ATLANTA — Thou­sands of absen­tee bal­lots get reject­ed in every pres­i­den­tial elec­tion. This year, that prob­lem could be much worse and poten­tial­ly piv­otal in hot­ly con­test­ed bat­tle­ground states.

With the coro­n­avirus cre­at­ing a surge in mail-in bal­lot­ing and postal delays report­ed across the coun­try, the num­ber of reject­ed bal­lots in Novem­ber is pro­ject­ed to be sig­nif­i­cant­ly high­er than pre­vi­ous elections.

If bal­lots are reject­ed at the same rate as dur­ing this year’s pri­maries, up to three times as many vot­ers in Novem­ber could be dis­en­fran­chised in key bat­tle­ground states when com­pared to the last pres­i­den­tial elec­tion, accord­ing to an Asso­ci­at­ed Press analy­sis of reject­ed bal­lots. It could be even more pro­nounced in some urban areas where Demo­c­ra­t­ic votes are con­cen­trat­ed and bal­lot rejec­tion rates trend­ed high­er dur­ing this year’s primaries.

“It is the num­ber one thing that keeps me up at night — the idea that vot­ers will do every­thing they can to ensure their bal­lot is returned on time and the sys­tem will still fail them,” said Michi­gan Sec­re­tary of State Joce­lyn Benson.

Bal­lot rejec­tions occur even under the best of cir­cum­stances. They go uncount­ed because they arrived too late in the mail, vot­ers for­got to sign them or sig­na­tures didn’t match the one on file at local elec­tion offices.

Those prob­lems could be com­pound­ed this year as mil­lions of vot­ers cast mail-in bal­lots for the first time because of elec­tion changes forced by the coro­n­avirus. Large num­bers of uncount­ed bal­lots could be used to sow doubts about the elec­tion; Pres­i­dent Don­ald Trump has been claim­ing for months with­out evi­dence that wide­spread mail-in vot­ing will lead to fraud.

The sud­den leap is wor­ri­some: 22 states are going from absen­tee bal­lots com­pris­ing less than 10% of all bal­lots four years ago to per­haps half or more this Novem­ber. Penn­syl­va­nia is among them: Near­ly 51% of all votes cast dur­ing its June pri­ma­ry were mail-in.

If vot­er turnout is the same as 2016 and the bal­lot rejec­tion rate equals the 1.4% from this year’s pri­ma­ry, near­ly 43,000 vot­ers in Penn­syl­va­nia could be dis­en­fran­chised this fall, accord­ing to AP’s analy­sis. That’s almost the same num­ber of votes by which Trump defeat­ed Demo­c­rat Hillary Clin­ton in the state four years ago, when some 2,100 bal­lots were rejected.

In Philadel­phia, a far high­er per­cent­age of absen­tee bal­lots cast in June were reject­ed — 3.9%. Most of those arrived too late to be count­ed, even though Gov. Tom Wolf extend­ed the dead­line by a week.

Amy Camp­bell, a 26-year-old Uni­ver­si­ty of Penn­syl­va­nia med­ical stu­dent, had her bal­lot reject­ed for anoth­er rea­son: a miss­ing signature.

Camp­bell is upset she didn’t have a chance to fix her bal­lot. She received an email two days after the vote count end­ed telling her offi­cials “couldn’t obtain (her) required signature.”

“The first com­mu­ni­ca­tion I got from the state was basi­cal­ly to tell me that my bal­lot had been can­celed,” Camp­bell said.

Philadel­phia Board of Elec­tions spokesman Nick Cus­to­dio said that’s not sup­posed to hap­pen, but couldn’t explain why it did.

Vote-by-mail rejec­tions could be of spe­cial con­cern to Democ­rats, who have seen a surge in absen­tee bal­lot appli­ca­tions this year. In Penn­syl­va­nia, for exam­ple, more than twice as many Democ­rats as Repub­li­cans vot­ed by mail in the June primary.

Only 21 states have defined pro­ce­dures for noti­fy­ing vot­ers if absen­tee bal­lots are reject­ed so they have a chance to fix it.

For its analy­sis, the AP also col­lect­ed absen­tee bal­lot data from Ari­zona, Col­orado, Flori­da, Michi­gan, Ohio and Wis­con­sin. Based on the per­cent­age of those bal­lots cast in each state’s pri­ma­ry this year, between 185,000 and 292,000 vot­ers in the sev­en states exam­ined could be dis­en­fran­chised if November’s turnout match­es that of four years ago and the rejec­tion rate remains flat. That com­pares to near­ly 87,000 bal­lots reject­ed in those states in 2016.

The bal­lot rejec­tions could be piv­otal in close races. In 2016, Trump won Wis­con­sin by rough­ly 23,000 votes.

Rea­sons for rejec­tion can vary. In a few states, a wit­ness or notary must sign the bal­lot enve­lope for the bal­lot to be valid. Where bal­lots received after Elec­tion Day are count­ed, a leg­i­ble post­mark from that day is gen­er­al­ly required.

“There could be a lot of peo­ple who are vot­ing this way for the first time, and they tend to make the errors that lead to lost votes,” said Lar­ry Nor­den, an elec­tions expert with the Bren­nan Cen­ter for Justice.

New, young, Black and His­pan­ic vot­ers — who tend to favor Demo­c­ra­t­ic can­di­dates — are dis­pro­por­tion­ate­ly affect­ed, accord­ing to research by Uni­ver­si­ty of Flori­da polit­i­cal sci­en­tist Daniel Smith. All have had absen­tee bal­lots reject­ed at high­er rates than white and more expe­ri­enced mail-in voters.

Being able to noti­fy vot­ers of prob­lems in time to fix them will be dif­fi­cult enough for under­staffed elec­tion offices. It doesn’t help that elec­tion offi­cials in some states, includ­ing Michi­gan and Penn­syl­va­nia, are not allowed to begin look­ing at absen­tee bal­lots until Elec­tion Day. That leaves a nar­row win­dow for iden­ti­fy­ing prob­lems and allow­ing vot­ers to fix them.

Experts say vot­ers must be dili­gent about request­ing and return­ing absen­tee bal­lots well ahead of Elec­tion Day; the Postal Ser­vice advis­es allow­ing sev­en days for a bal­lot to reach an elec­tion office.

Drop box­es can help. Michi­gan, Geor­gia, Col­orado and Ari­zona are among states adding more. Col­orado, which has been mail­ing bal­lots to all vot­ers since 2013, saw a slight increase in rejec­tions from about 1% in 2016 to 1.6% in this year’s pri­ma­ry; the rejec­tion rate dou­bled in Den­ver but still remained below 2%.

Joe Bernal, a 31-year-old tech ana­lyst in Mia­mi, expects to be dri­ving to the near­est drop box to deliv­er his bal­lot ahead of Nov. 3. His bal­lot was reject­ed in Florida’s March pres­i­den­tial pri­ma­ry for arriv­ing after the dead­line, even though Bernal said he mailed it at least three weeks before.

Bernal, who lost an aunt to COVID-19, said he is unwill­ing to vote in per­son: “I’m asth­mat­ic. I don’t want to have to risk it.”

But drop box­es only solve one prob­lem. The process of authen­ti­cat­ing vot­er sig­na­tures on bal­lot envelopes — an anti-fraud mea­sure — is so trou­ble­some that law­suits have been filed seek­ing to end the prac­tice bar­ring an ade­quate remedy.

Jen­nifer Mor­rell, a for­mer elec­tion offi­cial in Utah and Col­orado, said vot­ers need to be cer­tain they will be giv­en an oppor­tu­ni­ty to fix a prob­lem. Some juris­dic­tions are adding soft­ware that quick­ly alerts vot­ers via text or email, but Mor­rell said the prob­lem is being uneven­ly addressed.

“States that are real­ly new to this, most like­ly they will just be send­ing out a let­ter in the mail and hope the vot­er gets it in time and fills it out,” she said.

That wasn’t the case for Tasha Young, a 43-year-old teacher in metro Atlanta who found a let­ter from her local elec­tion office buried in a pile of mail after Georgia’s pri­ma­ry in June. It said she had failed to sign her bal­lot, but by then it was too late for her to sub­mit the required affidavit.

She plans to vote in per­son in November.

“For one thing, I don’t want to have to wor­ry about a dead­line or miss­ing a sig­na­ture,” Young said.

Bajak report­ed from Boston.