STUMP » Articles » Pew Pensions Report: Public Pensions Should Stress Test, also -- How the States Were Chosen » 31 May 2018, 12:16

Where Stu & MP spout off about everything.

Pew Pensions Report: Public Pensions Should Stress Test, also -- How the States Were Chosen  


31 May 2018, 12:16

In yesterday’s post, I looked at the recently-released report: Assessing the Risk of Fiscal Distress for Public Pensions: State Stress Test Analysis.

While the full report is 147 pages, you’ve got to realize that the primary text ends on numbered page 59, and the rest are appendices with technical detail and results. Also, there are some fairly generous margins, etc., in this paper and it reads rather rapidly. This isn’t the normal, deliberately opaquely-written and dense academic style of paper.

Someone wrote me about this study, and asked about California specifically… well, California isn’t covered in this paper, and let’s go see why.


To assess the sustainability of pension plans, we included 10 states of varying fiscal health in our analysis:
Colorado, Connecticut, Kentucky, New Jersey, North Carolina, Ohio, Pennsylvania, South Carolina, Virginia, and Wisconsin.3

[3] The analysis examines 18 pension plans across the 10 states and is limited to state worker and teacher plans in every case. For Connecticut, Ohio, Pennsylvania, and Wisconsin, the plans we cover represent virtually all of what is managed and reported at the state level. A full list of the plans is noted in the methodology section and appended to this analysis.

I’m not going to dig into that list – that’s not really my point.

Some of the states I’d be really interested in hearing about are: California, Illinois, and New York

I’m going to be very simplistic in making a ranking table of state pensions — I decided to include all state-level pensions from the Public Plans Database (which include plans like Calpers and IMRF, where the pensions are for employees of localities, but the funds are governed on a statewide basis in one fund. I also include safety officer pensions, which were sometimes excluded in the Pew study)

Going by GASB-reported liabilities for fiscal year 2016, this is the ranking I achieved:

The highlighted starts are covered in the study. Interestingly, the 4 largest states in terms of pension liabilities were not studied… why not?

Starting on page 23, they explain the choice of ten:

The states we chose to evaluate — Colorado, Connecticut, Kentucky, New Jersey, North Carolina, Ohio, Pennsylvania, South Carolina, Virginia, and Wisconsin — represent a range of fiscal positions and policies. The differences among these states allow us to use the stress test simulation model to demonstrate how market volatility and economic uncertainty can affect pension costs depending on a state’s current financial status and contribution rules.

Here are the three core research questions they used to select those states [page 13]

RQ 1: How do we assess fiscal distress and which states are at risk?

RQ 2: How might lower investment returns affect pension costs and therefore state budgets in the long term?

RQ 3: What is the impact of economic volatility on pension fiscal health, given states’ high exposure to stocks and other risky asset classes?

So rather than quote their full reasoning for selection, here’s my summarization:

  • Kentucky and New Jersey: super-low funded ratios, years of underfunding – most likely to run out of $$ first [RQ 1]
  • Pennsylvania and Connecticut: also low funded ratios, have greatly increased costs – looking at how much higher those costs may climb [RQ 2]
  • Colorado and Ohio: Contribution levels set as fixed percentage of payroll, unlikely to respond to changes in economic circumstances, checking volatility
  • Virginia and South Carolina: Fairly “traditional” practices for funding, which smooths out results – SC is poorly-funded, Virginia in the middle-of-the-pack. [RQ 3]
  • North Carolina and Wisconsin: relatively well-funded — NC has strong funding policies, WI has risk-sharing — how will they do under stress scenarios?

They do note that many of these states had recently made changes to the pensions, and they did factor that in to their projections.


To wit: they’re screwed.

Money runs out, plans become pay-as-they go.

Their New Jersey projection:

That looks a lot like this projection by me:

I have a few others of other NJ funds also showing such a runout.

My projections are only one plan at a time, not combined by state… and anyway, I think it’s nice that my projections accord qualitatively with theirs.

Here’s their Kentucky projection:

Now, that doesn’t quite accord with mine, but that’s because I have different asset run-out projections for the various Kentucky plans.


Key findings: Money doesn’t run out, but costs run realllly high. Which was already their baseline.

I really don’t want to hear about how DB plans are cheaper than DC plans. I think I could do really well if I had such high contributions to my 401(k).

Note: that’s just the baseline! They’re not even stressing anything there!

Here’s one of the stress scenarios: 5% level returns.

So yes, the costs can bump up even higher.


Key findings: unresponsive funding policy leads to decreasing fundedness in stress scenarios

No nifty graphs for this one (I mean, there are graphs, but they’re boring), so I’ll just directly quote:

Both Colorado and Ohio have taken measures to address eroding fiscal health; however, neither
state has built-in policies to adjust to market downturns or react to changing economic
circumstances. If investments fall short in the future, both states could once again face potential
insolvency. As demonstrated below with Virginia and South Carolina, adequate actuarial funding
policies can be instrumental in maintaining funding in uncertain and volatile investment climates.

Basically, they’re in precusor conditions to other states in poor positions re: their pensions. The point is that they don’t make up for bad times.


Key findings: can end up with very volatile funding requirements, due to their funded ratios and funded policies

These two states do diverge in what happens, because Virginia has an about-average-funded ratio of 72% and South Carolina is considerably worse at 52%. And it’s not only that: South Carolina has a floor on contributions, and Virginia does not. When they look at stochastic scenarios, here are the ranges they get:

Even though South Carolina is at a poor funded ratio level, they show less variability in where they’d end up for contributions.

Unlike Virginia, which sets contributions based on the latest actuarial
valuation and allows for required payments to either increase or decrease to offset return
volatility, South Carolina sets a contribution floor, but otherwise keeps the contribution rate
relatively stable. Specifically, the state’s funding policy does not automatically adjust
contributions downward to account for investment gains unless the plan is fully funded; nor does
it adjust them upward to offset investment losses, unless the new funding level falls below the
contribution floor. In other words, contribution rates and total contribution amounts are
designed to be less volatile.

A key finding of the stochastic analysis is that well-designed funding policy can mitigate much of
the cost volatility caused by the amount or timing of investment shortfalls. We conclude that
states can, in fact, solve for a fair amount of uncertainty with well-drafted contribution policies.
Like South Carolina, states can construct funding policies that explicitly maintain adequate
funding levels during good economic times to create a cushion, or surplus, to protect against
contribution shortfalls during economic downturns.

Imagine that – not taking a cushion away when times are good…. Calpers, are you listening?


I ran this yesterday, but again, we see less volatility for Wisconsin than for North Carolina.

So yay for Wisconsin.


Straight from the paper itself:

We find three key benefits to stress testing, generally:

  • First, stress testing can aid administrators and policymakers in planning for the next
    downturn, as well as protect state plans against the worst possible outcome: insolvency.

  • Second, stress testing provides a fuller understanding of the impacts caused by market
    fluctuations and can inform good funding policies and practices to better manage
    volatility and ultimately lower costs.

  • Finally, stress testing can provide a useful tool for considering a range of possible
    economic scenarios when scoring proposed reforms.

….. Stress testing should be a standard reporting practice for all public retirement systems.

That emphasis is in the original.

Two of the stress tests they have are entirely deterministic [level 5% returns, asset shock], and are similar to cash flow stress testing U.S. life insurers are required to perform.

Doing that sort of deterministic scenario testing is a good place to start, even if stochastic testing is beyond the capabilities of the plans right now. If you can’t run two alternative assumption sets for cash flows….

… there will be problems.