The effect of cow longevity on dynamic productivity growth of dairy farming


 Cow longevity is recognized as an important trait to improve farm economic performance while concurrently reducing environmental and societal impacts. However, there is an economic trade-off between longevity and herd genetic improvement, which may influence the evolution of dairy farms’ efficiency and productivity over time. This study used a panel data of 723 Dutch specialized dairy farms over 2007-2013 to empirically measure the effect of longevity on dynamic productivity change and its components. First, the productivity growth estimates were obtained using the Luenberger dynamic productivity indicator. Then, the estimates were regressed on longevity and other explanatory variables using dynamic panel data model. Results show that the average dynamic productivity growth was 1.1% per year, comprising of technical change (0.5%), scale inefficiency change (0.4%) and technical inefficiency change (0.2%). Longevity is found to have a statistically significant positive association with productivity growth and technical change, implying that farms with more matured cows were also those farms that recorded increased productivity through technical progress. However, it has a negative association with technical inefficiency change, which might follow from the reduced milk productivity of old cows. Dutch dairy farms have a potential to raise productivity growth by reducing technical inefficiency.


INTRODUCTION
The increased focus on milk productivity of modern dairy cows has been associated with a decline  where, is the peer weights or intensity vector for defining the reference frontier and is the 150 depreciation rate associated with the quasi-fixed inputs. Linear programming is used to solve Eq. 3-6.

151
In the empirical application of the current study, the quasi-fixed input constraint in Eq. 3-6 is rewritten

178
A second stage regression model is used to explain variations in productivity growth and 179 inefficiency scores over time within ones farm and across farms, and specifically, to measure the effect 180 of longevity on dynamic productivity growth and its components. The model can be written as: where is dynamic productivity growth and its components for farm ( = 1, 2, … , ) in year ( = where age at first calving (years) and culling rate (decimal).

190
Other explanatory variables ( , ) in Eq. 11 refer to factors that were not considered during the 191 estimation of the inefficiency scores, yet expected to influence the economic performance of farms   3 The models with one lag provide the best specification in terms of serial correlation and joint validity of instrument post estimation results (see the Results section). 4 According to Banker and Chang (2006, p. 1317), '… the use of a more stringent screen level such as 1 is likely to misclassify many uncontaminated efficient observations as outliers, while the use of a less stringent screen level such as 1.6 or greater may fail to remove many contaminated observations'. 5 It is assumed that the livestock value represent the value of the breeding stock since the sample farms are specialised dairy farms, where at least 85% of the total farm revenue is obtained from milk production.      The main driver of the fluctuation in productivity change was fluctuation in technical change 337 during the sample period (Figure 1). This might be due to volatility of milk and input prices. For The negative correlation between milk price fluctuation and dynamic productivity growth 8 implies that 341 'farmers are conservative (pessimistic) regarding price expectations and they devise production 342 structures that are optimal in low price frameworks' (Oude Lansink et al., 2015). As a result, farmers' 343 behavior is more conducive for achieving productivity growth during low milk price years than high 344 price years.

Effect of Cow Longevity on Dynamic Productivity Change and Its Components
362 Table 4 presents the estimation results of the two-step GMM for measuring the effect of cow 363 longevity on dynamic productivity change and its components. The first lag of dynamic productivity growth is statistically significant in explaining the variations 375 in dynamic productivity growth, implying the persistent nature of productivity. That is, farms with a 376 productivity decline last year would observe more productivity decline in the current production period.

377
Similarly, farms that were inefficient last year would become more inefficient in the current production directly comparable with the current study due to differences in the modelling approach (dynamic vs 400 static) and sample periods used. In the current study, the association between longevity and scale 401 inefficiency change was not statistically significant.

402
An increase in CI has a statistically significant negative, at the critical 10% level, association with 403 scale inefficiency change. An increase in CI by 1-d is associated with a 0.008 decrease in dynamic scale 404 inefficiency change (i.e., CI is positively associated with scale inefficiency). Although statistically 405 9 The same model structure used for the dynamic productivity change is fitted for its components (i.e., technical change, and technical and scale inefficiency changes). As a result, the post-estimation results show the presence of second-order autocorrelation for technical change and technical inefficiency change models (Table 4).