Mark Harris, the Republican who is the presumptive winner of last month’s race in North Carolina’s 9th Congressional District, released a statement last week agreeing that a new election should be held in the district — if allegations of election fraud were shown to have possibly affected the contest’s outcome. That echoed a similar comment first made by the executive director state Republican Party shortly after the fraud allegations emerged.

“There has to be enough votes in question to possibly change the outcome,” Dallas Woodhouse told the Charlotte Observer on Dec. 3.

That’s not true. State law allows the board of elections to call a new election under four conditions, one of which is if “irregularities or improprieties occurred to such an extent that they taint the results of the entire election and cast doubt on its fairness.” That’s a lower bar than the one Harris and Woodhouse are trying to establish.

Harris and Woodhouse, of course, have an incentive to set that higher bar: Without a new election, Harris is going to Washington. They also certainly know that proving that the results of the election were shifted in Harris’s favor would be almost impossible. The fraud that’s alleged to have occurred involved employees of a campaign consultant named Leslie McCrae Dowless having collected mail-in absentee ballots and potentially either altering votes or never submitting completed ballots to the state. Determining the scope of those changes with precision would be difficult.

The data we do have, though, suggest the possible change to votes in two counties — Bladen and Robeson — might have been enough to change the result of an election in which Harris leads by 905 votes.

There are two aspects to the vote in those counties that have drawn scrutiny. The first is the margin by which Harris won the mail-in absentee vote in Bladen County over Democrat Dan McCready. The second is the number of requested mail-in ballots that weren’t returned in those two counties.

Dowless’s effort, reporting has indicated, included having his employees get voters to request mail-in ballots and then to collect them — not necessarily after they were completed — for submission to the state. If Dowless and his team had access to those ballots, they could either have added votes for Harris to blank ballots or thrown away ballots that were already marked for McCready. In the first case, Harris might overperform in a county. In the second, McCready might underperform.

Data from the state and provided to The Washington Post by Catawba College politics professor Michael Bitzer shows how the vote looked in the 9th District, by county. What’s important is those places where Bladen and Robeson were unusual. (These are the counties that WSOC-TV in Charlotte, which has driven a lot of the reporting on this election, identified as ones in which Dowless’s team was active.)

We’ve flagged three things worth noting.

1: The percentage of votes received by the Democrat in Bladen County was much lower than in mail-in votes across the district, and the percentage received by Harris was much higher. Bladen was the only county where Harris won the mail-in vote.

2: In Bladen and, especially, Robeson County, the percentage of ballots requested but not returned was much higher than in the district overall. In Robeson, a much higher percentage of those ballots requested by Democrats didn’t end up being turned in than ballots from Republicans. Nearly two-thirds of requested ballots from Democrats in that county didn’t get turned in.

3: Remember that the oddities in Bladen and Robeson counties mean that in the rest of the district, percentages swung the other way. If Robeson and Bladen had an unusual number of unreturned ballots, the percentage of unreturned ballots in the rest of the district was lower than the overall figure.

So let’s compare the number of unreturned ballots in Bladen and Robeson Counties with those other six counties. We do a simple calculus: If ballots were returned in those two counties at the same rate as the other six, how many more ballots would have been returned?

In Bladen County, 78 of 164 ballots from Democrats, 96 of 129 ballots from Republicans and 128 of 202 ballots from unaffiliated or minor party voters would have been returned, a slight net gain for the Republican (ignoring how the independents voted).

In Robeson, the results are striking: 579 of 804 Democratic and 225 of 299 votes from unaffiliated and independent voters would have been returned. An additional 56 Republican ballots would have been returned, as well.

Remember, the theory is that Dowless’s team allegedly knew who those ballots had been cast for. The gain for McCready, then, could have been disproportionate among those non-major-party voters. McCready could have gained hundreds more votes.

Now, let’s look at those mail-in ballots in Bladen County. We can adjust that number two ways.

First, we have the vote for Harris reflect the vote in the other six counties where Dowless wasn’t working. That’s the middle pair of columns above; Harris goes from up 162 to down 181 votes.

But that’s unfair. The percentage of votes cast by mail-in ballot was also up in Bladen County. If the same percentage of the total vote there came from mail-in ballots and if the margin was the same as in the other six counties, Harris is now down 94 votes. Overall, he’s lost 291 votes from the actual results; McCready has lost 35.

That’s a swing of another 256 votes.

You can see how murky this gets. A net gain of 500 votes when we adjust for the unreturned ballots plus a question mark over those 225 third-party ballots. A swing of 256 when Bladen’s mail-in ballot is adjusted. In a race where Harris leads by 905 votes cast.

This is by no means definitive evidence that Harris won only because of fraudulent activity. The margin by which he leads is narrow, but most of the votes cast in the race came in the more populous counties of Mecklenburg and Union. Combined, Robeson and Bladen made up only about 14 percent of the vote.

What can’t be said at this point, though, is that the results couldn’t have been affected by alleged fraudulent activity. The state is conducting an investigation that may answer the question. Until that’s completed, the situation seems to fall within the purview of that fourth allowable reason for rerunning an election: irregularities that taint the results of the entire race.