Lessons and Myths of West Virginia's Pension Reform

When a state or local government's pension system gets in trouble, an oft-proposed solution is to close a defined-benefit (DB) system and open defined-contribution (DC) accounts. This eliminates the governing jurisdiction’s liability for future benefits while still allowing them to provide pensions to public employees who’ve been accruing promised benefits. You might have read such proposals on this blog from time to time.

But what happens when a government makes the switch to a defined contribution system and things still don't get better? This question is regularly posed when current systems are looking at reform. And it is one worth answering.

Three frequently cited states that have made the transition from a DB to DC system but have also seen their pension finances continue to decline in the years following reform are Michigan, Alaska, and West Virginia. Reason published a case study about Michigan earlier this year that addresses the looming question in context of the Wolverine state. And we have a forthcoming study looking at what has happened in Alaska (due out this summer).

For now we can look briefly at some arguments about West Virginia to get a sense of why most of the arguments about cost increases because of a DB to DC transition are generally cost increases in spite of the reform. Further, those that argue against transitioning from a DB system to a DC system because they believe that the changes have made a state worse off frequently betray a misunderstanding about how pension financing works, and are otherwise providing normative commentary.

A commentary by Diane Oakley, executive director of the National Institute on Retirement Security (a lobbying organization for pension plans) illustrates these points. The January piece for the website Pensions & Investments aimed at demonstrating the "flaws of adopting cost cutting in switching to DC plans.”

Oakley notes that in the 1990s, the West Virginia Teachers' Retirement System (WVTRS) was facing a large unfunded liability—state lawmakers were using bad actuarial assumptions in funding their plan and the irresponsible funding behavior had created a large gap between promised pension benefits and assets available to pay them. The state decided to close the DB fund to new teachers and put future teachers in a 401(k) styled DC system.

What's happened since then? There are at least three myths being perpetuated:

Myth #1: Pension Reform Caused the WVTRS Unfunded Liability to Get Worse

After laying out the facts about where reform came from Oakley argues, "More than a decade later, both the DB plan and the new DC plan faced challenges. The TRS DB plan was less than 20% funded, while teachers with DC accounts found their balances inadequate."

There are two things here. The first of these claims is a technical one: a decade after reform, the state still had large unfunded liabilities for its DB plan, despite reform. Seems simple enough, but Oakley's framing is taking advantage of "anchoring bias" on that 20% number (and it borders on deceptive). A system that is 20% funded is undeniably bad, but to know whether it was caused by the pension reform we would at least have to know how well the WVTRS was funded prior to reform.

We took a look. Publicly available data for 1990s is hard to come by, though at least one article, from the NEA, suggests the WVTRS was only 14% funded in 1990 prior to reform (the data isn't cited). In 1994, WVTRS was 11.6% funded, according to the earliest public data from West Virginia's comprehensive annual financial reports.

In 2000, a decade after reform, WVTRS was 21.4% funded. It isn't clear what year Oakley is referring to when she says "a decade later" the system was "less than 20% funded" but in 2003, the DB plan of WVTRS was 19.1% funded. Still, by 2005, the DB system was up to 24.6% funded.

Clearly, these are all terrible funding ratios. But Oakley's sentence inferred that the low funding ratio was because of the reform, when in fact it was already incredibly low. She inferred that WVTRS was getting worse under reform, when in fact it was improving, albeit very slowly.

The second claim is more normative: teachers' accounts were "inadequate." Such a claim requires making a decision about what a retirement benefit should be, what lifestyle a person should live in retirement, and whether it is appropriate to use taxpayer money to provide public sector workers promised retirement benefits that aren't widely available in the private sector. Further, it makes an assumption that 10 to 15 years' worth of savings in a 401(k) should be considered enough for measuring adequate levels of retirement. And it ignores that employee contributions to their DC retirement were lower than employee contributions to the DB system, almost certainly contributed to any disparity. But since this is mostly a normative debate (and one that I definitely have opinions on, but recognize are opinions that choose between competing goods in society) we will set it aside for another time. But I will suggest that there are ways a DC fund can be managed that yield results similar to DB systems. Improperly managed DC systems are not evidence of the failure of defined-contribution accounts. Just evidence that there are good and bad ways to run them.

Myth #2: WVTRS Got Worse Because of Demographic Shifts

Moving on, Oakley notes that: "While teachers made their required contributions to the TRS DB plan out of every paycheck, until 1991 state policymakers operated the system on an expensive a pay-as-you-go model that built up a significant unfunded liability. West Virginia adopted an actuarially based plan to reach full funding for the liability in the closed pension plan in 1994. But with the plan closed, demographics shifted quickly. By 2005, TRS paid pension benefits to nearly two retired teachers for every active teacher still contributing to TRS. When combined with funding percentage levels in the low 20s, this was a major concern."

This betrays a serious misunderstanding about pension funding (and one that I find confusing given Ms. Oakley's position). She correctly notes that operating a pension on a pay-as-you-go model is the wrong way to fund a pension system. Pension systems are supposed to be fully prefunded every year. That is, every year the government is managing a pension fund, it should be paying enough to cover all of the benefits earned that same year. 

But the demographics shouldn't matter in the way that Oakley suggests. There are two components to annual pension funding, the annual cost to prefund pension liabilities, known as “normal cost”, and the cost to pay off unfunded pension debt. Think about it mathematically: Normal Cost + Debt Payment = Annual Required Contribution (ARC) to prefund accrued pension benefits.

The Normal Cost is supposed to be prefunded every year. It is a combination of employee contributions towards their accrued benefits and the employer (government) contributions to cover the rest. So if West Virginia changed its actuarial practices to close out its unfunded liability presumably they were paying their full Normal Cost.

The transition to a DC system didn't change the Normal Cost payment for the state's DB system. And in fact, because the DB system was winding down, with fewer employees in it, Normal Cost should have started falling over time.

Additionally, in nearly every pension system around the country debt payments are not subsidized by employee contributions. Normal Cost funds the accrued benefits in a given year, and employees contribute towards that. Then on top of the government's contribution to Normal Cost they make an annual payment towards the pension debt (unfunded liabilities). Each year the state actuary would amortize the unfunded liability over the course of a certain number of years to determine the annual debt payment. But the debt payment would be the same no matter how many employees were in the system.

If West Virginia adopted a policy that used employee contributions to pay down debt it would be a policy choice separate and distinct from creating a defined-contribution system for new employees.

Demographics do matter in getting the Normal Cost projection right, though. Normal Cost is determined based on assumptions about how long retirees will live, salary growth, inflation, and what return on investments the pension fund's assets will get. If these numbers are off, then Normal Cost will be too low and additional unfunded liabilities will accrue.

It is critical to recognize that a state or city or county can close a DB system to new hires, create 401(k)-styled accounts for new employees, but still see unfunded liabilities grow if its actuarial assumptions are wrong.

And that is what happened to West Virginia. 

Myth #3: Closing the DC System Created Savings

Oakley writes that in 2005 West Virginia "closed the 401(k) plan and reopened the pension plan to new teachers. This generated an immediate savings for the state because the “normal cost” for TRS was roughly half of the required employer contribution to the 401(k) plan."

I do not question her numbers, that the Normal Cost for the DB system was half of the Normal Cost for the 401(k). However, this situation was primarily because the Normal Cost for the DB system was too low, which in turn was because the actuarial assumptions were wrong. The low Normal Cost projections directly contributed to the unfunded liability. The "savings" were not savings at all— they were achieved by perpetually underfunding more people's pensions.

Moreover, lawmakers could have just lowered the defined-contribution rate if they determined the cost of that system was too much. The rate that a government chooses to contribute to its DC accounts is a policy choice separate from the procedure of switching from a DB to DC plan. Any costs of the DC system itself would have been in spite of the reform, not because of the reform.

Have West Virginia's pension finances for teachers gotten better since ending their reform effort? Absolutely. Was this because shutting down the DC system created cost savings? Absolutely not. And to suggest otherwise is disingenuous. Just look at  the funding ratio of WVTRS and actual contributions to the ARC of WVTRS over the years after reform kicked in and after reform ended. The funding ratio improved slowly over time, though is still troubled by bad actuarial assumptions. But the drastic improvement in the funding ratio after reform ended wasn't because the DC accounts were ended and the DB system restored, it was because of the sharp increase in the percentage of Annual Required Contribution (see highlighted years).

Why the spike in ARC actual contribution percentage? We can actually turn to Oakley for the reason as she herself writes: "West Virginia demonstrated its renewed commitment to catch up on past pension funding payments by using $807 million from its tobacco settlement fund to shore up the TRS plan."

That is something the state could have done while maintaining the DC system. It was a good governance policy choice completely separate from whether the DC accounts were putting away enough for retirees. It was a policy choice that would have made the DB system stronger as it phased out.

Conclusion

West Virginia's experiment poses a lesson for other states (or local governments) considering pension reform. But the lesson is not that DB to DC transitions cost more. Inherently, such transitions move governments towards eliminating the concept of unfunded liabilities. Whatever liabilities remain in a defined-benefit system during reform exist in spite of reform, not because of reform. Instead, the lesson is that regardless of the type of system a government may choose, the transition still requires that the government use good actuarial assumptions or they will be underfunding their pension system in the midst of reform.

Research assistance for this blog post was provided by Truong Bui and Zachary Christensen.