Birth weight = 0 %U-old .990
Birth weight = 5 %U-old .995
Birth weight = 10 %U-old .997
Results so far are limited to the current scoring and evaluation procedure for calving ease
observations, which was not changed. Further work is needed to consider alternative
scoring and/or evaluation systems, some of which have already been identified. A major
step will be estimating the required variance and covariance components for each of the
alternatives. Preliminary analysis has shown that both the scoring system and the
(co)variances assumed for the evaluation can have important effects on the ABCs for
calving ease. Accuracy of the ABCs for predicting future % unassisted will be an
important criterion for comparing the alternative systems. Stability of the evaluations
over successive evaluation runs will also be of interest.