I've gone through the paper the first time, and here are some preliminary thoughts.
1) The paper says it uses gasoline vehicle emissions comes from the EPA Tier 2 emissions standards and the Argonne National Lab's GREET model. Under "Data sources for emissions of gasoline cars" the paper makes it clear that the authors use the EPA Tier 2 emissions standards to only model tailpipe emissions of SO2, CO2, NOx, VOCs, and PM. The authors bother to use Argonne National Labs GREET model to augment VOC and PM2.5 from tires and brake wear, but very clearly ignores all upstream emissions from gasoline production. That's completely disingenuous considering the amount of effort going into modeling the electricity production side.
2) on the electricity production side, under "Emissions for Electric Vehicles" it is unclear to me how the authors modeled the increase in electricity usage. I understand that the authors grabbed hourly emissions profiles for various power plants and the hourly electricity consumption. However, for the increase in load, it is not clear how the authors modeled the variance between power sources as load in increased. From the equation, I think the authors assumed an even increase across all power plants which clearly is not the case. For example, there is a big difference if the increase in load comes from a natural gas peaker plant or a hydro plant versus a coal plant. Further, it appears that the model doesn't account for any situations where the power plant may be over-generating electricity already, which may be the case in super-off-peak times. In other words, a coal plant at 2am might be at the minimum 40% idle level and is already generating above the demand load at that time - which means any additional charging, up to the point where the coal plant needs to increase power levels does not generate any additional emissions. The paper also uses some assumptions on charging profiles with time of day, but I don't see where it has the range presented.
3) Treating all emissions linearly for "damage" seems to me to be a very big assumption. I haven't yet sorted through how the authors map emissions to $, but that seems to be an area where the assumptions can cause significant distortion to the results.
4) Also, the point of emissions and the coverage of the resulting cloud is interesting and deserves more attention. I have seen some studies, but I think they are not quite giving us an accurate picture of the benefits or detriments of shifting the point of origin of the emissions.
5) The overall concept of tailoring incentives to regions of the country is interesting, but I think is flawed from the outset. First, the power plant data is old and always in flux. The contribution to emissions from an EV therefore is always in flux with a potential for much more change than with a gasoline vehicle. Similarly with #4, moving the point of emissions, even if the overall emissions is higher for now might be worth it. Also, there may be small steps at the power plant side that (scrubbers for instance) that might change the emissions profile dramatically in an economical fashion. As a result, while a BEV sold today may have slightly worse emissions than the equivalent gasoline car, that may very well not be the case in just 1 or 2 years with increase in wind energy and the reduction of electricity use overall, which may help the existing nuclear and hydro sources cover more of our super-off-peak load.
I think the lack of upstream calculations ultimately dooms this paper to irrelevancy. It is interesting that this paper attempts to discuss hourly power production of electricity, but falls far short of providing the critical information on the capacities of super-off-peak power production with hydro, wind, and nuclear electricity production. I suspect that given what we know of charging, 95% of charging is done at off-peak, and for some, 99+% is done at super off peak. So the charging profiles are almost always wrong for Model S owners which represent the future BEVs (2017 onwards, the behavior and products will reflect a longer range, more like current Model S owners).