The Revenue Demands of Public Employee Pension Promises

The recent financial and economic crisis and lingering low interest rates have heightened concerns about the ability of state and local governments to pay promised pension benefits to their current and future retired workers. By some estimates, the present value of pension promises may exceed pension system assets by $3 trillion.

While there are numerous studies of the extent of underfunding in public pensions, the past literature has generally not focused on the tax revenue implications of underfunding. This question is at the heart of a new paper by NBER researchers Robert Novy-Marx and Joshua Rauh, The Revenue Demands of Public Employee Pension Promises (NBER Working Paper 18489).

The primary data for the analysis comes from the Comprehensive Annual Financial Reports of nearly 200 state and local pension plans. The authors aggregate data for all plans within each state and combine this with tax revenue data from the U.S. Census Bureau and gross state product (GSP) data from the Bureau of Economic Analysis.

For their core calculations, the authors first forecast expected future benefit payments to retirees, current employees, and former employees who are vested in the plan but not yet receiving benefits. They next calculate new service costs, defined as any increase in the present value of expected future benefits from one year to the next (as plan participants progress in age and seniority) in excess of new contributions to the plan. Finally, they calculate the additional amount that state and localities would need to contribute to the plan annually for the next 30 years to completely amortize the unfunded pension liability.

The authors estimate that in aggregate, contributions from state and local governments to pay for public employee retirement benefits currently amount to 5.7 percent of all revenues generated by these entities (including state and local taxes, fees, and other charges). In their base case simulation, contributions would need to rise by 8.4 percent of own revenue, to a total of 14.1 percent of own revenue, to achieve fully funded systems in 30 years. This corresponds to an immediate and permanent increase of $1,385 per household per year.

Naturally, the required new revenue varies across states. While Indiana requires an increase of only $329 per household per year, the required increase in New York is $2,250; four other states also face required increases of at least $2,000 per household per year.

While benefits already accrued by current and former employees account for a sizeable share of the new revenue needs, all states (with the possible exception of Indiana) are increasing their unfunded liabilities each year, with newly accrued pension benefits exceeding new contributions. At least thirteen states would need to double contributions just to pay these newly incurred service costs.

Next, the authors explore how much the required contribution increases would be reduced under several policy changes that reduce future benefit accruals. One policy they examine is a "soft freeze" of pension plans, where new hires participate in a defined contribution (DC) pension plan rather than the traditional defined benefit (DB) plan. Utah and Alaska have instituted soft freezes, and this policy is also under consideration in Florida. They find that instituting a soft freeze in all states would have a moderate revenue-saving effect - the required average revenue increase declines from $1,385 to $1,210 per household.

One reason the decline is not greater is that affected employees would be eligible for Social Security and the authors assume that states would bear the full cost of the Social Security payroll tax in addition to contributing 10 percent of wages into the DC plan. Interestingly, for some states with high employee contribution rates and low Social Security coverage, instituting a soft freeze would be more costly than funding DB promises for new workers.

The authors also examine the effects of a "hard freeze," which would entail stopping all future benefit accruals in the DB plan, even for existing workers. Although this policy is not uncommon in the private sector, it has not yet been implemented by any public DB system. The authors estimate that if all plans were hard frozen, total required increases would average only 4.8 percent of own revenue, or $800 per household per year. This figure declines to 4.2 percent or $701 if Social Security costs for new workers could be shared equally between employers and employees.

Finally, the authors explore the sensitivity of the findings to some of their assumptions. If pension assets perform very poorly (at the 10th percentile of the asset return distribution), the required contribution rises to $2,468 per household per year; if assets perform well (90th percentile), the contribution drops to $756. They also estimate a version of the model in which taxpayers are allowed to move in response to rising tax rates - this raises the dispersion of required tax increases but does little to affect the average increase.

As the authors conclude, "one theme that emerges [from our work] is that substantial revenue increases or spending cuts are required to pay for pension promises to public employees even if pension promises are frozen at today's levels."

The authors acknowledge funding from the Zell Center for Risk Research at the Kellogg School of Management.