In 2006 and 2011 two vector borne diseases have emerged in the European ruminant population: bluetongue (BTV) and Schmallenberg. Milk yields collected as part of milk recording represent an abundant source of information that could be used for emerging disease surveillance. The aim of this work was to evaluate monthly collected milk yields as an indicator of vector borne disease emergence. Milk recording data were used to detect the 2007 BTV emergence in France. Expected test-day milk productions per cow were predicted for 2006 and 2007 from linear mixed models, based on 3 years of herd milk records history. Clusters of deviation between predicted and observed milk productions were detected using a scan statistic as implemented in SaTScan™. Log likelihood ratios (LLR) were used to rank the clusters and to set a threshold for the definition of an alarm. The choice of a threshold was a trade-off between the number of alarms before the emergence (false alarms) and the timeliness of the detection once the disease had emerged. Using an LLR of 50 (100), there were an average of 1.7 (0.8) false alarms per week and the BTV emergence was detected 7 (9) weeks after the first notification. The first cluster with an LLR>100 located in the emergence area was further investigated. A difference between observed and predicted production of greater than 1 kg/cow/day was observed around the time of emergence. However, a difference of equal magnitude was observed during the year preceding the outbreak. We conclude that milk production predicted from herd history alone does not allow the detection of an emerging infectious disease that would have had an effect on milk production similar to BTV.