DETROIT – Errors by election workers in Detroit led to problems in the number of ballots not lining up with the number of people recorded as having voted in scores of precincts, a Bureau of Elections audit released Thursday found, and as a result the bureau plans comprehensive training to assist the city in better preparing its election workers.

Separate from the Detroit audit, the Bureau of Elections conducted its first ever statewide review comparing the list of those who cast absentee ballots against those who voted at the precincts on Election Day and found 31 apparent instances of the same voter voting twice. The bureau has new technology that enabled this first of its kind review.

The names of the 31 persons who appear to have voted twice were referred to the Department of Attorney General for further investigation. It is not clear whether the 31 deliberately sought to vote twice or whether some might have voted absentee, forgot they did so, and then voted at the polls too.

“Even one is troubling and that’s why we did want to refer it for further investigation,” Department of State spokesperson Fred Woodhams said.

The audit criticized 45 cities or townships that downloaded electronic poll book data for their workers to use on laptops more than two weeks before Election Day. As a result, those workers may not know whether voter showing up at the precinct already cast an absentee ballot, and that was the likely weakness that enabled 31 voters to vote twice, the audit said.

The severity of the problems in Detroit with gaps between the number of ballots cast at each precinct and the number of voters having been recorded as voting in each precinct poll book came to light during the partial statewide recount of the presidential vote although the Wayne County Board of Canvassers reported prior to the recount that about 392 precincts appeared out of balance. Under Michigan law, any precinct where there is such a gap cannot be recounted.

During the partial recount, the county board of canvassers attempted to recount 263 Detroit precincts and absentee voter boards and found 26 percent could be recounted mainly because of the gaps.

In many cases, the difference was between one and five. But in a couple cases, the gap was in the hundreds. As suspected, those larger gaps were the result of election workers forgetting to remove ballots from the tabulator at the end of the night and placed them into the sealed container. The votes were counted, but not stored properly.

“The audit refutes suspicions that the relatively small number of ballots placed in the ballot container could have been illegally tabulated again and again,” the audit said.

The bureau selected 136 of the city’s “most irregular and questionable precincts” and absentee boards for the audit, the audit said.

Forty-eight percent of the audited precincts could have been brought into balance on election night had election workers or receiving boards “taken prompt remedial action” to address the discrepancies, the audit said.

In the precincts that the state could not balance, the gap was narrowed by nearly 70 percent.

Prior to the audit, the imbalance among the 136 precincts was 586. After the audit, that was down to 216

The audit found that many of the problems in Detroit stemmed from election workers lacking proficiency in using electronic poll book software or mishandling provisional envelope ballots, which voters cast when the election workers cannot find their name in the poll book, but the voter insists he or she is at the proper precinct.

The audit showed 17 percent of the 136 precincts reported tabulator malfunctions on Election Day. City elections officials and some Detroit legislators had blamed malfunctioning equipment for the city’s election problems, but the audit said the 17 percent figure “reinforces the conclusion that other errors unrelated to tabulator performance were the primary reason for imbalance.”

Electronic poll book errors occurred in 78 percent of the 136 precincts audited. The electronic poll book is supposed to improve upon the traditional paper poll book by automated functions and reducing the chance of error, the audit said.

“However, BOE observed numerous instances in which precinct inspectors failed to properly enter vital data into the EPB such as the names of all voters who received a ballot, instances in which an individual voted by provisional ballot or spoiled their original ballot and received a replacement, or other occurrences that should have been noted in the ‘Remarks’ section,” the audit said.

The audit said a paper backup would have eliminated many of, maybe most, of the errors. The 2016 election cycle was the first in which all of Detroit’s precincts relied solely on electronic poll books, the audit said.

The audit recommended improving the city’s election worker training program through using the Bureau of Election’s library of materials, broadening election worker recruitment efforts to find people familiar with using computers and enhancing training with electronic poll books. It also called for the Legislature to pass, and Governor Rick Snyder to sign, a bill requiring counties to disclose the number of out-of-balance precincts in the certified canvass reports.

And the audit called for legislation requiring precinct chairs to pass a written test of about provisional ballot procedures and other Election Day nuances.

“The city of Detroit and the entire state have made progress in election administration in recent years, but this audit highlights key areas in need of improvement,” Secretary of State Ruth Johnson said in a statement. “I have directed Bureau of Elections staff to assist city election officials in making needed changes to poll worker training and recruitment efforts. Voters in Detroit and across Michigan deserve no less.”