Solve trajectory projections using base year data #172
Description
The current trajectory projection code looks at corp disclosures of typically recent era and projects the trends of up to five years of data. For some companies, we only have 3 or 4 good years of data. For many companies, the past 5 years produce trajectory projections that head up, not down, creating nonsense scores (as there's no way a company is going to be increasing projections at that rate for the next 30 years and the cumulative overshoot is just massive).
Many of those same companies with marginal gains over the past 5 years have made great progress since their "baseline" years (which go back much further: 2012, 2005, etc). If we do a linear interpolation of the ancient base year to the first year of "real" report data, the historic intensity projections might make a lot more sense.
We have the base year data due to targets, and it should be straightforward to project into 2016, 2017, 2018, or wherever our data picks up. It is a way to "fill the data gaps" given how difficult it is to find good disclosure data from 2014, 2015, etc.
Comments @LeylaJavadova @ImkeHorten @hmorgan29 @kmarinushkin