The coastdown test is done to calculate the effect of mechanical resistance and aerodynamic drag on the car, which has a direct effect on fuel economy. The effect is most pronounced at highway speeds, which is why many of the EPA number adjustments concern only that figure. Simulating the coastdown test requires a road-force load number to be calculated, and that calculation is what the EPA is adjusting now.
The numbers are not changing much for 2017, but this is still something to consider when car shopping, especially when cross-comparing fuel economy ratings from 2016 and prior with 2017 numbers. Here are some adjusted figures for consideration: A 2017 Acura ILX will get 35 mpg highway instead of 36, a 2017 Kia Rio Eco will do 36 highway instead of 37, and a Mazda CX-3's highway reading drops from 35 to 34 mpg.
The coastdown test is something with which several manufacturers have struggled in the past. In 2014, Hyundai and Kia were found to have overestimated the fuel economy of about 1.2 million vehicles made from 2011 to 2013, which resulted in collective penalties of around $300 million. The Korean carmakers used false road load values in their dynamometers when fuel economy ratings were calculated in laboratory conditions. A similar calculation issue affected Japanese-market city cars made by Mitsubishi, which were amidst the recent Mitsubishi/Nissan fuel economy scandal. Ford hasn't had it easy either, as erroneous total road load horsepower test values and wind tunnel testing caused several of its hybrids' mpg numbers to be overstated.