>>/301093/
probly flawed. i read some analysis piece that looked at new voter registration over the past 4 years that claimed the models didn't take into account the shift seen. not to mention trump voters distrust/incentive to lie, oversampling etc etc