A multivariate and stochastic approach to identify key variables to rank dairy farms on profitability

The economic efficiency of dairy farms is the main goal of farmers. The objective of this work was to use routinely available information at the dairy farm level to develop an index of profitability to rank dairy farms and to assist the decision-making process of farmers to increase the economic efficiency of the entire system. A stochastic modeling approach as used to study the relationships between inputs and profitability (i.e., income over feed cost; IOFC) of dairy cattle farms. The IOFC was calculated as: milk revenue + value of male calves + culling revenue - herd feed costs. Two databases were created. The first one was a development database, which was created from technical and economic variables collected in 135 dairy farms. The second one was a synthetic database (sDB) created from 5,000 synthetic dairy farms using the Monte Carlo technique and based on the characteristics of the development database data. The sDB was used to develop a ranking index as follows: (1) principal component analysis (PCA), excluding IOFC, was used to identify principal components (sPC); and (2) coefficient estimates of a multiple regression of the IOFC on the sPC were obtained. Then, the eigenvectors of the sPC were used to compute the principal component values for the original 135 dairy farms that were used with the multiple regression coefficient estimates to predict IOFC (dRI; ranking index from development database). The dRI was used to rank the original 135 dairy farms. The PCA explained 77.6% of the sDB variability and 4 sPC were selected. The sPC were associated with herd profile, milk quality and payment, poor management, and reproduction based on the significant variables of the sPC. The mean IOFC in the sDB was 0.1377 +/- 0.0162 euros per liter of milk ((sic)/L). The dRI explained 81% of the variability of the IOFC calculated for the 135 original farms. When the number of farms below and above 1 standard deviation (SD) of the dRI were calculated, we found that 21 farms had dRI < -1 SD, 32 farms were between -1 SD and 0, 67 farms were between 0 and +1 SD, and 15 farms had dRI > +1 SD. The top 10% of the farms had a dRI greater than 0.170 (sic)/L, whereas the bottom 10% farms had a dRI lower than 0.116 (sic)/L. This stochastic approach allowed us to understand the relationships among the inputs of the studied dairy farms and to develop a ranking index for comparison purposes. The developed methodology may be improved by using more inputs at the dairy farm level and considering the actual cost to measure profitability.