Langer Research Associates formally launched our PARC knowledge management tool for survey research professionals at the annual conference of the American Association for Public Opinion Research in New Orleans this past week, as well as presented three papers, all focused on aspects of polling in the 2016 presidential election.

PARC is a secure, cloud-based system that enables researchers to instantly and accurately search their past projects, including individual survey questions. It’s invaluable for internal organization and efficiency, client support and retention and stakeholder or public-facing uses. We received great interest and feedback at our exhibit hall booth. (Our ad in the conference program looked pretty swell, too!)

Among our papers was “Predicting 2016 State Presidential Election Results with a National Tracking Poll and MRP,” in which we described our use of the national ABC News and ABC News/Washington Post tracking poll in 2016 to produce state-level vote estimates via multilevel regression with poststratification. Our model correctly predicts the winner in 50 of 51 states and the District of Columbia, outperforming publicly released pre-election prognostications. Our findings indicate the effectiveness of using a large, high-quality, probability-based national dataset to make state-level election predictions. See the full paper here and our presentation slides here.

We also offered our summary of “How and Why it’s President Trump,” slides here; and our postmortem analysis of the tracking poll we produced for ABC News, in which our final estimate came within 2 points of the actual popular vote margin, matching the average accuracy of ABC News pre-election polls since 1984 (slides here).