of the report, it is arrived at by fitting a two models with 4/6 fit parameters, extrapolating them, and taking the difference between that and the observed numbers. They fit their model through 2014, then show that it diverges from 2015 on. How do we know their fit is ever any good if extrapolated past the end points? For example, it would be nice to see how it performs at predicting 2012-2014 after being fit up to 2011.