The PPP was designed to keep employees of small businesses on payroll during the pandemic. Out of almost 4.9 million loans, the number of "jobs retained" is zero for 554,146 and blank for 324,122. Seven loans list negative job numbers.

Conversely, almost a thousand entries show 500 jobs for loans under $150,000, which is mathematically doubtful given that the aid is based on 2.5 times a firm's average monthly payroll.

In 209 of those cases, it implies an average monthly salary of $4 or less per employee. Taken together, those figures call into question the job numbers in one out of every five PPP loans.

The anomalies cast doubts about the accuracy of the data for the centerpiece of the $2.2 trillion relief package enacted in March, including whether it supported the 51.1 million jobs that the administration has touted.

Backlash over borrowings

The PPP program is already facing backlash for doling out millions of dollars to big-name law firms, Wall Street managers and companies with ties to President Donald Trump and other politicians.

Now critics say the data issues make it difficult to evaluate how well the program worked, especially because the names of borrowers were redacted for smaller loans that account for about 87% of the number of loans.

"We are spending, as American taxpayers, upwards of half a trillion dollars to purportedly help small businesses stay afloat," said Kyle Herrig, president of Accountable.US, a government-watchdog group that often criticizes the Trump administration. "We should know where the money went, how many jobs were saved, and right now with the data, we don't have that ability to say with any certainty."

The reported number of jobs is based on information provided by applicants, according to a spokesperson for the Treasury Department, which runs the PPP with the Small Business Administration.

While some borrowers may have erroneously omitted the jobs number, the total value of loans approved is consistent with supporting about 51 million jobs based on average small-business employee compensation, the spokesperson said.

Under the program, borrowers file their applications through an approved lender. After the SBA issues a loan-guarantee number for banks to disburse funds, the lender and borrower can agree to a lower the amount, the spokesperson said.

The SBA and Treasury didn't explain how some million-dollar loans in the data are much higher than some borrowers said they applied for and received.

Banks said the problem with the jobs data is that neither the PPP application nor SBA's electronic system that lenders use to submit applications required an input for "jobs retained."

The application had a box for "number of employees," and some lenders said they submitted that number while others said they left it blank.

Getting the jobs number right will matter even more when borrowers apply for loan forgiveness: The business owners will have to prove they maintained headcount and salaries to get their aid turned into a grant. If the SBA determines that a borrower is ineligible, the agency will direct the lender to deny loan forgiveness, the Treasury spokesperson said.

Supporters credit the PPP, created and implemented in a matter of days, for getting funding quickly to millions of U.S. small firms that self-certified they needed the funds as businesses were shuttered. They say such an unprecedented program rolled out so quickly was bound to have glitches.

Mistakes galore

Bloomberg News spoke to more than a dozen companies that, according to the government, received loans of more than $1 million with a reported one job retained. The borrowers all said there were mistakes in the dataset.

Frank Demandt, owner of a Miami architectural firm, said he had four employees and that the loan value was grossly overstated. He said he received $19,700, not the $1 million to $2 million range cited in the SBA data. Demandt's lender, BankUnited Inc., confirmed the loan amount and said it doesn't know why the data say otherwise.

In Star City, Arkansas, Gregory Smith was dumbfounded. The chief executive officer of C&L Electric Cooperative Corp., which supplies power to more than 20,000 customers in a rural part of the state, said in his application that the company had about 100 employees.

"If it shows one job retained, that's way off," Smith said.

Several companies, including scooter-rental service Bird Rides Inc., have complained that they appear in the data even though they didn't apply for or get aid.

"I actually never received a PPP loan," said Bridget Ottoh of the Ottoh Group in Mt. Juliet, Tennessee, who was listed as having been approved for a loan of $2 million to $5 million. "I applied for it and then withdrew it."

The SBA inspector general, who produced a report in May critical of the agency for not collecting demographic data to prioritize loans to underserved and rural areas as Congress intended, has a review ongoing about the implementation of PPP and is aware of public reporting about anomalies, spokesman Sheldon Shoemaker said.

Data analysis is part of the investigations and reviews, he said.

Miller, the Tennessee accountant whose loan was reported as between $2 million and $5 million in the SBA dataset, said he learned about the error when a friend saw his name in a list of PPP recipients published on a local newspaper's website and asked why Miller hadn't left town yet. The lender, Millennium Bank, couldn't be reached for comment.

"This whole thing is a farce," Miller said.