Bob Baldwin's Reflections on DB and Pension Design

Bob Baldwin, a former director of PSP Investments who runs Baldwin Consulting and is a director at Addenda Capital, sent me an important paper, DB and Pension Design: Reflections on Risks and Reporting:
Over the forty years that I have served as a pension specialist inside the trade union movement and outside of it since 2005, debates about the relative merits of defined benefit (DB) and defined contribution (DC) pension plans have been a prominent part of pension discourse. The intensity of the debate has ebbed and flowed over the years but has been more intense in recent years as there has been a shift from DB to DC plans that has been particularly notable in workplace pension plans (WPPs) in the UK and US and to a lesser but clear degree in Canada as well. This shift has been much stronger in the private sector compared to the public sector and includes a shift from registered DC plans to group RRSPs (See: Appendix 1 and Baldwin, 2015). This shift has left the remaining members of DB plans feeling threatened and for many the sense of threat has been compounded by the emergence of target benefit (TB) plans which they see as an inferior alternative to DB.

As happens in many intense debates, participants tend to become advocates for particular points of view and confirmation bias sets in. People take in information that seems to confirm their opening, big picture view of what is best, and find reasons to dismiss non-conforming evidence. Recent discussions of DB, TB and DC provide ample evidence of confirmation bias at work.

DB plans are often revered by their supporters on the grounds that they can provide benefits that are more predictable than DC plans and they are capable of providing adequate retirement incomes. Moreover, if things go wrong financially, the employer will pay to right the financial ship. There are important elements of truth in this characterization of DB plans. But this popular characterization also obscures features of DB plans that are less flattering. (Unless otherwise noted adequacy will refer to the ability of pension plans to replace pre-retirement earnings. More technically it refers to the ratio of retirement income to pre-retirement earnings also known as the gross replacement rate (GRR)). Discussions of DB, TB and DC in the abstract tend to obscure basic issues related to the similarities and differences among these plans that should be taken into account. 

The abstract discussions also provide little or no guidance of how to address pension design issues in a context where the economic, financial and demographic context have created difficulties for all forms of pension plan. Unfortunately, plan sponsors often react to current circumstances by reflexively turning to pure DC without considering the full range of alternatives, and plan members dig in in defence of DB plans where they still exist – unwittingly exposing young plan members to considerable risks by doing so.

My hope is to provoke an examination of some of these issues free from the hyperbole that is widely invoked in partisan debates about DB and DC.

Diversity under each of the DB and DC umbrellas

In debates about the general merits of DB and DC, proponents of DB plans tend to have a particular type of DB and DC plan in mind. The DB plan implied in much of the commentary is a “Best or Final Average Earnings” plan with benefits that are indexed to inflation after benefits begin to be paid. On the DC side it is implied that we are dealing with a plan in which investments are “self-managed” during the pre-retirement period and the run down of assets during retirement is also “self-managed”. If these types of DB and DC plans were the only types that existed, the preferential views for DB noted above would be easy to accept more or less at face value.

But the reality is that DB and DC plans come in quite a variety of shapes and sizes and not all of the DB plans include the virtues commonly ascribed to DB.

The world of DB includes plan designs that may fall quite short on income adequacy and to a lesser degree, on predictability. A member of a career average earnings (CAE) plan that includes no upgrade in the earnings base used to calculate benefits will likely provide inadequate benefits in periods of strong wage growth and benefits that will not be predictable until close to retirement. The same problem can arise with flat benefit (FB) plans. These DB plans can be managed so that benefits are predictable and adequate. But there is nothing inherent in the plan design that makes them so. But note: these are DB plans.

In an era when retirement periods commonly span 20 or more years, post retirement adjustments to reflect price or wage increases become increasingly important. Over a twenty year time span, annual inflation at the 2 per cent rate targeted by the Bank of Canada will reduce the purchasing power of a pension that provides no protection against inflation by one third. About one third of Canadian DB plans with just under half of DB plan members provide no automatic adjustments in the post retirement period. In these plans, benefits may be predictable at the date of retirement, but their purchasing power after retirement is not predictable. (See: Appendix 2) (Only one in five DB plan members gets full protection from inflation as measured by the Consumer Price Index (CPI). The remainder get some form of partial protection against inflation).

On the DC side, a similar problem arises.

The self-managed plans not only involve a high degree of uncertainty with respect to the benefits they will provide, but the contribution levels may be inadequate to provide adequate retirement income in low return environments like the current one. Self-managed DC plans also impose decision-making responsibilities on people who have little time to do the work required to make good choices and who don’t have appropriate investment expertise.

Over long time frames the adequacy problem can be addressed through higher contribution rates. The predictability problem can also be addressed through the choice of a specific plan designs that do not rely on self-management.

Prior to the early 1980s most DC pension plans in Canada provided benefits based on the purchase of deferred annuities. DC plans organized in this way provide much greater predictability than self-managed plans but will fall between self-managed DC plans and BAE or FAE DB plans in terms of predictability. Whether they produce adequate benefits conceived of in GRR terms will depend on a variety of factors including the contribution rate, the interest rates embedded in the annuities and the earnings trajectory of the plan members.

There are also a number of Canadian DC plans that provided DB guarantees in 1990 – the time of our last major reform to the tax rules governing pensions and RRSPs. The plans of this sort that were in existence at the time were allowed to keep operating but new plans of this sort could not be registered with the Canada Revenue Agency. These plans provide the same downside protection as DB plans while offering the upside investment risk associated with DC plans. They provide the same if not greater opportunity to provide adequate incomes.

Looking beyond the Canadian border, there are many examples of DC plans that provide minimum rate of return guarantees. The downside protection offered by these plans is similar to a career average adjusted plan like the Canada Pension Plan (CPP) while offering upside investment opportunities.

It is worth noting too that plans don’t always get administered in the way that the basic benefit design might suggest they should be. In the 1980s and 1990s both stock and bond returns were high and it was common for DB plans to be in a surplus position (i.e. plan assets exceeded plan liabilities). In many cases where DB surpluses existed, the surpluses were used in whole or in part to finance benefit improvements. To the extent that plans were managed in this way, they were managed as if they were collective DC plans with strong minimum DB guarantees! The DB formulas were guaranteeing a minimum level of benefits and surpluses were paying for enhancements.

The main point to be made here is that abstract debates about the virtues of DB and DC plans are of limited value. There is too much variation in the specifics of plan design under each heading. WPP design is more like a spectrum of choice rather than a binary choice between DB and DC. What is of most importance is to assess the retirement income risks faced by the plan members and make sure they are addressed in a way that is fair among plan members and is reasonable in terms of the level and volatility of required contributions. Some of the issues related to fairness and required contributions are addressed below.

What are we trying to achieve?

Generally workplace pension plans are designed to allow people to maintain their standard of living after they leave paid employment. As was noted above, this post employment period can be a long one. Recent annual reports of the Ontario Teachers’ Pension Plan draw attention to the fact that the average retirement period under that plan is greater than the average period of employment as a teacher.

Discussions of pension plans adequacy often focus exclusively on what plans produce in the retirement period and the consideration of adequacy is usually cast in terms of the GRRs they will produce. This perspective is too limited.

Pension plans affect living standards in both the pre-retirement period and the post-retirement period. In the post-retirement period we will be interested in the extent to which the pension replaces pre-retirement earnings. But in the pre-retirement period we have to be concerned with the extent to which the pension plan is depressing the ability of plan members to buy goods and services prior to retirement.

Ideally, the pre-retirement sacrifice will combine with post retirement benefits so that living standard will be the same in both periods. If too little is given up prior to retirement, living standards will drop in retirement. If too much is given up, living standards in the pre-retirement period will be depressed below their post-retirement level. (It is possible to have too much pension)! Neither over contributing/saving nor under contributing/saving is inherently desirable. Unfortunately, for reasons that will become clear below, it is impossible to get the balance of pre-retirement sacrifice and post-retirement benefits right for all member of a DB plan. This does not detract from the fact that the objective of continuity of living standards should be sought.

In thinking about getting the balance right, three additional things need to be considered.

First, in describing the objective of maintaining living standards in retirement, so far I have referred exclusively to the role of WPPs in achieving this goal. In most high income countries, publicly administered pension plans will also contribute to meeting this objective. Given the structure of Canada’s publicly administered pension plans (OAS, GIS and CPP), the objective of maintaining living standards in retirement will be met for Canadians with lower earnings - half average wages and salaries and below – by the publicly administered plans alone. But above that level of earnings, income from privately administered sources will be needed in order to maintain living standards in retirement and the need will grow quite rapidly through the middle and upper middle earnings ranges.

Second, it has been a tradition in both public policy discourse and retirement planning to argue that a retirement income amounting to 70 per cent of pre-retirement earnings is a reasonable benchmark for being able to maintain living standards in retirement. Recently this approach to assessing retirement income adequacy has been subject to two types of challenge.It has been argued that this benchmark too high for the middle earners who are of key policy interest. (See: Mintz, 2009). GRRs in the range of 50 to 60 per cent are becoming more common among pension experts in Canada.

It has also been argued that gross pre-retirement earnings and retirement income are weak indicators of living standards in the pre and post retirement periods. In the pre-retirement period, a significant wedge between gross earnings and the consumption possibilities of a pension plan member can be created by the need to support children, mortgage payments, taxes and, of course, pension contributions. In the post retirement period many of the factors that are relevant in the pre-retirement period will cease to be relevant for most pension plan beneficiaries (e.g. mortgage payments, transfers to children, etc). But taxes will still be relevant for most. Also some adjustment to income to take account of the value home equity is appropriate. The adjustment could take the form of imputed rent or annuitization. The presence and economic situation of a spouse is also important.

Recognizing that living standards are not well defined by gross earnings and pension income gives rise to a number of adjustments that give rise to a net replacement rate measure. A given gross replacement rate may translate into a range of net replacement rates. Within a DB plan that is designed to generate a given replacement rate (e.g. 70 per cent after 35 years of service) there will be a wide range of net replacement rates after 35 years of service reflecting differences among the members in terms of their family situations and whether they own their own home. The fact that DB plans are designed to generate GRRs rather than NRRs forces us to recognize that NRRs may cover a wide range and won’t be perfect for all members. However, this reality does not allow us to ignore the importance of NRRs.

Once it is accepted that pension contributions depress pre-retirement living standards and increase NRRs, an important question arises: what contributions should be entered into the calculation of pre-retirement living standards. For plan members who have deductions made from their paychecks to support a pension plan, the most obvious contributions to take into account are the deductions from each paycheck. But should employer contributions be ignored?

There is good reason for including some portion of employer contributions in the calculation of what plan members give up in order to get their pension benefits. In most situations it is fair to surmise that an employer is most worried about total labour costs not the component parts of the cost. To the extent this is true, a rational employer will discount other elements in the compensation package of employees to take account of the contributions to the pension plan that are predictable. Thus the economic burden of employer contributions will be shifted from employers to employees – rather like sales taxes being shifted from vendors to consumers. In DC plans the required contributions will usually be predictable but this will less often be the case in DB plans.

In DB plans contributions are unpredictable (more on this below). There may be situations where an employer is confronted with unexpected need to make special contributions to a DB plan based on an actuarial deficit showing up in an actuarial valuation report. The employer may find it impossible to shift the burden of the special payments to employees in the short term due to labour market conditions and/or contractual obligations to employees.

All things considered it would be a mistake to ignore employer contributions in considering what plan members give up in order to earn DB benefits.

Some changes that have affected retirement savings plans

It is a basic axiom of all types of pension plan that the effects of the plan will depend not only on the benefit and financing rules in the plan but also on the way the rules interact with an ever changing labour market, demographic, financial and economic environment. A stable set of plan rules does make the plan stable in terms of the benefits it will provide or the cost of providing the benefits. Some of the variables that affect the outcomes of pension plans are quite unpredictable over short and even longer time frames. This is especially true of investment returns which play a central role in the financing of pre-funded pension plans.

Any number of changes in the labour market, demographic, financial and economic environment have affected to operation of workplace pension plans in recent years. But three merit particular attention: investment returns, life expectancy and entry and exit ages from the labour force.

As was noted above, returns on both stocks and bonds were very strong in the 1980s and 1990s. Since 2000, returns have been weaker. The high returns of the 1980s and 1990s were important in their own right and their contribution to pension finance was accentuated by slow growth in wages and salaries. Slow wage and salary growth meant that earnings replacement targets were growing relatively slowly. In DB plans, plan liabilities were growing more slowly than would have been the case with rapid wage growth. At that time providing retirement incomes proved quite inexpensive. Changes in the financial environment have made retirement incomes more expensive. (More on this below). (See: Appendix 3)

Regarding life expectancy, in the 14th Actuarial Report on the Old Age Security (OAS) program, the Office of the Chief Actuary (OCA) points out that the average life expectancy of a 65 year old Canadian has increased by roughly six years between 1966 and 2016. By 2060, average life expectancy at 65 is likely to increase by another four years. (OCA, 2017) This is, of course, a very welcome development. But to state the obvious, it also means that if people keep starting their pensions at the same age, the pensions will be paid out over longer periods of time and will be more expensive.

Low returns in general - and lower interest rates in particular - have combined with increasing life expectancy to raise the cost of providing a dollar of annual retirement income. Thus, in establishing a lump sum value of a dollar of DB benefits in workplace pension plans, Statistics Canada estimates that the cost of an indexed benefit has increased over the relatively short period from 1999 to 2016 by 60 per cent.

These developments have put upward pressure on DB contribution rates (See: Appendix 4). They have also contributed to the shift from DB to DC forms of workplace pension plan – especially in the private sector. In the public sector they have contributed to a shift from pure DB to plans that are largely DB but place some financial risk on the benefit side of the plan by making the indexation of benefits in pay contingent on the funded status of the plan. Target benefit plans have also been introduced in the public sector in response to the financial difficulties of pure DB plans.

Another development meriting comment is changes in the ages of entry and exit from the labour force. There has been a general trend in recent years for the average age of both entry and exit to go up – with the increased age of exit being more clear than the entry age in the aggregate data. At the same time that the average age has been going up, entry and exit ages for the society/economy as a whole have been becoming more diverse. For public pension plans like OAS and C/QPP, these developments raise important questions about the appropriate age of eligibility for pensions and whether the role of specific chronological ages should continue to be the dominant criterion for eligibility. Should a person who enters the labour force at 18 qualify for benefits at the same age as someone who enters at age 30?

The entry and exit ages for specific workplace pension plans are likely to vary from the society/economy wide norms – in some cases quite significantly. The entry and exit ages in each plan need their own study and analysis. One question worth asking in the context of plans with special early retirement provisions is what portion of new entrants will qualify for them under the conditions when they first become available. In the case of the federal public service superannuation plan, between the 1980s and early 2000s, there was a significant decline in the portion of new plan members who would qualify for special early retirement benefits when they first become available. (See: Baldwin, 2012.)

What allows DB benefits to be predictable?

There is an accounting identity that applies to all types of pension plans:

Benefits = contributions + investment returns – expenses.

Given that the basic building blocks of all types of pension plans are the same, an obvious question arises: what allows a DB plan to provide more predictable benefits than a DC plan. The answer lies in two related but distinct features of DB plans.

First, the contributions to DB plans are adjusted on a regular basis so that retirement income objectives that are embedded in the DB benefit formulae can be met. In the Canadian context these adjustments will be made based on actuarial valuation reports that must be prepared with a frequency no greater than three years.

The second feature of DB plans that allows them to provide predictable benefits is cross subsidies within and between various cohorts of plan members. The regular adjustments to DB contribution rates mean that different cohorts of plan members get different effective rates of return on their pension contributions – hence the overlap between the cross-subsidies issue and the adjustments to the contribution rate.

There is also a wide range of cross-subsidies within cohorts. Those with short periods of retirement subsidize those with long periods of retirement, early entrants to the plan subsidize late entrants and so on.

There has been some interest in and measurement of cross-subsidies across cohorts among pension analysts (See for example: Cui, deJong and Ponds, 2011 and Kortleve and Ponds, 2010). Less interest has been shown in cross-subsidies within cohorts. An exception is provided by Young, 2012 who identifies a limited number of cross-subsidies and provides a measure of their impact. Blommestein et al, 2009 identify a number of cross-subsidies within cohorts but don’t provide a measure of their impact.

The level of contribution rates to a pension plan is important in determining both post-retirement benefit levels and the pre-retirement loss in consumption possibilities. But investment returns commonly account for two thirds to three quarters of benefit payments. By their nature, investment returns are unpredictable and the way that the returns are shared within and between cohorts in a DB plan is important in determining who is subsidizing whom.

In context, it is worth noting that both plan members and employers who sponsor DB plans face parallel dilemmas in terms of the degree of financial risk that should be accepted in order to get higher investment returns. Plan members would like to maximize the benefits they get from their contributions which would tend to push them to seek higher returns which in turn would mean higher levels of risk. But they also want benefit security which would push them in the opposite direction. Employer plan sponsors would like to minimize required contributions for promised benefits which would tend to push them toward higher risk investments. But they would also like predictable contributions which would push them in the opposite direction.

Some related transparency problems in DB

The reality of cross-subsidies within DB plans is not a problem in and of itself. Cross-subsidies are inherent in any form of insurance and people do place positive value on insurance – even when they know that they may get less of a “return” on their insurance premiums than other purchasers of the same insurance. Unfortunately the cross subsidies in DB plans are not always recognized and seldom if ever measured. As a result, it is very difficult for plan members to assess their impact and make an informed decision on whether they are “worth it.”

It is conjecture on my part, but my sense is that plan members have a dim sense of some cross-subsidies and accept them within limits without having a clear sense of their financial impact. Others are less clearly perceived and may not be as welcome if they were clearly perceived. An example of the former would be the cross-subsidy from members with short retirement periods due to early death to those with long lives and hence long retirement periods. This is an interesting cross-subsidy because it is constrained in many DB plans by the presence of a guaranteed minimum period of payments. The presence of minimum guarantee periods raises a question about the willingness of plan members to be an open-ended source of subsidization.

A source of cross-subsidy in final and best average earnings plans that is less clearly perceived is the cross-subsidy from plan members whose earnings are flat or declining as they approach the age of pension receipt to those whose earnings are increasing rapidly. It is a moot point whether this particular cross-subsidy would garner substantial plan member support if it was perceived and its financial effect was known.

The point here is not to argue that cross-subsidies are wrong. But, as far as possible, they should be identified and measured so that plan members can have some sense of whether they are “worth it.”

DB plans have a key premise and that is that the sponsor of the plan (and/or the plan members) has (have) an unlimited willingness and ability to contribute more to the plan. This is an implausible premise. As long as labour market, demographic and financial circumstances remain within limited boundaries, the implausibility of the premise will not be clearly visible. But in circumstances like those of the early 21st century, the difficulty with the premise becomes clear. This creates a transparency problem: plan governors do not articulate the outer limit of acceptable levels of contribution and don’t explain to plan members what happens if that limit is reached. This becomes an increasing problem as increases in the contribution rate run the risk of pushing pre-retirement living standards below post-retirement levels.

Lurking beneath the surface of much of the discussion in this note is a reflection on the extraordinary range and degree of risks in trying to provide an adequate and predictable retirement income. It seems simple enough to promise a 30 year old that for each year of service they put in under a DB plan, they will get payments amounting to 2 per cent of their best five years earnings for their retirement period and a smaller payment to a surviving spouse during their lifetime. In fact this is a promise that is full of uncertainties: the best 5 years’ earnings are unknown, the start date of payments is unknown, the end date of payments to the plan member and their spouse are unknown and the size of any post-retirement adjustments to reflect inflation are unknown. Moreover, we don’t know the rate of return that any money set aside today to help make the payments will earn. This money set aside today may or may not be sufficient to pay future benefits. The pervasiveness of these uncertainties dictates the need for regular financial assessments of DB plans.

These key sources of uncertainty are addressed in traditional financial reporting but they are addressed in ways that tend to divert attention from the uncertainties associated with the variables. Traditional financial analysis of pension plans relies on the projection of key variables at fixed rates through time. In other words, they rely on fixed rates of wage growth, fixed rates of return and so on through time. This form of modelling of the financial future of pension plans is known as deterministic modelling. Modelling in this form can answer basic “yes/no” questions such as: are the “normal” contributions high enough to cover the cost of the newly emerging benefits, are the assets in the plan sufficient to cover the cost of the benefits that have accrued to date?

The problem with this traditional approach is that it fails to capture all of the uncertainties that arise through time with the full range of variables. It may be true on the basis of single-valued assumptions about the future that are chosen for the purposes of analysis that a plan’s assets are sufficient to pay benefits promised to date. But, it may also be true that there is a reasonable chance that shortfalls will arise in the future and prompt the need for contribution rate increases or benefit reductions. These possible consequences of uncertainties are accentuated in plans that have risky investment portfolios and are mature plans. The potential need for these types of adjustments is not clearly brought into focus in traditional analysis.

In DB plans, the assumptions about the future that underpin the financing of a plan are made clear in actuarial valuation reports. In DC plans the assumptions about the future that underpin the plan may not be formalized but may be implicit in claims that contributing to the plan at a certain rate will generate a retirement income amounting to some per cent of pre-retirement earnings. In the case of both types of plan, there is a possibility – indeed a likelihood - that the future will be mis-estimated. In the case of a pure DB plan, the mis-estimations are corrected entirely through changes to the contribution rate and in a pure DC plan through changes in the benefits.

DB and TB

So far I have focused attention on the historic debate about the virtues of DB and DC. More recently, debates have also emerged about the virtues and vices of target benefit (TB) plans compared to DB.

TB and DB plans have a logic that has a common point of departure. Unlike DC plans both DB and TB plans promise plan members a benefit that will be paid for each year of service in the plan. Key risks are pooled in both types of plans so that there is a reasonable expectation that promised benefits will be paid. Using broadly similar actuarial methods, both types of plans establish a contribution rate required to pay the benefits promised by the plan. (In many if not most cases, the benefit level that is promised will reflect a limit that the plan sponsor or tax law will impose on required contributions and/or maximum benefits. To the extent this happens, it introduces a DC element into the logic of DB and DC).

The main thing that distinguishes DB from TB plans is what happens when financial problems arise.

In pure DB plans, when things go wrong financially, there are two options for correcting the problem:

1) Increase contributions;
2) Reduce future benefit accruals.

In TB plans these two options are available as is one more:

3) Reduce accrued benefits.

Option 3 is generally not available for DB plans because pension benefits law in all Canadian jurisdictions other than the federal jurisdiction prohibit the reduction in accrued benefits. 

It is important to note that the difference between DB and TB has different implications for different cohorts of plan members. The options open to DB plans create no risk for retired members and little risk for members close to retirement age. But they expose the young and future plan members to most of the financial risk of DB plans. The TB options expose all cohorts to financial risk that will vary depending on the specifics of the plan.

As is the case with DB and DC plans, there can be a range of specific designs of TB plans. Two variables in the design of TB plans are crucial:

1) Is there any room for variability in the contribution rate; (the more variability there can be in the contribution rate, the more TB will perform like DB);
2) To what extent is positive experience used to recoup benefit losses in earlier time periods (the higher the priority that is given to this use of surplus, the more TB will perform like DB),

TB plans have been operating in Canada for many years. These plans have been created at union initiative to provide pensions to workers in industries with large numbers of small employers in which it would not be practical to establish DB plans at each workplace. These plans are generally known as multi-employer plans (MEPPs).

MEPPs have several characteristics that are worth noting:

1) Contributions to them are fixed during the term of collective agreements that require employers to contribute to them;
2) Regulatory law includes specific provisions that apply to MEPPs and allows accrued benefits to be reduced;
3) Regulatory law requires that half of the governing body of a MEPP be made up of plan member representatives.

Until recently, TB plans have been restricted to the MEPP context. Several provincial jurisdictions have adopted or are considering legislation to permit single employers to adopt TB plans.
Regulatory changes to permit single employer TB plans have become particularly controversial in the federal jurisdiction. The federal government has introduced legislation (Bill C-27) to permit the registration of single employer TB plans. The changes have been vociferously opposed by some unions and retiree groups that would potentially be affected by the legislation. There are important respects in which Bill C-27 needs to be altered to protect plan member rights. But its general direction is consistent with the shared theme of the Ontario Expert Commission and the Joint Expert Panel on Pensions in Alberta and BC that Canada’s regulatory and tax law needs to be amended to accommodate plans that embody elements of DB and DC.

Concluding thoughts

I want to conclude this paper by summarizing what I see as key points made above. But, I also think it would be helpful to put the foregoing discussion that focuses of WPPs in context.

Canada’s RIS is structured in a way that middle and upper earners have to get income from either WPPs or individual savings plans in order to achieve a standard of living in retirement that is comparable to what they had beforehand. Based on evidence of a variety of types, this basically means participation in a WPP – preferably one with a DB element to it. While the success of WPPs is of primary importance to the well being of people coming to retirement age, it has wider importance too. (See: Baldwin and Moore, 2015, Baldwin, 2017 and Baldwin and Shillington, 2017) With a growing portion of the population being made up of retirees, the success of WPPs will also be important to maintain robust domestic demand for goods and services and success will also have a positive effect on fiscal balances in the future.

The declining participation in WPPs and the shift away from plans with DB elements are discouraging with respect to the potential role of WPPs in providing retirement income. Looking beyond these recent developments, another longer term reality needs attention. WPPs have never been widely available in the small employer context. Indeed most small employers lack the scale and expertise to serve as an adequate platform for administering a pension plan. Bearing in mind that the financial services industry has not come up with adequate solutions to this problem, finding an effective organizational platform for small employers is an important challenge. It is unlikely though, that plans that satisfy this need will be of a pure DB type. Hence there is a further need to keep exploring the space between pure DB and pure DC.

The need to keep exploring plan designs coupled with the need to adapt plans to a changing environment raises another issue. The regulatory law that governs WPPs was crafted at a point in time when most members of WPPs in both the public and private sector belonged to DB plans. The objective of the law was to protect DB plan members from errors and/or abuse by employers. The need for the regulatory law to be more flexible with respect to plan design was an important theme of the Ontario Expert Commission on Pensions and the Joint Expert Pension Panel created by the governments of Alberta and British Columbia. I would strongly endorse this theme and argue that it needs to be complimented by more flexibility to adapt to changing circumstances. But this latter adaptability needs to be made safe for plan members by encouraging joint plan member/employer governance as in the Jointly Sponsored Pension Plans in Ontario.

In previous sections of this paper I have argued a few key points:

1) There is enough variation in specific plan designs under each of the headings of DB, DC, and TB that there is little value in arguing the virtues and vices of plan design at that abstract level.
2) Too little attention has been paid to:
a. the impact of plan design on pre-retirement living standards.
b. the impact of changes in the socio-economic environment on pension plan benefits and costs.
c. the impact of financial risk in DB plans both in the aggregate and for specific cohorts of plan members.
3) DB plans have transparency problems in not identifying what happens when things go wrong, what cross-subsidies are embedded in the plans and what financial risks are embedded in the plans.

I don’t wish to end on a negative note. For plans that have DB elements I would recommend the following:

1) Establish a clear appreciation of current and future plan members’ financial needs throughout the retirement period.
2) Balance retirement income needs with impacts on pre-retirement living standards with a view achieving continuity of living standards.
3) Be aware and sensitive to differences within the membership group (working from averages and medians is never adequate).
4) Understand the risks in providing the benefit promises and who gets the rewards and burdens associated with the risks.
5) Be as clear as possible about cross-subsidies within and between cohorts of plan members.
6) Be clear about what will happen if things go wrong.

As noted above, pension plan design is more like a spectrum of choice rather than a binary choice between clearly defined DB and DC plans. The position of plans on the spectrum will be established by the way that financial risk is allocated between contribution and benefit rate variability and between and within cohorts of plan members and employers – to the extent that the latter bear financial risk.


Baldwin, Bob, 2016. Assessing the Retirement Income Prospects of Canada’s Future Elderly: A Review of Five Studies. (Toronto: CD Howe Institute).

Baldwin, Bob, 2015. “The Economic Impact on Plan Members of the Shift from Defined Benefit to Defined Contribution Pension Plans” in Canadian Labour and Employment Law Journal, Vil. 19 No1. (Toronto: Lancaster House).

Baldwin, Bob, 2012. The Federal Public Service Superannuation Plan: An Agenda for Reform. Montreal: Institute for Research on Public Policy).

Baldwin, Bob, 2017. The Pensions Canadians Want: The Results of a National Survey. (Toronto: Canadian Public Pension Leadership Council).

Blommestein, Hans, Pascal Janssen, Niels Kortleve and Juan Yermo, 2009. Evaluating the Design of Private Pension Plans: Costs and Benefits of Risk-Sharing. (Paris: Organization for Economic Cooperation and Development).

Canadian Institute of Actuaries (CIA), 2015. (Ottawa: CIA).

Cui, Jiajia, Frank deJong and Eduard Ponds, “Intergenerational Risk Sharing within funded pension schemes” in Journal of Pension Economics and Finance Volume 10, Issue1. (Cambridge: Cambridge University Press.

Dimson, Elroy, Paul Marsh and Mike Staunton, 2018. Credit Suisse Global Investment Returns Yearbook 2018: Summary Edition. (Credit Suisse Research Institute).

Kortleve, Niels and Eduard Ponds, 2010. How to close the Funding gap in Dutch Pension Plans? Impact on Generations. (Boston, Center for Retirement Research).

Landon, Stuart and Constance Smith, 2019. Managing Uncertainty:The Search for a Golden Discount-Rate Rule for Defined-Benefit Pensions. (Toronto: CD Howe Institute).

MacDonald, Bonnie-Jeanne, Lars Osberg and Kevin Moore, 2014. How Accurately does 70% Final Earnings Replacement Measure Retirement Income (In)Adequacy? (Toronto: International Centre for Pension Management).

Milligan, Kevin and Tammy Schirle, Rich Man, Poor Man: The Policy implications of Canadians Living Longer. (Toronto: CD Howe Institute).

Mintz, Jack, 2009. Summary Report on Retirement Income Adequacy Research. (Ottawa:Department of Finance).

Office of the Chief Actuary (OCA), 2017. 14th Actuarial Report on the Old Age Security Program as at 31 December, 2015. (Ottawa: Office of the Chief Actuary).

Ontario Teachers’Pension Plan (OTPP), 2018. Plan Sustainability: Annual Report for 2017. (Toronto: OTPP).

Organization for Economic Cooperation and Development, 2017. Pensions at a Glance 2017: OECD and G-20 Indicators. (Paris: OECD).

Pesando, Jim, 2008. Risky Assumptions: A Closer Look at the Bearing of Investment Risk in Defined Benefit Pension Plans. (Toronto: CD Howe Institute).

Rachel, Lukasz and Thomas D. Smith, 2015. Secular drivers of the global real interest rate. Staff Working Paper No. 571. (London: Bank of England).

Vettese, Fred, 2016. How Spending Declines with Age, and the Implications for Workplace Pension Plans. (Toronto: CD Howe Institute).

Wolfson, M, 2011. Projecting the Adequacy of Canadians’ Retirement Incomes: Current Prospects and Possible Reform Options. (Montreal: Institute for Research on Public Policy).

Young, Geoffrey, Winners and Losers: The Inequities within Government-Sector, Defined Benefit Pension Plans. (Toronto: CD Howe Institute).
First, let me thank Bob Baldwin for sending me such a rich contribution to the DB/ DC/ TB debate. This paper is well-researched, well-written and well-thought-out (note, I didn't include footnotes).

When I was working at PSP, I had the opportunity to hear Bob's views at some of the board of directors meetings I attended and I always found his comments balanced and very sensible.

I will keep my comments brief and on target. I myself am guilty of always extolling the virtues of DB plans and berating DC plans, portraying them in a very negative light.

I'm not going to lie, I hate DC plans, don't think they're real pensions as they don't have any income security attached to them, at least not any of the ones I am aware of. I see DC plans much like I see RRSPs (401 K(s)) or TFSAs, a supplementary savings plan which is tied to the whims and fancies of public stock markets. Bull market, you win, bear market you lose, especially a long and brutal bear market.

This is why I am in favor of large, well governed defined benefit (DB) plans which share the risk of the plan. But even here, I need to be a lot more specific. If we are talking about a mature plan especially (more retired than active members), then I believe adopting condition inflation protection is a must.

Why? Because it ensures inter-generational equity and it's mostly painless on retired members as it's typically implemented for a short period and in some cases, if the plan gets fully funded or over-funded, benefits are not only fully restored but retroactively fully restored.

There is something else Bob touched upon in his paper at the end, namely, how WPPs have never been widely available to smaller employers.

This is why in my comment revisiting the DB pension plan model failure, I explicitly stated:
I believe large corporations shouldn't be in the DB plan business (apart from some exceptions I cited previously). They don't get it and are incapable of fulfilling the pension promise. Businesses should focus on their core business and let pensions to the experts.

Moreover, I fundamentally believe we can offer Canada's corporations real, long-term solutions to their workers' pension needs not by de-risking them and doing away with them, but by bolstering them and creating a national plan that covers all workers properly just like CPPIB covers all Canadians properly.

As I stated last week: "If we want to improve corporate DB plans, all we need to do is look at the success of Canada's large public DB plans and model something based on their governance and investment approach."
CPP/ CPPIB is a great model, one I always believed we should expand (enhanced CPP) and even emulate in other pensions, specifically covering the workplace.

On the topic of Bob Baldwin's paper, Bernard Dussault, Canada's former Chief Actuary, shared this:
No surprise, Bob Baldwin has once again produced a magnificent paper on pensions. His analysis of the DB, DC and FB pensions is as usual thorough and as objective as can be. All explanations are provided with deep scientific rigour and the few opinions expressed in the paper are identified as such, politically impartial and not taking side for either the pension plan members and sponsoring employers. Bob’s paper deserves to be read by all Canadian pension actuaries.
Pretty high praise from Canada's former Chief Actuary but I agree with Bernard, Bob's paper should be read by actuaries and investment professionals.

Another eminent actuary, Malcolm Hamilton, who retired from Mercer and now enjoys being free of any constraints, shared this with me on Bob's paper:
Bob has an understandable DB bias given his long history with organized labour. In this paper he objectively examines the DB vs. DC debate and explains why it is virtually impossible to take that debate to a clear, unambiguous conclusion. There are too many employers (public sector, private sector, big, small,...), too many employees (married, single, young, old, male ,female, low income, high income, financially literate, financially illiterate...), too many economic environments (high inflation, low inflation, high interest, low interest...), too many plan designs (DB, DC , TB, hybrid...) and too many complications (How much do retired Canadians really need? How much do DB plans really cost? How should we trade off risk and return, low contribution vs. stable contribution, etc?).

Finding the best design for both employees and employers in every circumstance is too great a challenge. Bob did the reasonable thing. He sets out the reasons why designing a pension plan is difficult and warns people about the dangers of cutting corners or becoming wedded to a particular solution. Remember the old saying...
"To the man with only a hammer everything looks like a nail."
Readers hoping for a "magic bullet" solving all our problems will be disappointed, but this is as it should be. They have set their sights too high.

Personally, I would have spent more time on the following:
  • The importance of viewing pensions as a compensation element. Who should pay for the cost of the pension plan and how should the cost be measured for this purpose?
  • The link between cost, risk taking and risk bearing - how does the taking and bearing of risk affect pension cost? How should we price pension plans if we want to treat both employees and employers fairly?
  • What accounts for the large difference between public and private sector pensions?
Philip Cross and I addressed these topics in our paper for the Fraser Institute (Leo here: see my comment on the dirty secret behind Canada's pensions and more on Canada's dirty pension secret).
Interestingly, in an update to my comment revisiting the DB pension plan model failure, Bernard Dussault shared this on "The DB Pension Plan Model Failure":
When both the CPP and the three public pension plans covering the members of the federal public service, the Canadian Forces and the RCMP started investing their net contributions in private market (January1998 for the CPP and April 2000 for the public plans), the real rate of return assumed on the underlying funds for purposes of the triennial statutory actuarial reports was 4%.

Since then, 4% was tweaked a little bit (slight increases) a few times to come back after a while essentially to 4%.

Despite the disastrous negative return of 14% return on the CPP fund in 2008, the prescribed 9.9% contribution rate could until now be securely maintained each year since 1998. Likewise, even if the plan covering the federal public servants was subject to a significant 9.4% (of plan liabilities) deficit as at March 31, 2011, the plan had as early as March 31, 2014 developed a 3.9% (of liabilities) surplus.

In other words, following the 2008 extreme downturn, by far the worst economic downturn since 1929, two well designed and properly governed DB pension plan proved they could well survive without having to make any change to their DB design/structure. This tells me that if anything needs to be changed to the DB plans, it is their governance, particularly the financing policies, e.g. contribution holidays should be fully prohibited, severe measures should apply when the special contributions required following a deficit are not paid, valuation assumptions should be set on a realistic basis erring on the safe side, etc.

With such proper financing policy, the fluctuations in the level of contributions are normally minor and of short duration. If this would still be a major concern for a sponsoring employer, there would then be a case to offer to the plan members whether they are interested in taking over the underlying small financial risk pertaining to such fluctuations in the contribution rate.
  Malcolm Hamilton took issue with Bernard's comments:
Bernard must have forgotten the reasons why the CPP and the the public service pension plan (PSPP) performed so well in 2008:
  • The CPP had no problem because it was only 20% funded. The losses were small because the CPP had so little to lose relative to its liabilities. Even if the entire pension fund had been lost, a return of -100%, the contribution rate would have increased by less than 2 percent of covered payroll. What can we learn from this? Badly funded pension plans shine in bear markets because they have little to lose.
  • The PSPP had no problem because the assets supporting benefits earned prior to 2000, 75% of the pension fund in 2008, were held in a superannuation account earning a guaranteed (by the federal government) 7% return. Only 25% of the pension fund was actually invested in the capital markets. What can we learn from this? It's nice to have the federal government guaranteeing a 7% return on 75% of your pension fund in a catastrophic bear market.
The CPP and PSPP excelled in 2008 due to poor funding and government guarantees. Good design and exemplary governance had nothing to do with it. In fact, neither is evident.
I obviously don't agree with Malcolm on that last point, there's no question that good governance had something to do with mitigating the losses in 2008.

Bernard was kind enough to share this on Malcolm’s reaction to his observations on the CPP and the federal Public Service Pension Plan (“FPSPP”), i.e. the plan described in the PSSA (Public Service Superannuation Act):
Let me first clarify that my observations exclusively pertained to:
  • FPSPP2 implemented on April 1, 2000 subject to a fully funded financing policy, as opposed to FPSPP1 implemented on 1954 subject essentially to a pay-as-you-go (i.e. no or 0% funding) financing policy. FPDPP1 deals with pension accruals from 1954 to March 31, 2000 while FPSPP2 deals with post March 31, 2000 accruals. Actually, the sole difference between FPSPP1 as at March 31, 2000 and FPSPP2 as at April 1, 2000 is the financing policy, i.e. no funding vs. 100% funding, respectively. Some FPSPP2 provisions were amended/modified on a few occasions after April 1, 2000, but the financing policy was not.
  • CPP1, i.e. the existing 1966-implemented CPP, which is subject to a partially funded (about 20%) financing policy since 1998, as opposed to CPP2, which was approved in 2016 to become effective in 2019 subject to a fully funded financing policy.
As Malcolm’s comments on the FPSPP pertain as a whole to both FPSPP1 and FPSPP2, while mine pertain exclusively to FPSPP2, Malclom’s conclusions are not relevant as a rebuttal of my observations. By converting FPSPP1 into FPSPP2 on April 1, 2000, the federal government was successful in correcting as well as anyone could have the inadequacies of the pay-as-you-go financing method. Once you recognize being responsible of a problem (FPSPP1 paygo financing) that you have caused by addressing it as well as could be (FPSPP2 full funding policy), you have to live with the unchangeable reality and stop repeatedly conveying that you regret having caused it, as no more can be done to address it.

If CPP1 actual demographic and economic experience had in aggregate been less favourable than its statutory assumptions from 1998 to now, the 9.9% contribution rate would have had to be increased and/or the prescribed level of pension indexation benefits would have had to be decreased. Neither happened because experience was better than the well-set assumptions (investment return being the most impacting one) and because of the quality of CPP1’s designed and explicit partial-funding financing policy. True, in case of extreme financial hardship, e.g. no investment earnings at all, the CPP1 9.9% contribution would need to be increased by no more than 2% of contributory earnings. Besides, if this was to happen, it would be a problem, even if CPP1 is only 20% funded, not only because any big or small, temporary or permanent, increase in the contribution rate is the main reason why we are having that “DB Pension Plan Model Failure” discussion but also because CPP1 would not be meeting the objectives of its 20%-partial-funding financial policy. 

The failure to meet the objective(s) of a DB pension plan’s financing policy is always a problem irrespective of the targeted funding level. If CPP1 had been implemented on a full funding basis in 1966, the contribution rate would be no more than 6% rather than the actual 9.9%. Any increase beyond 9.9% would be most unwelcome as a major financial problem for employers and workers and a major political problem for the federal and provincial governments.

I thank Bernard and Malcolm for the comments and insights but think they need to be locked up in a room with some beer and hash it out among titans of actuaries.

As far as Bob Baldwin, I thank him for sharing his great paper on DB and pension plan design, it's a great contribution to an important and ongoing debate.

Lastly, take the time to read this Benefits Canada article which explains as membership in traditional defined benefit pension plans continues to decline, it’s becoming more common to see “contingent” plans — including target-benefit, shared-risk, multi-employer and jointly sponsored — which require members to take on at least some of the risk that benefits may or may not meet expectations.

Below, an older clip from my 2014 comment on the brutal truth on DC plans which I also posted in an even older (2012) comment on America's 401 (k) nightmare. As the default retirement plan of the United States (and Canada), the 401(k) falls short, argues CBS editor-in-chief Eric Schurenberg. He tells Jill Schlesinger why the plans don't work (2009).

Update: Wayne Kozun, former SVP at Ontario Teachers' and now CIO of Forthlane Partners, shared this after reading this comment:
I thought that Baldwin's post was quite good, but I think that "cross-subsidy" is not a good term to use when discussing the pooling of longevity risk as it has negative connotations. Maybe it is a technical term used by actuaries but using this term with the broader public might cause people to get angry and feel that they are being ripped off by this aspect of DB plans. (As an aside -I feel that it is one of the key benefits of DB over DC and there should be more work done to figure out how to bring longevity risk sharing to the DC world, perhaps using aspects of a tontine.)

I would prefer a term like "mortality risk sharing benefit" or talking about the over-saving penalty that is inherent in DC plans as everyone has to save enough to live to 100 when only a few of us will live that long.

There was a good article on this in the January 2007 issue of the FAJ by Waring and Siegel, Don't Kill the Golden Goose! Saving Pension Plans, that stated:
"A back-of-the-envelope estimate shows that from sharing longevity risk alone, a given dollar amount of retirement benefit is 35 percent cheaper to provide through a DB plan than through a DC plan."
Isn't that a good thing rather than a cross-subsidy?

The above article is: (this issue of the FAJ has several good articles on the DC vs DB debate and still holds up 12.5 years later).
I thank Wayne for sharing his insights and bringing Waring and Siegels' paper to my attention.

Malcolm Hamilton, a retired actuary, sent me his thoughts on Wayne's comment and the Waring and Siegel paper:
Let's start with the quote from the Waring and Siegel paper.
"A back-of-the-envelope estimate shows that from sharing longevity risk alone, a given dollar amount of retirement benefit is 35 percent cheaper to provide through a DB plan than through a DC plan."
Wayne suggests that this be viewed as a good thing and not as a "cross subsidy". To me, it is simply a misinterpretation of what the "back of the envelope calculation" demonstrated. To see this, read the description of the calculation appearing on page 33 of the paper. Here is an abbreviated summary:
  • The authors estimated that the cost of a life annuity paying $100,000 per annum was $1,180,000 in 2007 based on the 4.8% yield on long term U.S. Treasuries at that time.
  • The authors then estimated that the cost of constructing a bond portfolio that would deliver the same $100,000 per annum for a term certain of 40 years would have been $1,802,000 using the same 4.8% interest rate. This, they argued, is what a DC plan member needed to do to be sure of having a $100,000 income for life.
  • The authors then concluded:
    • "Compared with what is needed to provide this income to an unannuitized retiree, a DB annuity requires 35 percent less accumulated savings to providethe same guaranteed lifetime stream of payments!"
This statement is either misleading or untrue depending on how you interpret a "lifetime stream of payments". The two streams of guaranteed payments are clearly not the same. The annuity delivers $100,000 per annum for life, roughly 18 years on average based on the quoted price, and then it stops, leaving nothing for a surviving spouse, or children or the member's estate. The bond portfolio delivers $100,000 per annum for 40 years no matter when the member dies. The second stream costs more because it delivers more - 40 years of income versus an average of 18 years with the life annuity. Both options deliver the same 4.8% internal rate of return, hence both are fairly priced. The alleged 35% discount is not a discount. The DB members pay less because their survivors receive less.

What were the authors trying to say? They meant to say that if we look at the member's pension very narrowly and attach value only to amounts paid to the member, not to the member's survivors or estate, then the DB plan has a 35% cost advantage because the average member will collect the same $100,000 for life from each of the streams. This 35% is not a discount. It is the present value of the benefits assumed to have been foregone by DB survivors when compared to the benefits enjoyed by DC survivors. It has nothing to do with sharing longevity risks. It is about shifting the economic value of the benefit from survivors to pensioners. There is no magic here. No free lunch. No additional value added. The DB option discussed in the paper leaves nothing for survivors. The DC option gives survivors 35% of the total amount saved.

One might argue that paying more to members and less to survivors is a good thing, but it turns out that the DC member could have done this anyway. To quote the authors,
"We should note that, in theory, an annuity option could be added to a DC plan to capture that roughly 35 percent cost advantage. In practice, however, most DC plans just don’t foster sufficient accumulation for the benefit of annuitization to be meaningful even if it were offered."
So even if the goal is to die broke and leave nothing to your survivors, it turns out that you can do that in a DC plan - but you probably won't be able to fully exploit the opportunity because you won't have accumulated $1,180,000 like they do in DB plans.

In the 1970s, Canadians with RRSPs needed to withdraw their money by 71 (and pay tax on it) unless they bought a life annuity. This was so unpopular that the law was changed and RRIFs were introduced as an option for those who were prepared to forgo the "economies" associated with pooling mortality risk. It turns out that very few Canadians will voluntarily buy annuities for a variety of reasons. They don't like the illiquidity, or the irreversibility, or the inflexibility. They don't like locking-in low returns, or the assumption of near perfect health built into the annuity rate. Finally, and perhaps most importantly, they don't mind leaving their unspent savings to their spouses and children and grandchildren.
I thank Malcolm for his excellent and critical insights.

Bernard Dussault, Canada's former Chief Actuary, shared this with me:
Malcolm’s observations make much sense, I do not disagree. Obviously DB plans are not perfect.

However, DB plans are generally superior to DC plans from the members’ perspective for at least the following important reasons:
  1. DB retirement benefits are paid until the death of the pensioner, and many DB plans include survivor benefits. DC plans carry the serious risk that the pensioner survive his/her pension fund.
  2. Upon retirement, the pensioner is responsible for the investment returns on a DC fund, while under a DB plan higher returns are obtained because the DB plan fund is generally bigger, more diversified than the individual DC fund, and contrary to DC investments, DB investments are generally pooled and looked after by experts who thereby generally experience higher investment returns.
I thank Bernard for his wise comments and I have advocated that CPP benefits should carry survivor benefits for a spouse or even a child in cases that child is disabled.

Lastly, Bob Baldwin shared this with me:
It is not entirely clear to me whether Malcolm is objecting to the alleged 35% magnitude of the advantage that has been cited, or to the suggestion of any advantage. Personally, I am suspicious of the 35% number because I am skeptical that people managing the rundown of their assets routinely manage the rundown with an end date of age 105 in mind. Having said that, I don't know of empirical evidence on the point. (Trying to determine the implicit life expectancy in patterns of annual withdrawals would be an interesting research project). While I am suspicious of the the 35% advantage, I suspect that fears of running out of money cause individuals to run down their assets at a slower pace than is warranted by their life expectancy. Thus I suspect there is some DB advantage.(Adverse selection in annuity markets suggests that large populations of people have some dim sense of their life expectancy).

But even if it is true that "on average" there is a DB cost advantage of some magnitude, we have to be careful in what we say about individuals in DB and DC arrangements. The cost advantage of DB stems precisely from the fact that plan members who die young will over-save for their own retirement and the "over-saving" will be used by someone else. (It is interesting that many DB plans constrain this cross-subsidy by providing periods of guaranteed payments which may suggest a limit on people's willingness to be a source of subsidies). It also happens in DB plans that whole cohorts of plan members may experience mortality improvements that were greater than anticipated. When this happens - all else equal - younger cohorts will save not only to provide for their own lifespans but to make up for under-saving by older cohorts.

Malcolm has suggested that savings in excess of those needed for the life span of a DC plan member are compensated by payments to surviving family members. Maybe. But I am not sure that this will be a strong motivator for people saving for retirement. A few years ago I was working with some survey data created by the Canadian Public Pension Leadership Council on retirement and pension issues facing the still employed population. I was surprised at the lack of weight respondents gave to bequest motives as a reason for wanting more than enough to maintain their standard of living.

I would add two general points. Although the cross-subsidies discussed here relate to longevity, DB plans have many other cross-subsidies as well. Second, there is nothing inherently wrong with cross-subsidies. But, if they are not identified and measured, it is hard for people to make an informed choice about which ones are fair and reasonable and which ones aren't.


P.S. Your recent blog on the use of an 8% discount rate in US state and local plans reminded me that if you are not aware of the public pensions database on the website of the Center for Retirement Research, it is worth a look. It is very good. I have wished for years that we had a comparable one here in Canada.
I thank Bob for his excellent insights and replied:
"The cost saving on DB plans isn’t just stemming from some members over-saving, subsidizing others who live longer. The cost savings Jim Keohane and others refer to are due to pooling of resources and expert investment strategies across public and private strategies. I think it’s an important distinction. I am aware of the database you’re referring to but to be honest, haven’t looked at it in a while."
I also agree with him, we need a similar open database in Canada to track what Canadian public pensions are doing.

Before closing, Malcolm did come back to me to clarify something:
Two elaborations...
  • I realize that most DB plans include survivor benefits however the 35% "discount" to which Wayne referred did not allow for this. If you add a 100% survivor benefit the 35% turns out to be more like 15% to 20%.
  • I did not claim that bequest motives are an important motivator of retirement savers. I do not believe that they are, or that they should be. I simply observed that buying annuities to prevent survivors from inheriting your unused retirement savings has been an ineffective motivator of prospective annuity purchasers. Canadians may not save heavily to enrich their survivors but they equally do not go out of their way to deny their survivors an inheritance. They appear not to like the idea that pension plans (or insurance companies) might be enriched by their untimely demise.
DB pension plans cost effectively deliver retirement incomes. I do not dispute this but they are not 35%, or 20%, more cost effective than DC plans. Any estimate of the advantage DB plans enjoy should not proceed from the assumption that money left to survivors matters not to plan members.
I thank all the contributors who took their time to share their wise insights on pension plan design.