Updates on Election Analysis Project (2018)

December 5, 2018

Author: Yair Ghitza, Chief Scientist

First, we just updated the national congressional data that we posted a few weeks ago. This mainly1For transparency (and careful readers/analysts who download the data), we also cleaned up some backend data/modeling issues that came up as we continued to analyze the data. For example, birthdate is missing from only 1% of records on the voter file nationally, but more in some states. We have to impute an estimated age for these missing records — this doesn’t affect things much at the national level, but it can be important in places with lots of missing data. We noticed a problem in how some of the imputed age records were computed and fixed it for this round of analysis. reflects the additional results that have come in since we posted. The Democratic margin of victory increased as more results came in, and we see that reflected in many of the voting blocs shown in the spreadsheet. Qualitatively, little has changed: our estimates only changed by a point or two for each reported group.

Second, this week we will begin publishing analysis of states and congressional districts, starting with the Georgia governor’s race. Here is our tentative publication schedule. More states and CDs will be added as more precinct data comes in, along with official turnout data from the Secretaries of State around the country.


This shows our plan for specific elections, but we’ll also have posts about more general trends that we’re seeing across the country. Some will describe technical pieces of our data; some will describe political/substantive trends; and some both.

Uncontested elections skew the national topline, by about 1 point in margin in 2018.