Reviving this thread as it's been mentioned in another

Surely this study is only as good as the dataset used. Reading the supplementary information, the "million song dataset" is slightly misleading in this study as more than half of it was discarded due to lack of date information or duplication.

Also, Jack, Country is included...