Every four years, Colorado’s Public Employees Retirement Association (PERA), the state’s public pension plan, conducts an experience study to re-evaluate its underlying economic and demographic assumptions. From the results of that study, PERA will usually adjust those assumptions, and recalculate its liability. For instance, the 2016 study made headlines when it revealed significant new liabilities as a result of lower mortality. It became the impetus for Senate Bill 200, the PERA reform package passed by the legislature in 2018.
SB-200 created the Pension Review Subcommittee, composed of legislators from both houses and both parties, and their appointed subject matter experts. I currently serve on that committee as an appointee of the House Republicans.
Last year, the subcommittee voted to commission an audit of the experience study and PERA’s adjustments in response. The purpose was to look for systematic biases in the assumptions that might understate or overstate liability, as well as reporting issues or ambiguities. We chose the firm of Gabriel, Roeder, Smith (GRS) to conduct the audit. It presented the results to the subcommittee in two parts, on August 19 and on September 2.
GRS found three anomalies with biases in them. These involved the withdrawal assumption, the retirement assumption, and the accounting for new entrants into the system. The withdrawal assumption is the percentage of members PERA expects to stay in the system until they are vested. PERA adjusted that number upwards, but GRS’s experience with similar large pension funds strongly suggests that the adjustment wasn’t large enough, meaning that it continued to understate the effect of employee withdrawal on the future liability.
Similarly, PERA adjusted its retirement assumption, but perhaps not by enough. With more people retiring early, PERA began paying benefits before it had realized all of the contributions it expected in order to fund them. In both cases, PERA’s likely under-adjustment came from over-weighting previous experience, and under-weighting the current experience found by the study.
Finally, new entrants into the system create an immediate obligation of about $220 million per year. Incorrectly, this is not recognized immediately, underestimating the normal cost of the pension and leading PERA to understate the future obligation in 2048 by around $9 billion. (The normal cost of a pension is the benefits earned in the current year.) Collectively, these three biases amount to “leakage” of about 5% of payroll, understating the increase in the obligation by about $500 million a year.
In response to these anomalies, all of which tend to understate PERA’s future obligations, GRS recommends that the next experience study be moved up two years to 2022. First, it will give PERA a chance to see how much, if any, of the 2020 results are a result of transient Covid effects. As important, there is a chance that PERA’s benefit and contribution adjustment formula might indicate a relaxation of those stringencies, erroneously leading to higher benefits and lower contributions that will have to be reversed within a few years.
GRS also identified a problem with how losses are amortized, that is, the number of years available to pay them off. According to SB-200, PERA must target 2048 for full-funding. As a result, the overall liability prior to 2018 is amortized to 2048. However, new losses are amortized at 30 years from when they’re incurred. This creates confusion in the reporting and allows losses to be amortized past the 2048 target date for full funding. GRS recommends amortizing new losses to 2048, at least until 2033, when volatility could begin to be a problem. This practice is recommended by the Public Plans Community of the Conference of Consulting Actuaries (CCA).
In recent years, in addition to its Comprehensive Annual Financial Report (CAFR), PERA has issued what it calls a Signal Light Report, showing the likelihood that PERA will meet its funding obligations by various dates. GRS recommends that should include historical reconciliation of gains/losses by source. Not only would this help validate the assumptions, or not, it could point to where future changes will be needed to correct for biases in those assumptions.
In addition to recommendations, the auditor pointed out the current difference between the actuarial value of assets, which is smoothed over four years, and the market value of assets, which is not. In the past, because of sudden market downturns, smoothed returns have overstated the asset values, giving the impression that things were better than they were. Currently, because of two years of good returns, the smoothed assets understate the level of funding. The auditors noted that when the deferred returns are finally recognized in the actuarial assets, they could force the plan to relax the adjustments prematurely.
Unfortunately, the auditors didn’t stop at these recommendations. The Sept. 2 discussion moved on to some terrible ideas that we thought had been buried, but now seem on their way to resurrection. That will be the subject of a future article.
Joshua Sharf is a Senior Fellow at the Independence Institute, a free market think tank in Denver, and serves on the Pension Review Subcommittee.