A simple measure of pension generosity

ABSTRACT

Accurately assessing the generosity of government pensions is an important policy issue. This article proposes to measure generosity by the present value of benefits workers accrue per additional year of working net of their contributions. We show how a simple adjustment to standard pension metrics will approximate this measure. We conclude by documenting differences between our measure and other generosity metrics using data from Boston University’s Public Plans Database.

KEYWORDS:

• Pensions
• pension generosity
• retirement

JEL CLASSIFICATION:

• J3
• J45
• H75

Acknowledgements

We are especially grateful to Andrew Biggs for many valuable comments that have improved our article. This research received no specific grant from any funding agency in the public, commercial or not-for-profit sectors. Any errors are our own.

Disclosure statement

No potential conflict of interest was reported by the authors.

Notes

¹ In the U.S. states, Aubry and Crawford (2017) note that two-thirds of states making pension reforms adjusted benefit levels that apply only to new employees. The most common adjustments applied to age/tenure requirements and final salary calculations.

² The Public Plans Database is freely accessible at: https://crr.bc.edu/data/public-plans-database/[accessed 8/15/2019].

³ The replacement rate is the percentage of a retiree’s initial pension benefits expressed as percentage of final working income. Studies using replacement rates to gauge generosity are generally interested in adequacy issues related to retirement.

⁴ Cribb and Emmerson (2016) show that changing the indexation of benefits from the retail price index to the consumer price index (which is estimated to be 1.4 percentage points less per year) has a larger effect on reducing the value of benefits than changing the normal pension age from 60 to 65.

⁵ In addition, as Brown and Wilcox (2009), Novy-Marx and Rauh (2011) and Biggs (2011) (and others) have noted, pension liabilities should be discounted at a risk-free rate to reflect the risk borne by a plan’s sponsor for being obligated to make the promised benefit payments even if actual investment returns fall short of the plan’s expected returns. The present value of a plan’s liabilities increases (sometimes significantly) when they are discounted using a risk-free rate, which will also increase a plan’s normal cost.

⁶ The duration of newly accrued benefits is a weighted average of the timing of future benefits weighted by the present value of those payments (discounted by the pension-specific expected investment return) so it is basically a measure of the average time of future benefits.

⁷ The PPD also publishes plan payroll amounts so it is simple to convert the adjusted employer normal cost into dollars per active member as well.

⁸ In the Public Plans Database, the normal cost is coded as NormCostRate_tot, the employee’s normal cost is coded as NormCostRate_EE and the assumed investment return is coded as InvestmentReturnAssumption_GASB.

⁹ A plan’s complete sequence of expected liabilities for current workers is required to precisely calculate a plan’s duration of newly accrued liabilities. To our knowledge, no plans report this sequence. Some plans do report the sequence of expected liabilities for all workers, which can be used to estimate the duration of total liabilities. We calculated the duration of total liabilities for three plans that publish their complete expected sequence of benefits for workers and retirees: Arkansas Public Employees Retirement System, Teacher Retirement System of Texas and the Wisconsin Retirement System. The estimated durations were 10.4 (AR), 10.8 (TX) and 10.6 (WI). Adding 15 to those figures implies durations of newly accrued liabilities of 25.4, 25.8 and 25.6, respectively, which are very close to 25. As long as the age distribution of workers and retirees across pension plans is similar, then using a fixed duration value of 25 will not introduce bias into the generosity measure.

¹⁰ Some plans in the PPD have missing data for some years. The scatterplots are the averages for all plans with data in a given year.