After Pennsylvania’s biggest pension plan botched a crucial financial calculation, the FBI launched an investigation, the fund’s board began its own probe, and 100,000 public school employees suddenly faced paying more into the retirement system.

Now The Inquirer and Spotlight PA have obtained new internal fund documents that shed light on that consequential mistake. The material traces the error to “data corruption” in just one month — April 2015 — over the near-decade-long period reviewed for the calculation.

The error was small. It falsely boosted the $64 billion PSERS fund’s performance by only about a third of a percentage point over a financial quarter. Even so, it was just enough to wrongly lift the fund’s financial returns over a key state-mandated hurdle used to gauge performance.

The documents reveal that a fund consultant, Aon, blamed the mistake on its clerical staff for inputting bad data. The material also shows that even though the fund hired a consultant, the ACA Compliance Group, to check the calculations, the consultant made only limited checks, and skipped over the month with the critical errors.

Experts say the fund pursued a flawed path from the beginning. They said the plan executives erred months before the bungled calculation by rejecting a warning that it avoid unaudited numbers — those not verified by independent analysts. That choice to use new, untested numbers had the effect of making results look better than they really were. Then, these critics said, PSERS hamstrung the ACA review by limiting its scope.

As Robert Lavenberg, a CPA and lawyer who formerly headed the accounting giant BDO’s pension audits, said: If PSERS wanted to do a thorough review, “you would do a clean sweep” and check all months, not just some.

The headaches for PSERS — the Public School Employees’ Retirement System — began this spring when its board admitted endorsing a bad number for its investment profits. The mistake was significant because under a state pension reform, known as risk-sharing, working teachers and other school employees hired since 2011 must pay more to finance the pension system for 265,000 retirees if returns fail to meet benchmarks.

The board had declared triumphantly last year that its investment strategies had paid off. It claimed to have put the fund over the benchmark hurdle, sparing teachers and other staff from an increase. This April, it reversed course and abandoned the old performance figure. It officially adopted a new, lower one, and announced that the workers would indeed pay more. The increase kicks in July 1.

The board had little choice but to fix the number. A top tax lawyer warned the board that failure to do so would be “catastrophic” and force half a million current and retired school workers to pay future income taxes on pensions immediately.

Soon after the board’s reversal, The Inquirer reported that the FBI was investigating the calculation, along with the fund’s spending on Harrisburg real estate. In subpoenas, federal prosecutors have demanded every scrap of work paper, from drafts to final versions, from the pension fund and from consultants Aon and ACA dealing with the mistaken number.

PSERS has been conducting its own in-house probe. Records show that some involved with the review wonder whether the first figure was no mistake at all, but a deliberate attempt to inflate the fund’s performance.

And in the documents obtained by The Inquirer and Spotlight PA, state Treasury Department officials — Treasurer Stacy Garrity is a pension board member — asked: “Who determined or wrote the scope of work for the ACA project and how was it decided?”

ACA says it was not to blame and merely followed what it was asked to do. “ACA was hired to validate the calculation methodology over a defined period by testing on a sample basis, not to validate or calculate the performance return,” it said in a brief statement. “ACA performed the services it was engaged to provide — our calculation was based on the data provided to us, which we now understand contained an error.”

Indeed, in the documents, consultant Aon, a global firm with a $763,000 PSERS contract, says it was at fault for the bad numbers for April 2015. It blames the problem only on “inadvertent clerical mistakes at a data-entry level.”

In apologetic letters to the pension system, Aon was contrite even as it played down the mistake.

“On behalf of Aon, please know that we much appreciate PSERS patience as we have endeavored to unravel what very much appears to have been clerical data-entry mistakes, however unfortunate,” Steve Voss, head of Aon’s Chicago-based North American Investments, wrote on April 16.

In the letter, the firm does not name the employee or employees who made the mistake. It declined to discuss the calculation, as has PSERS.

Among the firm’s accounting tasks for the pension system, perhaps its most crucial was to compile that performance figure. To do that, Aon was required to review nine years of fund profits, from mid-2011 to mid-2020. The fund had to clear a “hurdle” of a 6.36% average annual return to spare teachers and school employees a new levy — and the fund’s leadership embarrassment.

To be sure, the bar seemed relatively low. After all, the S&P 500 index of big U.S. stocks paid over 10% a year during those years.

But the test loomed amid mounting criticism that the plan had for many years posted lackluster returns at best and, at worst, was a dupe of price-gouging, poor-performing hedge funds, and private equity billionaires.

Moreover, even before the calculation, PSERS board member Joe Torsella, then the state treasurer, had been questioning the fund’s executive director, Glen Grell, about why Aon and Grell had gone back years to reenter and improve the performance of some fund investments, using unaudited numbers. Grell replied that the changes were routine.

In sum, while the pension fund had easily cleared two earlier “shared risk” exams, in 2014 and 2017, the test for 2020 was a nail-biter. PSERS managers were nervous.

In a memo. Morgan Lewis, one of at least three law firms hired by PSERS after the scandal broke, said Aon was told about the fund’s anxiety about whether it would fall short, or as the Philadelphia law firm put it, “Aon was made aware of this sensitivity.”

Lavenberg questioned why Aon had been so alerted, saying that might have telegraphed the leaders’ desired outcome to the consultant. “Wink, wink. Nudge, nudge,” he said.

The board was also anxious enough that it hired ACA, based in New York, for $60,000 to check up on Aon’s work.

The board majority in December 2020 said it had cleared the key hurdle — albeit barely. It voted to state officially that annual investments had grown by 6.38%.

Torsella and two others on the 15-member board abstained from the vote, expressing doubts about the number. But the fund’s management insisted the new result was solid.

In fact, fund leaders in a PowerPoint presentation to the board and in a statement to the media said Aon and ACA had both endorsed the figure.

“ACA confirmed the nine-year market value return as 6.38%, which is higher than the 6.36% risk-share threshold,” the board was told.

But had ACA, in fact, verified the number?

In an April 16 private memo to the board, ACA explained its method was a sampling, not a verification. In the same memo, ACA explained why it had failed to catch the error. It said it sampled the math in only 40 of the 108 months over the review period from 2011 to 2020.

“April 2015 was not one of the months in ACA’s sample selection and therefore ACA did not detect missing cash flows in the AON calculations for that month,” the firm wrote.

As it happens, Torsella has long raised doubts about how the fund was measuring performance for the year 2015 — doing so months before the December vote.

Torsella first asked about the 2015 figures in a letter last August to Grell, noting skeptically that the fund staff was using unaudited figures and boosting performance for the year “five years after the fact.”

Grell’s reply was that the fund had new and better data for the performance of some investments in private companies. Experts say the valuation of such private investments is inherently subjective, unlike the explicit values placed on firms sold on stock markets.

John McLaughlin, an auditor who runs his own firm in Newtown Square after stints with Aramark and BDO, said the fund’s managers seemed in too much of a rush to approve the new numbers. “The board needs to push back [against management] if you need more time to get the right answer,” McLaughlin said in an interview. “Don’t blame the accountant. Blame the management.”

None of the documents obtained by The Inquirer and Spotlight PA explore or even mention the role of PSERS management in the reevaluation of the 2015 figures.

But according to experts who reviewed the documents, ACA in the end corrected PSERS’ data to the numbers its auditors had approved.

Said Lavenberg: “Why you would ever use an estimate, if you had an audited number for that same period, makes no sense whatever.”