Confronting Myths About Closing Defined Benefit Plans
Photo 39302665 © Andrew Oxley -


Confronting Myths About Closing Defined Benefit Plans

Evidence from Pension Reform in Michigan and Alaska

Can closing a defined benefit pension plan directly lead to the creation of unfunded liabilities? The question is an important one as policymakers consider the various options on the table for responding to the public sector pension crisis facing most states in America. Unfortunately, there are several predominant myths about the closing of defined benefit plans that are often repeated when this question is considered.

In a recent policy study, Truong Bui and I analyzed the actuarial experience of Michigan and Alaska since their respective pension reform efforts that closed defined benefit plans and replaced them with defined contribution plans for new hires. For each state, we forecast several counterfactual scenarios to determine what the level of unfunded liabilities in their pension plans would have been if pension reform had not been adopted.

Across the board we find that closing defined benefit pension plans has made states more solvent than if they hadn’t adopted their respective reform efforts:

  • Michigan is more than $7 billion better off for closing the defined benefit tier of its general employee plan than without pension reform; and
  • Alaska is more than $4 billion better off for closing its two defined benefit plans because of reform.[1]

Our findings support the arguments others have made about the pension reform efforts; in particular that closing defined benefit plans has helped improve Alaska’s credit rating and saved Michigan taxpayers money. However, our paper also stands in contrast to analyses of Michigan and Alaska’s pension reform efforts that claim that closing the defined benefit plans has caused debt to rise in both states. These latter claims should be directly addressed and dismissed.

Myth: Unfunded liabilities increased in Michigan and Alaska after pension reform was adopted, therefore pension reform has been a failure.

Truth: There are many factors that have caused the growth of unfunded liabilities, but pension reform is not one of them; correlation of reform with debt growth does not necessarily mean reform caused debt growth.

Since Michigan and Alaska closed defined benefit pension plans, both states have seen unfunded liabilities increase. Critics of pension reforms that close defined benefit plans, and offer new hires defined contribution plans, often argue that pension reform has caused this increase in unfunded liabilities. In a March 2015 op-ed opposing pension reform in Nevada, economist Teresa Ghilarducci claimed:

States that have experimented with a transition to 401(k)-style systems have exacerbated rather than solved their debt problems. Michigan started enrolling all new state employees in a 401(k)-type plan in 1997. The system’s unfunded liabilities increased from $697 million in 1997 to $4.1 billion 13 years later.[2]

This claim is both overly simplistic and misleading. The inference is that because new employees were enrolled in a defined contribution plan, while the outstanding defined benefit plan was closed, that the unfunded liabilities increased. Or at least, this is inferring that creating a defined contribution plan didn’t prevent the growth of pension debt. However, simplistic correlation analysis doesn’t allow for the possibility that pension debt would have been even higher without the reform, and that the growth of unfunded liabilities could be structurally unrelated to closing a defined benefit plan.

As we discuss in detail in our recent policy study on the effects of pension reform on plan solvency-as well as in an updated case study of Michigan’s pension reform for the State Employees Retirement System (MSERS)-the proximate causes of an increase in unfunded liabilities in Michigan over the past two decades have been:

  1. The systematic underfunding of MSERS’s required contributions by about $570 million since 1997. Michigan simply stopped paying 100% of the actuarially determined contributions necessary to properly fund the defined benefit plan, which was a policy decision inherently separate from putting new employees in to a defined contribution plan.
  2. The underperformance of plan assets, with an average return of 7.1% between 1997 and 2014 instead of the 8% assumed rate of return. This poor investment return was not a result of the MSERS pension reform. Investment strategy did not change to fulfill any requirements relative to the reform, and the actual rate of return would have been less than the amount expected with or without closing the defined benefit plan.
  3. The use of a discount rate that consistently undervalued MSERS accrued liabilities. We estimate that normal cost rates have been at least $2 billion less than necessary to pay future benefits simply because the plan is undervaluing accrued liabilities.[3]
  4. The misuse of pension assets to guarantee private sector investments. The state pledged pension assets as a backstop for an $18 billion bond issuance to Michigan Motion Picture Studios. The business ultimately failed to meet revenue projections, missed a bond payment, and triggered a multi million payout to bond holders from pension fund assets.
  5. Retrospective benefit changes. An early retirement incentive program that offered an increase in benefits to leave the plan but without funding the actuarial costs of the change to MSERS’s obligations.

Just because unfunded liabilities increased between 1997 and 2010 doesn’t mean it was the closing of the defined benefit tier of the plan that caused the increase.

Myth: Closing defined benefit plans inherently increases unfunded liabilities because of the declining number of active members in the plan.

Truth: For employees in Michigan and Alaska, defined benefit plan contributions are intended to pay only for normal cost, with employer contributions covering all unfunded liabilities. Defined benefit pensions differ from Social Security where the cash flow of active members is needed to pay for retiree benefits (i.e. pay-as-you-go). With defined benefit pension plans employee contributions cover a portion of their own benefits and are always refundable (i.e. pre-funded).

There is a technical claim that closing a defined benefit plan inherently leads to increases in unfunded liabilities because it reduces contributions into the plan, specifically, because unfunded liability contributions based on the payroll of active members gets reduced overtime. Citing a 2015 paper from the National Institute on Retirement Security (NIRS), columnist Michael Hiltzik wrote:

The main problem with closing defined benefit plans is that the demographics within the closed plans change quickly. Without new members coming in, the number of active workers making contributions shrinks. The loss of young members making contributions for years before retirement is especially damaging.[4]

That argument was based in part on this claim from NIRS about Alaska:

Generally speaking, when a DB plan is frozen, plan costs will increase. This is because the plan’s demographics tend to change rapidly… New members of a DB pension, by definition, do not start with any unfunded obligation for benefits. So, if Alaska kept open the DB pensions instead, these new employees would have resulted in the DB pensions getting a net funding contribution from a stable or growing group of employees rather than an ever smaller payroll base over which to spread the payments to meet the unfunded liabilities.[5]

If this statement were accurate, it would mean the best strategy for reforming a pension plan would be to hire as many people as possible. But, instead, the argument fails to accurately articulate how defined benefits plans are funded. (See Reason’s backgrounder on how pensions are funded.) It is true that new members of a plan would start out with zero unfunded liabilities. However, the new members would only mean additional contributions to a plan relative to the additional accrued liabilities-i.e. promised pension benefits-they would bring. The new members would not have resulted in extra contributions to help pay for previous member liabilities-funded or unfunded.[6]

Consider a counterfactual scenario where Alaska didn’t close the defined benefit plan for teachers (ATRS) but instead put all new teacher employees in a new Tier 2 with slightly altered benefits, but still defined benefits. The annual contributions made to the plan as a whole on behalf of the Tier 2 members would be only what is actuarially determined as necessary to fund the accrued liabilities of Tier 2 members in a given year, assuming a particular rate of return on the assets. The “net” contributions to the plan would be neutral relative to the additional liabilities from the new members. The contributions would have no bearing on the normal cost of Tier 1 members-for which separate contributions would be made-or for the unfunded liabilities of Tier 1 members. In practice this would be no different from keeping a plan open in general without separate tiers; it is just that the contributions on behalf of various employees are all bunched together, making it difficult to recognize at a glance how the funding actually works.

Moreover, the NIRS statement tacitly confuses expenses and costs. It is true that, when a pension plan is closing, the payroll gets reduced over time. However, the same dollar amount can be 10% of one amount of payroll and 50% of another amount of payroll, but it will still be the same expense. If a pension plan has a level-dollar amortization method, then over time the amortization payments will be measured as increasing as a percentage of payroll, but the percentage of payroll figure isn’t particularly relevant, as it is the dollar amount paying down pension debt that is measured in dollars.

The use of payroll as a medium for determining the rate an employer should pay toward pension debt isn’t anything more than a method for calculating a dollar amount. Every year the dollars that make the amortization payments in Michigan and Alaska flow out of general revenue of employers and into the plan assets of MSERS, APERS, and ATRS. Payroll also flows out of general revenue and into the bank accounts of employees. Additional employees would mean additional paychecks, but the dollars to pay the unfunded liabilities would come from the same source, with or without those employees.

Theoretically, it doesn’t matter if those dollar amounts are determined using a percentage of the payroll of active members or a percentage of income tax revenue, so long as there is a plan in place to pay off the unfunded liability. Having a declining payroll only means the dollar payments will have a larger percentage of payroll numerical representation, not that the actual costs of paying off the unfunded liability are somehow higher.

Myth: “Transitioning” from defined benefit to defined contribution inherently increases taxpayer costs.

Truth: Closing a defined benefit plan does not require a state to change the annual dollar amounts it is planning to pay toward its remaining unfunded liabilities.

Finally, there is a claim that the very nature of “transitioning” from defined benefit to defined contribution entails costs. In testimony before a legislative committee in Nevada that was considering closing the state’s defined benefit plan, Ghilarducci claimed:

[Since Alaska’s] 401(k)-type plan for new state and public school employees began in 2006…. Alaska required contribution rates nearly doubled. For Alaska state employees, the actuarially determined employer contribution rate required to pay off the unfunded liabilities increased from 12.4 percent of salary in 2006 to 22.5 percent in 2012. For teachers it was worse, the rate increased from 24.6 percent to 36 percent.[7]

There is nothing about closing a defined benefit plan that automatically increases contribution rates. The policy decisions of a state may increase contributions in the near-term due to a change in the amortization schedule that pays the debt off faster or defined contribution rates that are designed to be higher than defined benefit normal cost. The increase in contribution rate requirements in the years following a pension reform in Michigan and Alaska was specifically due to an increase in unfunded liability amortization payments.

The increase in amortization payments was not directly related to the closing of a defined benefit plan either. As outlined before, there are direct causes for the growth of pension debt due to the actuarial experience of the plans and funding policy of the plans-both of which would have affected Michigan and Alaska’s plans without reform. (And in Alaska, part of the increase in unfunded liabilities is due to a change in the discount rate, meaning the growth in pension debt was in part due to more honest and accurate accounting that does not undervalue pension liabilities as badly as before.[8])

Thus, Ghilarducci mischaracterizes the causes of the increase in unfunded liabilities, inappropriately laying it at the feet of the nature of closing a defined benefit plan instead of at the feet of a mismanagement of the closed pension plans. The growth of pension debt experienced by Alaska’s APERS and ATRS plans would not have happened if the return on investments had met the assumed average return and if the state had paid its actuarially determined employer contributions. Adopting a more manageable investment return target and paying the full costs of the plan shouldn’t be that much to expect of those responsible for funding public sector retirement benefits.

Moreover, even with the mismanagement, Alaska’s APERS and ATRS plans still have less in unfunded liabilities today than they would without reform. Closing a defined benefit plan closes the plan off to the liabilities that come with new hires-liabilities that are exposed to the undervaluing of improper discount rates and underperforming investment returns.

The reason that NIRS, Ghilarducci, and others perpetuate the myths about Michigan and Alaska is because they underappreciate the dramatic role that underperforming investment assets and underfunding contribution rates has played in causing the growth in unfunded liabilities. For Michigan, NIRS acknowledges that “other factors” have contributed to Michigan’s unfunded liability, including underperforming investments and underfunding the employer contributions.[9] However, it fails to attribute proper weight to these “factors.” The losses during the financial crisis years were so bad that even several strong years of investment returns have not been enough for MSERS to achieve its expected rate of return on average. And even though Michigan has made “larger payments” in recent years, those payments still haven’t been at or above 100% of the actuarially determined rate.

That said, even if required contributions to APERS and ATRS had not increased in dollar terms (assuming everything went well for the plans), the contribution rates would necessarily increase because of the simple arithmetic of the measurement point. With payroll declining, reporting contributions as a percentage of payroll is no longer a meaningful benchmark for how expensive a plan is. With a $1 billion payroll, contributions of $100 million are 10% of payroll, and with a $10 million payroll, contributions of $5 million are 50% of payroll-the latter is clearly less, even though the rate in percentage terms is higher. It is important to always distinguish between the expenses of a plan and the actual, long-term costs.


In the fall of 2014, the city of Phoenix was considering a ballot measure that would have introduced a defined contribution plan for new members. During a hearing at city hall a number of public sector union members stood up in opposition to the pension reform effort, citing the growth of unfunded liabilities in Michigan and Alaska as evidence that changing from defined benefit to defined contribution is bad for a municipal pension fund.

In the spring of 2015, the state of Nevada was considering a legislative effort to introduce defined contribution benefits for state employees. During a state assembly hearing, a number of public sector union members stood up in opposition to the pension reform effort, also citing Michigan’s and Alaska’s unfunded liability growth in the years following their pension reforms as a reason to reject the proposed reform.

In both cases, the local press repeated the claims in stories about the pension reform efforts, spreading the misinformed analysis about Michigan and Alaska. In both cases, the pension reform efforts failed-with Phoenix residents voting down their measure, and Nevada legislators killing their effort before it even got out of committee.

The individuals presenting these arguments about Michigan’s and Alaska’s pension reform efforts are sometimes innocent of the analytic failures behind them, such as when they are union members sent forward by their leadership, armed with talking points and tasked with presenting them. Other times those advancing the misleading arguments are in a position to know better, such as educated economists or actuaries with a personal or ideological stake in stopping the proposed reform.

A defined contribution plan might offer even more generous benefits than a defined benefit plan, and thus have a higher normal cost that increases employer contribution rates-but in such a case, the higher costs would be an explicit and intentional policy decision by elected officials, since the costs of a defined contribution plan are only the defined contribution rate. By contrast, unfunded liability pension costs are not due to elected officials choosing to budget for them, but instead are costs related to previous funding policy failures by elected officials.

In light of this, journalists covering pension reform stories should not repeat the myths about changing from defined benefit to defined contribution plans. There may well be good reasons to vote down a pension reform proposal-neither the Phoenix nor the Nevada measures were perfectly designed reforms. However, when informing the public about the merits or detractions of any particular reform that closes a defined benefit plan, the press should not repeat the myths that such a closure and replacement with defined contribution benefits will cause or has caused increases in unfunded liabilities.

— Footnotes —

[1] In both states, employers underfunded the plan by skipping out on required contributions, while at the same time the allocation of pension plan assets resulted in investment returns that significantly underperformed the long-run assumed rate of return. The actual savings have been much lower, but only relative to the funding policy failures of each state that are unrelated to pension reform. If Michigan and Alaska had managed the pension reform process correctly, they would have experienced the $7 billion and $4 billion in savings, respectively. The fact that their pension plans still have significant unfunded liabilities is due to mismanagement of the closed defined benefit plans, the costs of which have consumed the majority of the savings that pension reform created.

[2] Teresa Ghilarducci (2015), “Nevada PERS a model for nation; don’t change it,” Review Journal, March 16, 2015.

[3] Total required contributions to MSERS defined benefit plan between 1997 and 2014 have been roughly $5.5 billion, using the plan’s 8% discount rate to value accrued liabilities and determine contribution rates. Using a 7% discount rate, the contribution requirements would have been closer to $7.6 billion over the same time frame. For more on our policy findings about the most appropriate way to determine a discount rate, see: Anthony Randazzo and Truong Bui (2015), “Why Discount Rates Should Reflect Liabilities: Best Practices for Setting Public Sector Pension Fund Discount Rates,” Reason Foundation Policy Brief 130.

[4] Michael Hiltzik (2015), “Public pension shocker: Shutting a pension plan actually costs taxpayers money,” Los Angeles Times, August 21, 2015.

[5] National Institute on Retirement Security (2015), “Case Studies of State Pension Plans that Switched to Defined Contribution Plans,” February 2015, p. 10.

[6] Another way this particular myth is expressed is by arguing that because closing a plan cuts off new members, it cuts off contributions from those members. Ghilarducci claimed as much when she argued that Michigan’s growth in unfunded liabilities was because “new employees were not contributing to the old plan, and had to contribute to the new plan, [while] the system still had to pay benefits to the old members while assets were falling.” In one respect this can’t possibly be true, because employees in MSERS weren’t required to make contributions to the plan until 2011. Prior to reform MSERS was a “non-contributory” plan for employees, and after reform only members of the defined contribution plan had to make employee contributions. But even if there were required employee contributions, they would not be necessary from new members to help pay down the debts of previous members. Employee contributions are only supposed to be relative to their own benefits (see pages 6-7 of our recent policy study, “Did Pension Reform Improve the Sustainability of Pension Plans?“).

[7]Teresa Ghilarducci (2015), “Nevada Pension Reform: First, Do No Harm,” Oral testimony to the Nevada State Legislature.

[8] Changes to actuarial assumptions (in particular, the discount rate) have moved the state toward a higher-though more accurate-recognition of the value of accrued liabilities, and thus are reporting larger unfunded liabilities today simply by virtue of a change in the discount rate relative to the pre-reform years. Thus, at a very simple level it is inaccurate to suggest that the increased contribution rates in Alaska are because of the closed defined benefit portion of the plan.

[9] National Institute on Retirement Security (2015), p. 5.

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.

This field is for validation purposes and should be left unchanged.