Interesting paper, but the data set is poor, limited, and during the height of the crisis in 2008. I bet calm alert and happy were highly correlated to daily m

Interesting paper, but the data set is poor, limited, and during the height of the crisis in 2008. I bet calm alert and happy were highly correlated to daily movements then. The model would have a much higher error rate if sampled from 2009 and 2010.