As the Fresno Bee recently reported, the city of Fresno’s pension systems are in much better financial shape than the Fresno County Employees’ Retirement Association (FCERA). As of June 30, 2015, the city’s two systems reported a combined $349 million of assets (at market value) in excess of actuarially accrued liabilities. By contrast, FCERA’s assets were $1.043 billion below its liabilities. Actuarial surpluses are rare in California, and the discrepancy between the city and county is so great that we thought it would be worth diving into the finances of Fresno’s retirement system to explain the contrast.
The systems provide extensive financial reports on their websites. The two most useful are Comprehensive Annual Financial Report (which includes financial statements and 10-year histories for many data points) and the Actuarial Valuation Study (which provides in-depth data about system assets, contributions and benefit payments). FCERA posts its reports at http://www.fcera.org. The two city systems – one for Fire and Police, and one for non-public safety employees – publish their reports at http://www.cfrs-ca.org/.
Table 1 below compares some key metrics across the plans.
The Valuation Value of Assets (VVA) is used by system actuaries to determine future contributions. But for our purposes, VVA is less useful than the Market Value of Assets (MVA). While MVA is simply the total market value of all the bonds, stocks and other investments the system holds, VVA includes various smoothing adjustments – reporting practices, some of them legitimate, that can mask liabilities.
Contributing to the difference in the financial health of Fresno’s city and county systems is the difference in benefit levels. For example, the City of Fresno limits public safety pensions to no more than 75% of final average salary (as per Section 3-333 of the municipal code). The county imposes no similar cap and also provides very generous benefit accrual rates, in some cases exceeding 3% per year of service. According to calculations we performed using the county system’s Benefit Calculator, a Tier 1 public safety employee retiring at age 60 with 30 years of service would get a pension equal to 97% of final salary. Tier 1 employees were hired before 2007; newer county employees receive less generous benefits.
The city does not cap miscellaneous employee benefits, but its employees earn substantially less credit for each year of service. According to the city’s Benefit Calculator, a miscellaneous employee retiring at age 65 with 30 years of service would receive 72% of final compensation, compared to 97% for a county employee retiring at the same age and the same number of service years.
Another potential distinction between the city and county systems is investment performance. A pension plan can improve its actuarial balance by achieving higher asset returns. Over the five years ended June 30, 2015, the city’s investments outperformed the county’s. FCERA generated annualized investment returns net of fees of 9.8%. The two Fresno city systems, whose assets are jointly managed, achieved net returns of 10.9% over the same period. This 1.1% difference compounded over five years is fairly significant. One billion dollars growing at the county’s rate of 9.8% becomes $1.596 billion after five years, while the same amount growing at the city’s 10.9% annual rate becomes $1.677 billion – $81 million more.
However, when we look at the 10-year period that includes the Great Recession, the performance numbers reverse. Over the 10 years ended June 30, 2015, county assets grew at an annual rate of 6.8% versus 6.4% for the city. Both of these return rates are below the annual asset return rates assumed by each system (more on this below).
Further, it’s worth noting that funding levels for all systems declined over the 10-year period. Between June 30, 2005 and June 30, 2015, FCSERA’s funded ratio based on VVA declined from 91.5% to 80.7%. The Fresno Fire & Police plan saw a decline from 126.4% to 119.6%, while the city’s Employee Retirement System witnessed a funded ratio decline from 139.8% to 109.2%.
In general, the market value of a plan’s assets is fairly easy to determine and is not subject to substantial estimation error. Most plan assets are invested in stocks and bonds that trade frequently and whose values are easy to establish independently.
By contrast, plan liabilities are based on numerous assumptions. How much a plan will have to pay in the future depends upon when employees retire and when they pass away. Expressing these future benefit payments in current dollars requires the choice of a discount rate – a choice subject to controversy.
Fresno city plans use a higher discount rate than FCERA. The city’s ERS and Fire & Police plans both assume annual returns of 7.50% and then use that rate to discount future benefit payments. FCERA uses a slightly more conservative rate of 7.25%. Both of these assumptions exceed the actual 10-year returns experienced by the city and county pension systems, and thus should arguably be reduced.
But to compare the systems, we don’t need to determine the ideal discount rate; we simply need to apply the same rate to each system. If we reduce the city’s discount rate from 7.5% to 7.25%, pension liabilities across the two city systems would increase about $61 billion and their funded ratio would fall by about 3.5%. (These estimates are discussed in an appendix at the end of this study). While significant, this fact only explains a small portion of the 38.3% gap in funded ratios between the city and county systems.
While the pension literature includes much discussion of discount rates, less has been written about mortality assumptions. But good death rate estimates are important: if beneficiaries live a lot longer than expected, pension payments will be much greater than forecast. This recently became clear in Detroit, where city officials faced a sudden spike in projected retirement payments after its pensions actuary switched to a new mortality table.
Mortality tables are produced by the Society of Actuaries. Most public pension plans use a table from the Society’s RP-2000 Mortality Tables Report produced in the year 2000. The large increase in Detroit’s projected pension costs occurred after actuarial firm Gabriel Roeder switched to the Society’s new RP-2014 Mortality Tables.
The RP-2000 report included a supplemental schedule that can be used to scale mortality rates to future years. The scaling procedure assumes a steady improvement in longevity, and thus a steady decrease in mortality rates over time. By applying the adjustment factor from the scaling schedule multiple times, an actuary can approximate what a future mortality table might look like. For example, by applying the scaling factors to the 2000 mortality rates 15 times an actuary can approximate 2015 mortality rates. In Detroit, Gabriel Roeder did not apply the scaling factor, thereby causing the big change when it transitioned to the newer mortality table.
Both the Fresno city plans and FCERA use the RP-2000 Combined Healthy Mortality Table and then scale the death rates from this table with factors in Mortality Projection Scale AA. However, there is an important difference. The city performs the scaling six extra times: it uses mortality rates scaled to 2021, while FCERA uses death rates scaled to 2015. This means that the city plans are projecting fewer deaths at any given retiree age – and therefore greater liability – than does FCERA.
The county’s mortality projections are thus more “optimistic” than those of the city plans, in the sense that its approach anticipates shorter-lived recipients – and that translates into lower expected benefit payments. The sooner an employee is assumed to pass away the less he or she is projected to receive from the system. If FCERA performed the same scaling as the city plans, its reported funding level would be worse. Without more data, we cannot say how much worse.
Finally, it ‘s worth noting that retirement rate assumptions differ between the city and county systems. The difference may be justified, and the impact is unclear. Since the plans have different benefit structures, they present different incentives to workers timing their retirements. When an employee retires early, he or she will receive benefits for more years but generally at a lower rate. So a change in retirement-age assumptions, may raise or lower projected system costs.
Overall, our conclusion is mixed. Fresno’s Employee Retirement System and Fire & Police Retirement System offer less generous benefits that the Fresno County Employees’ Retirement Association. This difference in benefit levels makes a substantial contribution to funding disparities between the systems.
FCERA uses a more conservative discount rate, while the city plans use more (financially) conservative mortality assumptions. These modeling differences affect the disparity between reported city and county funding levels, but they do not represent real differences and simply muddy our understanding of relative system performance. Ideally, all California pension systems would use the same actuarial assumptions (unless there are real demographic differences between workforces) so that we would be able to perform accurate comparisons.
Appendix: Recalculating AAL Using a Different Discount Rate
A pension system’s AAL is the discounted amount of future benefit payments. Unless one has a table of projected future benefit payments, it is impossible to precisely calculate AAL using another discount rate.
In 2013, Moody’s adjusted pension liabilities by using more conservative discount rate assumptions. The rating agency’s method of restating liabilities involves projecting forward the system’s reported liability for 13 years and then discounting the result back for 13 years using the more conservative rate. Moody’s refers to the 13-year re-discounting period as a “common duration” and recognizes that applying the same duration to all plans could be a source of estimation error.
Moody’s also noted at the time that more precise estimates would be possible once pension plans implemented enhanced reporting required under Government Accounting Standards Board Statements 67 and 68.
Under these new rules, pension systems must report the “Sensitivity of Net Pension Liability to Changes in the Discount Rate.” This new schedule shows the Net Pension Liability calculated using the current discount rate, a rate 1% higher and another rate 1% lower. For the Fresno city systems, we have Net Pension Liabilities based on rates of 6.5%, 7.5% and 8.5%.
We can estimate the impact on Net Pension Liability by linearly interpolating between the 6.5% and 7.5% values. For the two Fresno systems combined, the estimated impact of a change in discount rate from 7.5% to 7.25% is $69 billion.
Net Pension Liability as reported under GASB Statement 67 is higher than each system’s Actuarially Accrued Liability. In the case of the Fresno city systems, the difference is about 11.5%. If we reduce the Net Pension Liability difference of $68 billion by 11.5%, we arrive at the $61 billion estimate presented in the main text.
The author wishes to thank Lisa Schilling at the Society of Actuaries and Bill Bergman of Truth in Accounting for their assistance with some technical points in this study. Any errors are my responsibility.