Monday, December 3, 2018

NC's Competitive, and Now Contested, 9th Congressional District

With the national attention that the Old North State's Ninth Congressional District is getting this past  week and will likely get in the next few days, I thought it would be good to give some comparison perspective and an overview of the main issue at hand: the absentee by mail ballots in the 9th and what we know about them, from a data point of view.

An Overview of NC Voting Methods and Patterns:

For those unfamiliar with North Carolina voting methods, there are three methods that are most used by North Carolina registered voters to cast a ballot with: in-person on Election Day, in-person through early voting (known as absentee one-stop), and through absentee by mail (ABM).

With a record turnout for a blue-moon election cycle in the state, 2018's mid-term election saw the first time that more ballots were cast before Election Day than in recent elections. Typically, this is true in NC's presidential elections, but mid-terms tend to be ones that see the majority of ballots come on Election Day, rather than prior. This year was substantially different, as was the case that several congressional contests were notably competitive (the 9th, 13th, and 2nd), despite the normal partisan advantage given to these districts.

While the significant majority of absentee ballots came through one-stop/in-person, absentee by mail ballots surpassed 2014's mid-term election final numbers:

While the numbers were almost double for the one-stop/in-person method, more North Carolinians also requested and returned absentee by mail ballots than four years ago. And there was a distinct partisan difference in the voting methods than in year's past as well.

The following charts show the three primary methods--absentee by mail, absentee one-stop/in-person, and Election Day--and the party registration of voters using those methods to cast their ballots. Distinct trends are evident in both presidential and mid-term elections in North Carolina:

First, the 2008 and 2010 elections:

In 2008, with the Obama ground game, 61 percent of all the ballots cast came before Election Day in North Carolina, while two years later, in the mid-term, 63 percent of 2010's ballots came on Election Day.

Next, the 2012 and 2014 elections:

Again, 2012's presidential contest saw 61 percent of the ballots cast before Election Day, while the 2014 mid-term saw 58 percent of the ballots come on Election Day.

Finally, the 2016 and 2018 elections (note that for 2018, a few of the 100 counties have not finalized their Election Day data, so that field is left blank):

A couple of observations to make based on the proceeding three charts:

  • absentee by mail ballots usually constitute anywhere from 3 to 5 percent of the ballots cast. In 2018, it was on the high end, with 5 percent of the total absentee ballots being cast (once the final Election Day data is incorporated, it will like slip down);
  • historically, registered Republicans dominated in absentee by mail, as noted by their percentages of registered voters in that vote method;
  • while conversely, registered Democrats tend to dominate in absentee one-stop/in-person; and,
  • Election Days tend to bounce back and forth or are much closer when it comes to party registrations.

But, in 2018, Democrats, obviously through a concerted strategy and effort, got their registered voters to use absentee by mail this year, with 41 percent of the state-wide ballots being returned and accepted, while registered Republicans saw their historic advantage slip to third behind registered unaffiliated voters requesting and returning their absentee by mail ballots for accepted votes.

Now, another important explanation about North Carolina's absentee by mail vote method: 
  • voters requesting an absentee by mail ballot submit a form to their county board of elections, who records the request with the date that it came in;
  • next, the county board sends the voter an absentee ballot, again recording when that ballot was sent;
  • the voter should receive the ballot, complete it and the requirements (such as signing the ballot and witnessing it), and then return the ballot to the county board;
  • when the county board receives the ballot, it notes the date and, following a review of the ballot and the various requirements the voter was to complete, will accept the ballot as a vote and process it on Election Day. 

Through the courtesy of the North Carolina State Board of Elections and Ethics Enforcement, all of this data and records for each voter is recorded in a file that is publicly available each morning during the voting period leading up to Election Day and beyond. This file ("") is what I use to conduct the following analysis, open to anyone with sufficient computing power.

The State-Wide Picture of Absentee by Mail Voting

With that, I'll give an overview of the absentee by mail data, starting with the "big picture" of the entire state, drilling down into the 9th District, and then finally into various counties within the 9th, especially focusing on Bladen and Robeson counties. I'll do this by analyzing the data for three primary factors: absentee by mail ballots that were returned and accepted, those that were returned but rejected, and those that were never returned. This analysis will be done through the lens of voter race and voter party registration. 

First, at the state level: again, North Carolina did not have a state-wide major race (like U.S. Senate or gubernatorial), so much of the action occurred at the congressional level. First, the thirteen congressional districts and the state-wide total for the accepted rates of requested ABM ballots:

And then the non-return rates by congressional district and state-wide:

The 9th Congressional District has the distinction of being the district with the lowest accepted ABM ballot rate (73 percent) while having the highest non-return ABM rate of the 13 districts (24 percent). 

In breaking down the state-wide rate numbers by voter party registration, slight differences are evident from the data for the three major registration groups (registered Democrats, unaffiliated, and Republicans):

For the specific numbers by party, the following gives the cross-tab analysis conducted in SPSS of all requested ABMs from across the state with the returned ballot status (accepted, rejected and the various reasons, and those not returned):

So, out of the little over 122,000 ABMs requested across the state, a little under 20,000 were never returned to the county boards of election. Here's the breakdown by Congressional district for these non-returned ballots:

With nearly 3,500 requested but non-returned ABM ballots, the 9th Congressional District led the state. 

The 9th Congressional District:

In turning to the district at the heart of the controversy, the 9th Congressional District stretches from southeast Mecklenburg County (and includes parts of Charlotte) down along the NC/SC border east to Fayetteville in Cumberland County and ends in the upper-half of Bladen County, near the coast. 

The 2018 congressional election was noted as a competitive one, even though the partisan dynamics indicated a GOP advantage. In the end, Republican Mark Harris defeated Democrat Dan McCready by 905 votes (unofficially at this point). 

Harris won 2 counties--suburban Union County and rural Bladen County--while McCready won the remaining six counties. By vote method, Harris won on Election Day, while McCready won both absentee methods (one-stop/in-person and by mail):

What originally piqued my interest was the decision by the NC State Board of Elections to withhold certification of the 9th district's results, and the potential issue with Bladen and Robeson counties that I was hearing about in regards to allegations of missing votes. 

In doing some investigations (and writing up some preliminary analysis), I discovered that in Bladen County, the 9th's absentee by mail ballots went 61 percent for Republican Mark Harris, but that the party registration within that vote method's accepted ballots was only 19 percent registered Republican:

As I noted to the Charlotte Observer's Jim Morrill:

Republican Mark Harris won 61 percent even though registered Republicans accounted for only 19 percent of the county’s accepted absentee ballots.
Unaffiliated voters accounted for 39 percent. Bitzer said Harris’ margin “could potentially come from all those unaffiliated voters.”
“But to have each and every one of those unaffiliated voters vote Republican, that’s pretty astonishing,” he added. “If that’s the case, there’s a very concerted effort to use that method to one candidate’s advantage. . . . But at that level there’s something else beyond a concerted effort that could be at work.”

As noted in a previous analysis, Bladen County was the only county where Republican Harris won the absentee by mail vote:

With these questions and the NC State Board of Election's 7-2 decision to initiate an investigation, I decided to break down the counties' ABMs, first for all of Bladen County, and then the portion of voters who were only in the 9th, and finally for Robeson County (the entire county is within the 9th).

First, in looking at the entire 9th and voters' use of ABMs, there were differences regarding the return rates among voter party registrations:

Of course, some of the "non-return ballot" voters could have used in-person voting on Election Day if they didn't get their mail-in ballots, forgot to complete it, or didn't have the opportunity to complete that vote method's requirements. The final Election Day data is not yet finished from all the counties, so that will be another project to investigate, but in looking at all the 9th's non-returned ABM ballot registered voters (3,478), 73 percent (2,546) did not end up casting an Election Day vote (either in-person, provisional, transfer, or curbside). 

So, over 2,500 voters who requested absentee by mail ballots--but did not return those ballots--did not end up voting in the 2018 Ninth Congressional District election. 

In looking at the eight counties within the 9th, there are two notable counties with lower than the district's average accepted ballot rates--Robeson and Bladen:

Taking the idea of voters who didn't return their absentee by mail ballots and whether they voted in the 2018 election on Election Day, preliminary numbers show that out of 1,675 "non-return ABM" 9th CD voters in Bladen and Robeson counties, 1,192 did not cast a ballot on Election Day, while 483 did cast a ballot on Election Day.

Thus, the focus on these two counties and their rates by both voter party registration and voter race.

Diving Deep into Bladen & Robeson Counties:

Two counties within the 9th Congressional District stuck out to me in regards to their non-returned ABM rate.

Within Robeson County, 62 percent of the requested absentee by mail ballots were not returned, making it the highest county in the 9th District with that non-return rate percentage. The county with the second highest non-return rate percentage of requested ABMs in the 9th was Bladen County.

The analysis first turns to Bladen County and its breakdown by voter party registration and the various rates, first for the entire county:

Across Bladen County, 61 percent of the absentee by mail (ABMs) ballots were returned (56 percent accepted and 5 percent rejected), with nearly 40 percent not returned. But by party registration, Bladen County registered Republicans had a higher than county-wide average for not returning their ballots, while registered Democrats had a higher than county-wide average for returning and having their ballots accepted. 

Then in looking at the portion of Bladen within the 9th Congressional District:

Over a third of registered Democrats in Bladen's 9th Congressional District did not return their ballots, while 41 percent of registered unaffiliateds and 47 percent of registered Republicans did not return their ABM ballots. Bladen County registered Democrats in the 9th had the highest return and accepted ballot rate, while 54 percent of registered unaffiliateds and 47 percent of registered Republicans returned their ballots and were accepted. 

In terms of real numbers, 164 of the 473 registered Democrats in Bladen's 9th District didn't return their ballots, while 202 out of 487 registered unaffiliateds didn't return their ballots, and 129 out of 276 registered Republicans did not return their ballots. 

The main observation would be that Bladen's 9th Congressional District ABMs had a higher than state and 9th district non-return rates, but there were differences among the registered parties, in both percentages and raw numbers. While percentage-wise, registered Republicans appear to have suffered the most, registered unaffiliated and Democratic voters had more (in actual numbers) of ABM ballots not returned.

Among Robeson County (which is entirely within the 9th District), a different pattern emerges among absentee by mail ballots:

In terms of raw numbers, 802 out of 1,232 registered Democrats didn't return their ballots, 296 out of 484 registered unaffiliated didn't return their ABMs, and 81 out of 185 registered Republicans didn't return their ABMs in Robeson County. 

I decided to look at two other counties within the 9th which produced the largest percentage of votes in the election: Mecklenburg and Union counties.

First Mecklenburg County, which is divided between the 9th and the 12th districts, so this is the entire county by voter party registration and their respective rates:

And then for the portion of Mecklenburg that contains the 9th:

The 9th in Mecklenburg had a slightly higher accepted rate and lower non-return rate, while the party registration percentages were nearly the same between the 9th and the full county.

For Union County, which is completely inside the 9th:

In comparison to Mecklenburg's 9th district component, Union had a slightly lower accepted ballot rate and higher non-returned ballot rate for its absentee by mail ballots, with some variation amongst the party registrations. 

From Party Registration to Voter Race:

While party registration may give some indications regarding the use of absentee by mail vote methods, another factor that may give insight is the voter's race within the counties. That is where the analysis next turns.

First, the state-wide rates by voter race (the percentages on the left side indicate the percent of the absentee by mail electorate for each racial category):

State-wide, white voters generally had a higher ballot accepted rate, while black/African-American voters had a lower rate than the state's overall rate, and Native American voters had the lowest, with only a majority of requested ballots being returned and accepted. 

For the 9th Congressional District, this same analysis shows a similar pattern to the state's rates by race:

In the 9th, whites again had an above average accepted ballot rate for the district, while black voters had a rate of 54 percent and Native Americans were at 38 percent and a non-return ballot rate of 61 percent. 

In turning to all of Bladen County, 56 percent of voters who requested absentee by mail ballots returned them and had the ballots accepted, with black voters having the highest percentage of the groups with an accepted ballot rate. Half of the Native American voters did not return their ballots, but their percentage of the ABM electorate was 1.5 percent of requested ballots.

And within the 9th's portion of Bladen County, the racial composition of the ABM electorate was more white than the county as a whole (whites were 68 percent of the ABM requests to 60 percent for the entire county):

With Bladen's 9th District, white and black voters had similar accepted ballot rates and non-return ballot rates. 

In turning to Robeson County, the differences between racial groups is much more substantial than in Bladen County (again, a reminder that all of Robeson is within the 9th District):

While black voters were a plurality of the requested ABMs, a little over a quarter of those ballots were returned and accepted, while barely a third of the ballots from Native American voters were returned and accepted. White voters in Robeson had an accepted ballot rate of 54 percent. 

Again, I compared the rates within Mecklenburg (for the entire county and that portion in the 9th) and Union County; first, all of Mecklenburg:

And by the 9th within Mecklenburg:

Again, much like with the voter party registration, there is not as great disparities between racial groups, but black Mecklenburg voters in the county and in the 9th did have a higher non-return rate. 

In Union County, there was a lower accepted rate compared to Mecklenburg's 9th portion, but again, there were differences among racial categories of voters:

Where Might This Lead?

While the allegations of suspected voter fraud, mainly by collecting, sometimes called "harvesting," absentee by mail ballots, is currently under investigation, the data shows that two counties--Bladen and Robeson--have potentially enough anomalies in their absentee by mail rates--both in accepted and non-return--that may warrant greater inspection. 

Perhaps voters in these two counties simply decided to request, but never return, their absentee by mail ballots; it may be as simple as that. It could also be that the campaigns, on both sides of the political fence, failed to motivate and/or follow-up with their ABM voters to get their ballots completed and returned; it could be as simple as that as well. Maybe it's the fact that both Bladen and Robeson voters (two rural counties) are just that different from Mecklenburg (urban) and Union (suburban) voters in having a different approach to voting; it could be as simple as that. But if there are other issues at hand than voters simply not returning their own ballots, the integrity of the voting process in the 9th Congressional District's election may be called into question.

Ultimately, the State Board of Elections may make that determination (based on whether it's current composition, or a new structure and membership, is developed) of whether to certify the election or not. Perhaps their scheduling of an evidentiary hearing before December 21 may give the voters a better sense of potential issues within the district. Ultimately, according to the North Carolina General Statutes, the State Board has the authority to consider the 9th's election contest and whether a new election would be warranted, based on one of the four following findings and a vote from 5 members of the board to call a new election:

  1. Ineligible voters sufficient in number to change the outcome of the election were allowed to vote in the election, and it is not possible from examination of the official ballots to determine how those ineligible voters voted and to correct the totals.
  2. Eligible voters sufficient in number to change the outcome of the election were improperly prevented from voting. 
  3. Other irregularities affected a sufficient number of votes to change the outcome of the election. 
  4. Irregularities or improprieties occurred to such an extent that they taint the results of the entire election and cast doubt on its fairness.
Nowhere in the statute does the Board have the power to certify an election because the number of questionable ballots would not produce a change in the overall election. But this will not prevent one side from arguing that the election should be finalized and certified, while the other side will contend that enough irregularities or improprieties have occurred to question, or 'taint,' the 9th's results. Needless to say, in North Carolina's hyper-polarized environment, this potential dispute over an extremely competitive election could only further divide the partisans.

As the national media has begun to cover this story over the past week, I ended a conversation with a reporter last week when she noted that I lived in a very "exciting state." I could only respond with "oh, you have no idea."