STUMP » Articles » Math Ain't Magic: Playing With Numbers Doesn't Make Pensions Cheaper » 15 February 2016, 19:15

Where Stu & MP spout off about everything.

Math Ain't Magic: Playing With Numbers Doesn't Make Pensions Cheaper  


15 February 2016, 19:15

Steve Malanga explains:

Scary Pension Math: Even After Bull Market, Governments Still Owe $1 Trillion

Wall Street has only recently ended the third-longest bull market in history. But in July, when stocks were still near their 2015 high, the Pew Charitable Trusts reported that state and local pension debt nationally had barely fallen since 2009, despite years of market gains.

Now, with markets down more than 12% since summer, new pension debt is piling up rapidly again, putting fresh pressure on budgets. Facing a bleaker outlook, some pension officials are admitting they’ve underestimated how hard it is to bounce back from volatile market declines like 2008.

Absent some startling development — like an even bigger runaway bull market, which few market watchers anticipate — the crisis will worsen.

At the root of the problem is a change in the financial structure of public pension funds. Whereas in the early days of government pensions governments expected more than half the money to pay benefits would come from taxpayer and worker contributions, pension systems increasingly came to rely on risky investments to pay retirees.

Since 1984, investment earnings have accounted for 62% of the money pensions need, according to a study by the National Association of State Retirement Administrators.

Another problem: When funds did well in the market, politicians often gave away the gains as new benefits, rather than banking the surpluses. Officials should have known better.

Even when fully funded, government pensions must earn on average 7.7% annually just to stop the accrual of new debt as today’s employs accumulate additional benefits. But few systems are fully funded, meaning the pressure on returns is even greater.
Admitting they’ve overshot their estimations, some pension funds are lowering their horizons. CalPERS staff have told the fund’s board they want to gradually reduce the system’s return rate, now 7.5%, to 6.5%.

But such reductions to projections do little to slash the $1 trillion or more in current debt. Instead, retirement systems will have to ask taxpayers to ante up even more to pay for lower investment earnings in the future.

The assumed rate of return is used to develop what contributions need to be made year-to-year to appropriately fund the pension and is one of the most important assumptions there is for determining liability value.

Of course, reality intrudes and you get whatever return you get… the cash flows will be whatever they are, and the farther off your original assumption was, the faster the shortfall grows.

You can put whatever you like as the assumed rate of return in a public pension valuation, supposedly. I have asked others in the past if there was any return so absurd you couldn’t use it for a valuation, and I got no definitive answer.

A few of the pensions in the Public Plans Database have assumed rates of return above 8%, still.

The highest rate I see in there currently is 8.5%…. for the Connecticut Teachers plan.


On top of the absurdly high expected return of the Connecticut Teachers plan, there’s a silly proposal to split the Connecticut Employees plan (a different plan) into an unfunded, pay-as-you-go part and a we’ll-eventually-fully-fund-it-isn’t-80%-good-enough part.

Here’s the set up:

General Electric Co.‘s planned move to Boston and warnings from bond-ratings firms have injected new urgency into discussions about the best way to overhaul Connecticut’s pension system.

The state has funded 39% of its obligations to provide retirements for the state’s 96,000 employees and retirees, according to the state comptroller’s office. To fully fund the system by 2032, as required under an agreement with state-employee unions, the state would need to more than double its $1.5 billion annual contribution in upcoming years.

Democratic Gov. Dannel Malloy wants to make the pension payments easier to handle by stretching them out over a longer period of time and by splitting the pension fund into two accounts. His proposal has raised concerns from the state comptroller and treasurer’s offices as well as bond-rating firms.
Both Mr. Malloy and Mr. Lembo want to change how the state pays down its pension obligations from backloading the payments—with the costs increasing in later years—to front-loading them.

Such a shift would mean higher payments in the near term rather the long term, a change SEBAC has long recommended and still supports. Both the Malloy administration and the comptroller’s office say the state’s annual payments also would be more predictable.

Clouding certainty on how much the annual pension payment could be under any overhaul scenario is how much the state receives on its return on investing the money.

The state’s pension agreement now assumes an 8% return. Boston College published a report commissioned by the Malloy administration that characterized the 8% rate as too optimistic. It said the return has averaged 5.5% over the past 15 years. If that rate continues, the administration estimates the state could pay as much as $6.65 billion in 2032 toward pensions.

But Mr. Lembo says that assumption was too pessimistic, and he calls for lowering the assumed rate of return to 7% from 8%. That means the state’s pension payment could rise to $3.8 billion in 2032—still more than the state can afford, he said.


Seriously, saying you can’t change the assumption set to something a little more in line with your actual experience and expectations, because it would be too expensive… means that in reality, the promises you made were too expensive.

Changing the valuation assumptions only makes things look more affordable. It doesn’t make it actually more affordable.

Here’s John Bury’s take:

And just like magic those liabilities are off the books. As with bond repayments on POBs they move into debt. As for the other participants.

“A separate fund would cover workers and retirees hired on or after July 1, 1984. Benefits for these workers would be almost entirely paid for using the state employee pension fund’s current assets of $10.6 billion which would continue being invested.”

This then becomes the plan that is valued and doing some quick math (in billions):
Total liabilities: $14.9 + $10.6 = $25.5
Funded ratio: $10.6 / $25.5 = 42%
Pre-1984 participant liabilities: .72 x $14.9 = $10.7
Post-1984 participant liabilities: $25.5 – $10.7 = $14.8
Official funded ratio after the split: $10.6 / $14.8 = 72%

Without changing any benefits Connecticut raises their magic number from a scary 42% to what other big companies and bond-buyers might be gulled into considering a manageable 72% with the only question being whether there will be a footnote in the pension part of the financial statements about the pay-go part. I don’t remember ever seeing one about New Jersey POBs and that passes for honest disclosure these days.

New Jersey and Illinois have both gotten dinged by the SEC in the past for misrepresenting their pension liabilities.

Just something that came to mind while reading Bury’s post.


Mark Glennon reports that an actuary used by many Illinois fire and police pension plans is possibly getting disciplined, but as I see no public notice yet, I’m not going to comment on that specifically.

That said, there have been previous stories on a few of the plans this actuary was used for, and I want to point a few of those stories. It involves assumptions other than the discount rate.

From 2012 and the Rockford Star:

Old data may be skewing Rockford police pension funding

ROCKFORD — Demographic assumptions about the city’s police force that date back to the Reagan administration may be understating the city’s debt to its police pension fund, an independent financial review of the fund has found.

The Police Pension Board hired Foster & Foster, an actuarial firm with offices in Fort Myers, Fla., and Oakbrook Terrace, to review the annual reports done by the city’s actuary, Tim Sharpe, who is based in Geneva. The review was prompted by discussions at a statewide training session about the methods used to calculate pension fund liabilities, said board President Jeff Nielsen.

Foster & Foster listed a number of suggested changes in its report to the board, several of which are outside the board’s power and would require a change in state law. But Jason Franken with Foster & Foster did note that Sharpe is using a mortality table from [b]1984[/b] when at least two new such tables have been published in the subsequent quarter century.

And here’s something from 2014:

Police and fire pension funds report $200,000 shortfall

Actuarial change leads to property tax boost to cover pensions

By Jean Lotus

Forest Park’s police and fire pension systems took a hit this year because a simple actuarial change recalculated how long safety personnel can be expected to live. Actuary Timothy W. Sharpe, of west suburban Geneva, changed one element of his calculations last year, revealing a $104,000 shortfall in the police pension fund and a $94,000 shortfall in the fire pension fund.

That money was added to the village’s tax appropriation levy in July, boosting property taxes in town by about $200,000.

The change came when Sharpe switched last year from a 1971 mortality table to a new table that more accurately reflected the lifespans of police officers and firefighters living in 2000.

For more than a decade, Sharpe had been using a group annuity mortality table called the GAM-1971. As its name implies the table was created in 1971 using mortality data from police officers and firefighters collected between 1964 and 1968. Life expectancies on the tables tracked public safety workers who, at age 50, would have been born between 1914 and 1918.

According to the updated table, called the RP-2000, male police and fire personnel at age 50 in 2000 could be predicted to live an average of 4 years longer than they did on the 1971 table. Statistically, most firefighters and police officers are male. Longer-living employees mean more money needs to be socked away in pension plans to cover their retirement.

Sharpe was criticized for using the 1971 tables in his calculations.

Jim Palermo, a village trustee for LaGrange filed a complaint in 2012 against Sharpe with the Actuarial Board of Counselling and Discipline (ABCD) in Washington, D.C. for using out-of-date tables.

And finally, something from Tia Goss Sawhney, who did a guest post at Wirepoints about some of these assumption sets. An excerpt of Tia’s post at Wirepoints:

The WirePoints article observes that, with respect to Forest Park, in spite of “pretty good returns” the “dollar total of its pensions’ unfunded liability has increased in each of the last two years.” Unfunded liability increases can occur under four circumstances:

1) the total contributions made to the plan are less than the plans’ current year costs (actuaries call this amount the “normal cost”) and the interest on the unfunded liability (where the interest rate is the assumed asset return rate),

2) the assumed asset return rate is not met,

3) other assumptions (mortality, retirements, etc.) are not met, and/or

4) the actuarial methods or assumptions change.

There is reason to believe that Illinois pension assumptions are biased. Let’s start with the Forest Park Police Plan. In order to calculate the tax levy (“bill”) for fiscal year 2013, Forest Park’s actuary assumed a 7.5% asset return and a mortality rate equal to the 1971 Group Mortality Table with no allowance for the mortality improvements observed since 1971 or the improvements likely for the future. Furthermore, he made no apparent assumptions for service related deaths or disabilities, even though there are distinct benefits associated with service related deaths and disabilities [5],6. In evaluating the 7.5% assumed asset return, consider that, by law, Forest Park can only invest 55% in non-cash and bond investments [7].

You can follow the links for more details than I want to cover right now.

Just to nerd out for a moment: the higher the discount rate, the less the other assumptions matter. If you’re discounting at 8%, having an expected retirement of 25 years as opposed to 20 years doesn’t make as much of a difference.

I could do some life expectancy comparisons of 1971 GAM and RP-2000, but meh.

Let me just repost my recent death curves:

Note that those death curves represent the general population, which includes disabled people, etc. You can see that the “death peak” has been moving to older and older ages, adding on a few extra years of expected retirement over the decades, at about a year per decade (very roughly).

And overall mortality improvement hasn’t been an even trend. In a recent NY Times article, it was noted that life expectancy improved for higher-income people more than low-income people (and in some cases those lower-income people’s life expectancies lowered… but more on that another time.)

Many of those public employees are in the higher income brackets. Actually, the people in the lowest income brackets tend not to be employed at all.

But back to the matter at hand: using older mortality tables, for pensions, made the liabilities look cheaper.

But reality is that people aren’t dropping dead as often in their 60s as they did 40 years ago.

Every year that more retirees live longer, that experience is translated into actual pension cash flows out of the assets and to the retirees.

Reality always catches up.


Even if a small plan had an excellent mortality assumption, modest rate of return assumption, etc., it can still run into trouble just due to natural variability.

Think about life insurance for a moment. As an individual, you are exposed to a relatively large amount of uncertainty with respect to when you’re going to die. I looked at those death distributions and find the “central” 50% of probability of death. I basically went from the 25th to 75th percentile ages for death (percentiles determined from birth). These are only approximate, because I didn’t want to interpolate anything.

Cohort Lower limit (closest to 25th %ile) Upper limit (closest to 75th %ile) Interval width (years)
1920 37 76 39
1930 48 77 29
1940 55 79 24
1950 60 81 21
1960 62 82 20

Even for the youngest Boomers, there’s a pretty wide window just to get 50% confidence that you’ll live to a specific amount. And you only die once — it makes it hard to plan as an individual exactly when you’re going to die.

On the other hand, the insurance company handling multiple-thousands of policies will have much less uncertainty about their results. They’ve got a pretty good chance that their overall distribution of deaths will fall a certain way.

But the fewer policies there are, the more variability in their results. This is partly why smaller insurers need reinsurance more than the largest insurers.

Similarly, if a pension plan has only a few dozen participants, even with up-to-date assumption sets, reality is going to diverge quite a bit from the average.

At least, chances are good that will happen.


But back from my actuarial geeking: putting a thumb on the scale, with the assumptions all in an optimistic direction, “helps” the plan only in the short-term.

For a while, especially when you have many more active workers than retirees, the plan will look relatively cheap. Some of the investment experience will start to filter through before retirement occurs, as well as changes such as salary experience, but that won’t necessarily have a large effect on the funds.

Until people start retiring. And living a lot longer in retirement than originally expected. Money keeps going out the door, and one finds that one has to liquidate assets in order to meet pension cash flow needs.

Even if one keeps the optimistic assumptions, cash goes out of the fund, and the unfunded liability climbs and/or the required annual contribution starts to climb rapidly. This is not unique to Illinois plans – I’ve seen it at multiple plans in the public pensions database. The more the assumptions diverge from reality, the faster this demand on more cash will climb.

The point is that the assumption sets are there to help pensions to plan their contributions, so that they’re stable. But if the assumptions all diverge from reality in the same direction: the direction of lowering the liability value on paper compared to what more realistic assumptions would, then one ends up with an unsustainable funding path. That’s what Connecticut is seeing. That’s what many Illinois plans are seeing.

Too “optimistic” ends up with something very pessimistic indeed, as reality is revealed.

And it always will get revealed… eventually. The question is whether various decision-makers are dead before those revelations occur.

Compilation of Connecticut posts

Related Posts
South Carolina Pensions: Liability Trends
A Proposal for Public Pension Reform: The Idea and Coming Attractions
Kentucky Pension Liabilities: Trends in ERS, County, and Teachers Plans