Clarity's Likely Voter models were used for a compelling study on vote by mail:

The challenge in trying to evaluate the impact of vote at home on turnout is that there’s never a control group. Turnout rates in the states where everyone can vote at home — Oregon, Washington and Colorado — have increased since the system was adopted, and they’re now among the highest in the country. But because elections are complicated, and each state’s demographics are unique, it’s hard to prove that vote at home is the cause. 

 So Pantheon Analytics did the next best thing. The firm looked at voter files — the records of individual voters kept by state elections officials — in Colorado in 2014 and compared them with the predictions of a respected voter turnout model created by the Democratic-leaning consulting firm Clarity Campaign Labs. Turnout models use publicly available voter file information — age, sex, voting history — combined with consumer data to generate a probability score for each registered voter. 

Read more at https://www.washingtonpost.com/outlook/letting-people-vote-at-home-increases-voter-turnout-heres-proof/2018/01/26/d637b9d2-017a-11e8-bb03-722769454f82_story.html