UNC stats class finds 1,500 unaccounted absentee ballots in NC 9th District election

A UNC statistics class analyzed data from the most recent 2018 North Carolina general election and found that more than 1,500 absentee ballots were unaccounted for in Bladen and Robeson counties, which are part of North Carolina’s 9th Congressional District. 

Richard Smith, distinguished professor in the Department of Statistics and Operations Research, teaches the approximately 70-student class of STOR 556, “Advanced Methods of Data Analysis” and had students analyze election data from all 100 North Carolina counties. 

“It started out as a homework exercise,” Smith said. “There was an example in the text which was about the election results from the Bush v. Gore election, which is 19 years ago, and I said to the students, ‘Why are we talking about this example that is 19 years old when we have lots of data right here? There is an argument going on in North Carolina.' So I went online, downloaded some data and made it a homework exercise. I wanted to see what students would make of it.” 

The basis of Smith’s assignment revolved around the undecided election of Republican candidate Mark Harris for the U.S. House in North Carolina’s 9th Congressional District.

Harris won the election in November 2018 by 905 votes over Democrat Dan McCready, but the North Carolina State Board of Elections refused to certify his election after doubts arose concerning the legitimacy of the results due to fraudulent practices with absentee ballots. The State Board of Elections held a hearing this week about these allegations of fraud. 

The Board could order a new election if the 9th Congressional District ballots aren't certified.

“One of the issues is, can we actually prove that the number of missing votes was greater than 905?” Smith said. 

Smith said he collected data of the number of absentee ballots for all 100 counties in North Carolina. He then found the proportion of absentee ballots that were not returned was less than 4 percent for 98 counties, slightly higher than the usual 1 to 2 percent rate, and around 11 percent in both Bladen and Robeson counties. 

Students were first instructed to create a linear model for all 100 counties. 

Rachel Allen, a senior in Smith's class, said it was obvious on the graphs and modeling she created that of all the counties, Bladen and Robeson counties were outliers. 

Smith then had students create a linear model, without Bladen and Robeson counties, to predict the number of absentee ballots that should have not been returned for the two counties given various demographic variables and data from the other 98 counties under normal circumstances. The class then compared the estimates to the number of ballots they had found that were actually unaccounted for in Bladen and Robeson counties last November.  

“If you look at what we found, it is basically impossible for these two counties to have the number of unaccounted for absentee ballots that they do (without intervention)," said Michael Quint, a junior in Smith's class.

The class findings suggest that more than 1,500 ballots were not accounted for this past election. They report that data with a 99 percent confidence interval, which is a 99 percent confidence level, Smith said.

Smith said the point of doing the calculation is to show that 1,500 is greater than 905 so it proves this fraud could have changed the results of the election. 

“We don’t have any means of knowing whose votes were intended for the Republican or Democrat, but what the suggestion in the hearings is that there was someone going around collecting ballots and systematically removed all the ballots for the Democrat candidate,” Smith said.

Smith had sent an earlier version of his findings to the North Carolina State Board of Elections this past January. After reviewing the findings of his students in his class this semester and adding in some of their findings, he revised his paper and submitted it to the Board again. He has not received a response.

The Daily Tar Heel contacted the North Carolina State Board of Elections for comment. At time of publication, the DTH had not received a response.

“Professor Smith wasn't doing anything explicitly political with his paper besides just stating that there was something odd going on,” Quint said. “But I think through that work he was able to cement that and help with furthering the investigation to prove what was happening.” 

university@dailytarheel.com