I finally had a chance to read through the study. A number of points:
1. The paper makes numerous assumptions (all of which are spelled out) and concedes that some of its methodologies and assumptions "may be subject to large uncertainties and controversies." One example might be its figures based on traffic accidents. While the fuel might change, if a cleaner fuel allows for at least the same intensity of driving as is presently the case, there would be little accident-related savings.
2. The paper's approach for estimating the subsidies based on estimated externalities is a fairly common methodology.
3. The paper attributes only 22.3% of the total subsidies (pre- and post-tax) to climate change-related effects and 46.0% to local air pollution (particulates, etc.).
4. The paper also employs sensitivity analysis to "test" sets of scenarios. Based on reducing/increasing estimated air pollution damages by 50%, the post-tax subsidy would range from 4.7% to 8.3% of global GDP vs. the baseline estimate of 6.5% for 2015.
Having said that, IMO, the analysis isn't sufficiently complete for a comprehensive policy shift. It arrives at a figure for the extent of subsidies based on externalities. What it fails to do is to robustly examine the full impact on global GDP for the status quo (baseline) vs. alternative scenarios (though some other work is emerging in this area). For example, coal (no surprise given its air pollution intensity) accounts for the largest share of subsidies. If all coal usage were eliminated (a reasonable medium or long-term goal, but not practical in the short-term), there would be insufficient energy substitutes at present, leading to an energy price impact and global economic impact. Having an estimate of that global economic impact would be valuable, as it could better inform policy makers as to the opportunity costs associated with various policy choices. Estimates of subsidies provides part of the picture, but the full picture requires a robust assessment of the overall economic impact, given the macroeconomic linkages to incomes, poverty, etc. Clearly, a transition period will be required and until one has strong opportunity cost data, the details will remain murky on account of competing interests being able to leverage the opportunity cost uncertainty.
As noted in numerous past threads, I accept the scientific consensus on climate change (including the anthropogenic driver) and favor a policy path that reduces greenhouse gas emissions. Nevertheless, I recognize that one can't get from the present energy mix to a largely carbon-free one for some time (lack of alternatives, need for changing the energy infrastructure, etc.). A realistic path will need to consider the overall macroeconomic and socioeconomic impacts of various alternative policy paths. This study provides useful information i.e., some share of the subsidies could be utilized for financing cleaner technology/R&D, reducing coal usage would be a logical starting point for a cleaner energy path, etc., but does not provide the amount of information needed to materially move policy from the business-as-usual path.