Which pending voting-rule changes could materially affect competitive midterm contests, and through what mechanisms?

A’s strongest point is that operational ballot-processing changes deserve investigation even when statewide turnout barely moves. I agree. The weakest point is evidentiary: A supplies neither a measured incremental failure rate nor a race-specific margin distribution. The illustrative 5–20% failure range establishes mathematical possibility, not demonstrated electoral risk. A watchlist is justified; prioritizing these disputes as likely determinants is not yet justified.

First, North Carolina’s administrative instructions describe recovery opportunities, not automatic exclusion. Registration repair permits updates online, in person, or through prepaid mail; unresolved voters can supply identification information when casting provisional ballots. People lacking both identification numbers have an alternative procedure. Thus, the approximately 82,000-person mailing cohort cannot substitute for the number still unresolved, likely to vote, and ultimately unable to have votes counted. Voter-roll maintenance means keeping registration records accurate through updates and lawful removals; this repair program should not be conflated with mass removal. (ncsbe.gov)

Likewise, ballot curing—correcting eligible defects through documentation or attestation—remains available by direct email, fax, mail, or personal delivery under the September 4, 2026 memo. The deadline is noon on the fifth business day after Election Day. The third-party portal prohibition is therefore not a prohibition on electronic curing. My inference is that its incremental effect depends on how many voters cannot successfully substitute another permitted channel, not everyone receiving a cure notice. A must also distinguish newly imposed restrictions from previously existing defect rules. (ncsbe.gov)

Second, comparable research cautions against converting burdens into net turnout losses. Cantoni and Pons’s 2021 study used 1.6 billion voter observations from 2008–2018: its demanding specification estimated a −0.1-percentage-point turnout effect from strict identification laws. However, its 95% confidence interval, −3.0 to +2.8 points, does not exclude consequential effects. Campaign contact with nonwhite voters increased 4.7 points, suggesting mobilization might offset modest burdens. This is evidence against assuming mechanical disenfranchisement, not proof that North Carolina’s distinct procedures are harmless. (doi.org)

Third, legal status narrows the question. Watson’s June 29, 2026 decision rejected the asserted federal election-day receipt requirement. A correctly excludes that nationwide threat. A stay temporarily suspends an order; Purcell generally disfavors late federal-court changes, rather than resolving their legality. Merrill illustrates why an injunction and the rules actually governing voting can differ. (supremecourt.gov)

My proposed threshold is race-specific incremental ballot loss after adaptation, multiplied by the affected voters’ net candidate preference. Under A’s hypothetical 60–40 split, 500–2,000 lost votes shift the margin only 100–400 votes: potentially decisive against a 100-vote margin, not 1,000 or 10,000. Those are stress tests, not forecasts. My methodological recommendation is to retain candidate strength and mobilization as the baseline until administrative outcomes, comparable causal evidence, operative orders, and multiple plausible margins support a rule-specific exception.

A’s specific weakness is the gap between identifying a mechanism worth investigating and establishing a potentially decisive net effect. A concedes that its failure rates are unsupported sensitivity assumptions, that Texas cannot supply North Carolina’s effect size, and that the cure memo requires a version comparison. Those concessions substantially narrow our disagreement: A has defended an administrative watchlist, not demonstrated an exception to candidate strength and mobilization as the explanatory baseline. I accept that investigation requires less evidence than prediction.

The Texas evidence strengthens that watchlist, but does not complete the causal chain. Morris and Grange report that nearly 90% of approximately 30,000 people whose mail applications or ballots were rejected did not participate in the 2022 primary. Their summary also reports estimated subsequent general-election turnout reductions of 16 percentage points following application rejection and approximately 1.5 points following ballot rejection. Those differentiated estimates are more informative than treating all rejected voters as interchangeable. They demonstrate substantial harm in a particular setting, not a transferable North Carolina failure rate. The reported slightly higher ballot-rejection rates among Republicans also preclude assigning the aggregate loss to Democrats. (brennancenter.org)

North Carolina’s procedures supply additional reasons to measure rather than presume either successful adaptation or exclusion. The repair program permits advance updates and collection of missing identification information when voters cast provisional ballots; alternatives exist for people lacking both identification numbers. Its approximately 82,000-person mailing cohort is therefore an exposure denominator, not a lost-vote estimate. A’s proposed district-level tracking is appropriate, but proposed measurement is not yet measured impact. (ncsbe.gov)

For ballot curing—correcting eligible defects with documentation or attestation—the September 4, 2026 memo requires initial review within one business day and notification within one business day after identifying a curable deficiency. It preserves direct email and fax submissions alongside physical delivery, despite prohibiting third-party portals. These safeguards do not prove successful recovery; they identify implementation variables that any comparison must include. The operative cure deadline is noon on the fifth business day after Election Day. (ncsbe.gov)

Legal filtering further limits the relevant universe. Watson’s June 29, 2026 decision rejected the federal statutory theory requiring Election Day receipt, without eliminating state deadlines. A stay temporarily suspends an order; the Purcell principle generally disfavors late federal-court changes to election rules. Consequently, a complaint or appeal cannot substitute for checking the operative order and implementation calendar. (supremecourt.gov)

My proposed decision rule remains demanding but falsifiable: estimate incremental counted-vote losses after recovery and mobilization, allocate them across actual contests, and test candidate-preference uncertainty against several independently justified margins. Hypothetically, 2,000 lost votes split 60–40 shift a margin by 400; split 50–50, they shift nothing. The former could overturn a 100-vote deficit, not 1,000 or 10,000. A supplies no evidence selecting those inputs for a competitive 2026 race. My methodological recommendation is therefore targeted scrutiny without upgrading an unquantified risk into an electoral explanation.

A’s strongest argument is that investigation must precede definitive causal estimates: an operative requirement, an identifiable population and a credible ballot-loss mechanism justify scrutiny before anyone proves an election will change hands. Texas’s experience demonstrates that alternative voting channels do not guarantee recovery, so administrative safeguards cannot establish harmlessness. (brennancenter.org)

I accept that research threshold. My position remains stronger for assessing electoral consequences because A’s final argument establishes a watchlist, not a demonstrated exception to the candidate-strength-and-mobilization baseline. Conversely, my baseline is a methodological recommendation—not proof that most disputes have negligible effects. Neither uncertainty nor failure to exclude a consequential effect establishes that such an effect is probable.

North Carolina illustrates the missing evidence. Its approximately 82,000-person mailing cohort is not the current unresolved population; the state provides a daily-updated search tool, advance repair options and identification completion through provisional voting. Voter-roll maintenance means keeping registration records accurate; this particular repair process should not be described as automatic removal. We need district-linked unresolved counts and subsequent ballot dispositions before estimating electoral reach. (ncsbe.gov)

Ballot curing means correcting eligible defects through documentation or attestation. The September 4, 2026 memo prohibits third-party submission portals but permits email, fax, mail and personal delivery, with receipt required by noon on the fifth business day after Election Day. Consequently, the relevant mechanism is incremental failure to complete an authorized cure—not every defective ballot or prohibited submission. A version comparison must first establish which restrictions actually changed. (ncsbe.gov)

Texas provides a warning, not a North Carolina coefficient. Morris and Grange report nearly 90% nonparticipation among approximately 30,000 rejected applicants or ballot submitters in the 2022 primary, alongside estimated subsequent general-election turnout reductions of 16 percentage points after application rejection and approximately 1.5 points after ballot rejection. These distinct outcomes cannot be pooled into a transferable failure rate or partisan margin effect. (brennancenter.org)

Legal implementation is another necessary filter. Watson’s June 29, 2026 decision rejected the asserted federal Election Day receipt requirement. A stay temporarily suspends an order; the Purcell principle generally disfavors late federal judicial changes to election rules, as discussed in the Merrill concurrence. Pending appeals therefore require checking operative orders and implementation calendars, not assuming intervention. (supremecourt.gov)

I would change my assessment upon seeing audited, district-linked records identifying rule-specific failures, notification timing, cures, replacement voting and final dispositions, combined with a credible comparison design accounting for prior participation, demographics and campaign assistance. The estimated counted-vote effect must then be translated into candidate-margin effects using defensible preference estimates, not party registration alone. Hypothetically, 2,000 lost votes split 60–40 shift a margin by 400 votes: consequential against 100 votes, not 1,000 or 10,000. Evidence supporting such an effect across several independently justified competitive-margin scenarios would establish A’s potentially decisive exception; proof of an actual flipped winner is unnecessary.