A recently updated Brookings Institution report on public pensions suggests that underfunding is not a serious threat to state and local finances and “there is no imminent ‘crisis’, in the sense that [public pension] plans are likely to exhaust their assets within the next two decades.” The study’s authors find that public pension benefit payments as a percentage of gross domestic product are unlikely to rise much further and suggest pension liabilities can be stabilized with relatively minor fiscal adjustments that do not need to be implemented in the near term. However, the report relies on a small sample and makes strong assumptions that should be examined.
The authors — Jamie Lenney, Byron Lutz, Finn Schüle, and Louise Sheiner — assert that full pension funding is not essential for government fiscal sustainability, citing a 1958 Paul Samuelson paper explaining how a “pay-as-you-go” (PAYGO) retirement system can be sustained across generations. Although the authors of the Brookings report recognize that some degree of prefunding can stabilize future employer contributions, their modeling shows that most plans already have sufficient assets to avoid exhaustion in the near to intermediate-term even under conservative asset return assumptions.
Under certain conditions, PAYGO pension systems can function for an extended period as shown by the experience of Portland, Oregon’s Fire and Police Disability and Retirement Fund (FPDR). As I discussed last year, FPDR has been paying benefits for over 70 years without prefunding.
But past experience does not guarantee the security of future Portland public safety retirees (or taxpayers), let alone the financial sustainability of other state and local governments that experiment with PAYGO.
When the Brookings report was presented at a conference this spring, discussants offered technical criticisms of the authors’ modeling approach. For example, Stanford University’s Joshua Rauh argued that the model suffers by failing to consider the volatility of future asset returns. MIT’s Deborah Lucas suggested that the report’s future risk-free rate assumption is lower than levels implied by financial market activity.
Drilling down into the Brookings study’s data and modeling raises further concerns as described in the following sections.
The Potential Problems with Using a Small, Non-Random Sample
First, the power of the Brookings study’s findings is limited by sample size and composition. The authors analyzed 40 plans in 22 states, plus the District of Columbia. The sampling strategy is non-probabilistic, meaning that public pension plans were not chosen randomly, but instead based on data availability, plan size and funded ratio. It is thus impossible to know how well the chosen sample represents the nation’s 6,000 state and municipal government pension plans, nor can the analysis yield any reliable confidence intervals or margins of error.
While the sample includes six California plans, it excludes the California Public Employees’ Retirement System, CalPERS—the nation’s largest public pension system. Other major pension systems excluded from the sample, include the Ohio Public Employees Retirement System, the Wisconsin Retirement System, and the Texas Employees Retirement System.
Similarly, although relatively small municipal plans (with less than $2 billion of assets and accrued liabilities) in Baton Rouge and Kansas City are included in the analysis, none of Chicago’s much larger and deeply underfunded public pension plans are there. Nor are pension plans from the large population states of North Carolina, Virginia and Washington.
The authors say that the sample “matches the national distribution of plans in terms of both mean and variance for multiple plan characteristics – e.g. the funding ratio.” While this may technically appear to be the case, the small and geographically unrepresentative sample calls into question some of the report’s broader conclusions.
For example, the Brookings study concludes that “many of the most poorly funded plans have in recent years undertaken the largest reforms and increased contribution rates the most; in so doing, many of these poorly funded plans have already made significant progress toward stabilizing their pension debt.”
But since there aren’t that many plans in the overall sample, let alone in the sub-sample of “poorly funded” plans, the data would not appear sufficient to support this finding. In Chicago’s case, modest new hire reforms and contribution increases during the last decade were insufficient to halt declining funding levels for the city’s public safety plans (which were 21 percent and 18 percent for the police and fire systems respectively as of 2019). And, just recently, the Illinois state legislature increased cost-of-living adjustments for some retired Chicago firefighters, thereby reducing the fire plan’s funded ratio.
The sample used in the study seems inadequate to support the author’s baseline projection that nationwide state and local pension benefits as a percentage of GDP will peak at around 1.7 percent of GDP in the 2030s before falling back to 1.4 percent by 2117.
Pension Benefit to GDP Ratio Makes Questionable Assumptions About Economic Growth
These projections also must rely on strong assumptions about economic growth and mortality. If mortality does not improve gradually in line with the Society of Actuary rates the Brookings authors use, and instead life expectancy rises sharply because of new medical breakthroughs, public pension benefit payments could be much higher than expected.
Also, if the sharp decline in 2021 birth rates as projected by Brookings affiliated scholars Melissa S. Kearney and Phillip B. Levine is not reversed, the smaller labor force could result in lower-than-expected growth in long-term GDP.
Given the retirement of the baby boom generations, there is a very strong expectation that pension benefits to GDP should be rising. Instead, they have flattened out, very likely reflecting the effect of pension reforms.
A better approach to modeling would involve simulating not only asset returns but also macroeconomic and demographic values. Such an analysis would yield a distribution of pension cost and asset outcomes over an extended time period.
In a 2013 study, I used a multivariate simulation to determine that pension costs were unlikely to trigger a bond default by the state of Illinois over a 30-year time horizon. But unlikely and impossible are not the same thing: a few simulation paths led to default.
It is more instructive to conceive of outcomes such as asset exhaustion or government insolvency in probabilistic terms. Public pension reform measures can then be assessed by the degree to which they decrease the likelihood of extremely adverse outcomes, and by the amount of pension cost crowd-out that occurs under less extreme scenarios.
Is PAYGO or Pension Underfunding Sustainable?
While Sheiner and her coauthors may have stretched limited data and insufficient quantitative analysis to reach their conclusions, their observations nonetheless contain an element of truth. Under the right economic and demographic conditions, a pension system with little or no funding can operate for an extended period without causing financial distress for the sponsoring state or local government.
But a proper simulation analysis—which would include scenarios of economic contraction, negative asset returns and lower-than-anticipated mortality—would likely show that underfunding poses a risk for many systems.
Pension reforms that eliminate unfunded liabilities minimize insolvency risk. In the 2010s, Detroit and Puerto Rico became unable to service their pension liabilities due to population losses and the resulting declines in tax revenues.
Today, many states and cities are experiencing small population losses. If these losses accelerate, already affected governments may also lack the financial resources to cover the cost of public pension benefits earned in the past. Instead, governments should avoid benefit structures that create unfunded pension liabilities in the first place and implement pension reforms to pay down liabilities they have already accrued.
Finally, irrespective of whether unfunded liabilities are fiscally sustainable, there is also the question of whether they are appropriate. Pension benefits are earned at the time services are rendered. By not paying for the benefits as they are earned, political leaders transfer public pension costs onto future taxpayers who do not yet have a vote. From the perspective of intergenerational equity, funded ratios should always be 100 percent or higher based on conservative actuarial assumptions.
Dr. Louise Sheiner graciously emailed the following response to Marc Joffe:
In terms of your discussion of our paper, there are obviously a lot of points I disagree with, but I won’t go into a point by point rebuttal. I would like to note the following basic points:
- We do think our sample is broadly representative, but of course we can only show it is representative on the dimensions we can examine. Our modeling of each plan is very detailed which has clear upsides in terms of the analysis permitted. But it precludes modeling every plan. Moreover, we think the paper is extremely clear that there is immense heterogeneity across plans and that the statements we make for pensions in the U.S. overall do not apply for all plans.
- The fact that many plans have made changes that lower future obligations has been documented for a wider sample of plans elsewhere. E.g. https://crr.bc.edu/briefs/cola-cuts-in-statelocal-pensions/ and https://www.nber.org/brd/cost-living-adjustments-public-sector-retirement-plans. Moreover, the effect of these reforms is now visible in the official national, aggregate statistics. Given the retirement of the baby boom generations, there is a very strong expectation that pension benefits to GDP should be rising. Instead they have flattened out, very likely reflecting the effect of pension reforms.
- We note that, at risk-free discount rates, plans have long been underfunded so today’s taxpayers inherited this debt, they didn’t create it. Thus, the notion that plans should be funded from an intergenerational equity perspective doesn’t really apply—getting to full funding means that current generations pay for more than the cost of the services they receive—they have to pay their own costs plus the costs left to them, while future generations are absolved of the need to bear this burden. We think it is reasonable to instead spread the costs of the legacy debt across generations.
- Evaluating sustainability at the risk-free rate is a reasonable and widely accepted way of accounting for risk, because rates are expected to be much higher than the risk-free rate. Furthermore, in our final version of the paper, we do a stochastic simulation which shows that in virtually all cases, plans will be able to meet their obligations if they contribute as much as is necessary using the risk-free rate of return. No amount of funding (even 100%) can guarantee that plans will be able to meet obligations in all states of the world.
- In any case, the basic point of our paper is to point out that unfunded is not the same as unsustainable, and so a failure to be fully funded doesn’t imply that a plan will not be able to meet its obligations.
Stay in Touch with Our Pension Experts
Reason Foundation’s Pension Integrity Project has helped policymakers in states like Arizona, Colorado, Michigan, and Montana implement substantive pension reforms. Our monthly newsletter highlights the latest actuarial analysis and policy insights from our team.