The graph doesn't have an 2015 data on it, but the average in 2015 has dropped further.
I initially set out to model soft costs on their own, separate from module costs. However, that effort revealed other factors. Solar capacity factor is rising and operational costs (which are not included in soft costs are dropping. There are more variables at play. For some of those variables, there isn't robust data available over lon times.
All of those variables ultimately feed into electricity prices, in the end, which is why I chose that modeling approach.
So since the cost drivers changing - why does it make sense to model future price on old cost drivers ?