Help the Government of Canada organize its website Canada.ca!
Complete an anonymous 5-minute questionnaire. Start now.
Archived - Efficiency and Effectiveness of Savings Instruments Design
Archived information is provided for reference, research or recordkeeping purposes. It is not subject to the Government of Canada Web Standards and has not been altered or updated since it was archived. Please contact us to request a format other than those available.
-- by James Davies
Department of Economics
University of Western Ontario
December 3, 2009
Paper prepared for the Research Working Group on Retirement Income Adequacy
James B. Davies
Department of Economics
University of Western Ontario
London, Ontario N6A 5C2
Table of Contents
Current Participation in Pension and Saving Plans in Canada
Different Plan Types
Substitution Across Plans
Savers' Risk in Different Savings and Pension Plans
Other Impacts of Savings Instruments and Design Parameters
This paper reviews the efficiency and effectiveness of current tax-sheltered saving plans toward to retirement income in Canada. Efficiency is referred to if these plans create any kinds of distortions in individuals' saving behaviour. Effectiveness is referred to if these plans are effective in encouraging people to save adequately for retirement. The paper starts with a review of current participation in pension and saving plans in Canada (such as RPPs, RRSPs, TFSAs), using Survey of Financial Security (2005) and CRA tax data. This paper then discusses the rationale for multiple saving plans and multiple plan types. The author then discusses how various savings plans fit together, how they can be substituted for each other and non-sheltered savings, and how they interact with taxes on public pension income (CPP/ QPP/ OAS) and clawbacks on OAS/GIS. The paper then looks at the implications of savers' risk for plan design and examines the risk characteristics of different saving plans. Lastly, the author discusses the possible impacts of savings plan design on macro-economic stability, productivity, labour mobility and labour market mobility, labour supply and retirement age.
The paper concludes that a small minority appear to save too little in registered savings instruments to allow even 50 % replacement of income in retirement, noting that the difference may be made up by savings and accumulating assets in non-sheltered assets, such as housing, there is likely still under-savings by some groups. The paper also notes that, for low income Canadians, lack of saving for retirement is not generally irrational behaviour, given the high replacement rate for their pre-retirement earnings from public pensions. The paper notes that DC pension plans and group RRSPs appear to be good substitutes, with the latter offering some advantages to employers and employees, in terms of greater flexibility and lower administrative costs.
The paper underlines that both DC and DB pension plans have significant risks, with DB plans seemingly safer in normal times, but possibly having greater "catastrophic" risk. Risk also seems to be associated with the investment profile of DC plans and RRSPs (and the winding of up of DB plans), which, in the current context, are somewhat mitigated by the robustness of housing prices and popularity of less risky asset types.
Finally, the paper concludes that, while the effects of macro-economic stability and labour mobility are difficult to assess, there appears to be some concern regarding the decreased labour mobility associated with the lock-in effect of pension assets and some enhancement of macroeconomic stability associated with the large fraction of private savings that are in pension or savings plans in Canada.
This paper reviews the efficiency and effectiveness of tax-sheltered savings instruments design in the Canadian context. Both employer-based pension plans and private savings plans are considered. The former take the form mostly of Registered Pension Plans (RPPs) while the main examples of the latter are Registered Retirement Savings Plans (RRSPs), Deferred Profit Sharing Plans (DPSPs) and Tax-Free Savings Accounts (TFSAs). Design issues are both changes to these plans and the possible introduction of new plans. "Efficiency" is being used in the economist's sense of avoiding unnecessary distortions in behaviour, in this case mainly savings behaviour. "Effectiveness" means achievement of policy goals.
Not distorting savings behaviour requires making sure that the after-tax rate of return is the same as the before-tax rate of return, and that there is adequate opportunity for people to save in sheltered form. The criterion of equality between before- and after-tax rates of return is clearly met via tax-prepaid savings instruments like the TFSA (Individual Retirement Arrangements (IRAs) in the United States are another example), in which no tax is ever levied on the investment income earned. It is also met by tax-deferred savings instruments like RPPs and RRSPs if the marginal tax rate on withdrawal is the same as at the time of contribution. This is a criterion that is likely seldom met precisely in practice, which has some implications for savings instrument design as we shall see.
Effectiveness is likely to require the achievement of more than one goal. What are the policy goals for savings instruments? They may vary over time or place, including one or more of the following:
- Allowing people to save adequately for defined purposes.
- Encouraging people to save for defined, and/or possibly for undefined purposes.
- Ensuring adequate saving for defined purposes.
Defined purposes for tax-sheltered saving may include education, housing, and medical or long-term care expenses, as well as retirement. Here we are mostly interested in retirement saving, but saving for other purposes may have an indirect effect on retirement income adequacy, so should not be completely ignored.
How does one allow people to save adequately for a particular purpose like retirement? First, tax-sheltered savings instruments remove the taxation of investment returns. Second, they must allow adequate contributions to be made — contribution limits cannot be too low. Third, savers must be allowed to choose investments with a sufficiently high rate of return. While this package will allow people to save adequately, it may not necessarily encourage them to do so sufficiently, and it certainly does not ensure that they will do so.
Encouraging people to save can be attempted through advertising and education, but also through some aspects of savings instrument design. There has been considerable discussion in recent years, for example, of the use of the negative option for participation in voluntary employer-based savings plans. Research shows that a significantly higher proportion of workers will participate if they must opt-out to avoid participation than if they must opt-in. This is why, for example, in the September 2009 savings plan changes initiated by the U.S. administration there are strong incentives for plans to use opt-out rather than opt-in. The issue has been less salient in Canada, where participation is generally mandatory for employees at firms with RPPs, but has come to the fore in the discussion of proposals for the creation of broad new tax-sheltered savings vehicles.
There has been an increase in concern over the last year or two in the situation of Canadians not covered by RPPs who either do not contribute sufficiently to RRSPs or earn sufficient returns thereby to achieve adequate retirement incomes. Ambachtsheer (2009), for example, reports that of 17.8 million workers in Canada, 8.8 million are neither RPP members nor RRSP contributors. It has been asked what can be done to encourage greater participation in savings plans, and larger contributions, by people in this group. Tax incentives could conceivably be introduced, although this might be opposed by those concerned with maintaining strict economic efficiency. But attention has recently turned to the creation of broad new public savings schemes that would provide higher and more secure returns than the typical RRSP and which would encourage participation through opt-out rather than opt-in provisions.1
What should be done if, despite encouragement, not everyone appears to save enough, or to receive a high enough return on their saving, to achieve adequate retirement income? A possible answer is nothing. Establishing whether people are in fact saving enough for retirement is not a simple matter, and one should not jump to the conclusion that savings are insufficient. Very substantial saving takes place in non-financial form, via housing, other real estate, business equity, and other assets. And income requirements in retirement are lower than while people are working, raising children, and paying off mortgages. Pension income and age credits, plus the lack of Canada Pension Plan/Quebec Pension Plan (CPP/QPP) or Employment Insurance (EI) contributions also reduce needed income in retirement. Some careful studies that have taken into account these various aspects have found that on average and overall, people are not under-saving for retirement, as we shall discuss. However, if government is determined to ensure that everyone has adequate retirement income, then it may wish to expand CPP/QPP or make participation in a new broad-based sheltered saving plan mandatory for workers who are not covered by RPPs or able to show adequate voluntary RRSP accumulation.
There are clearly many important design issues for savings instruments. It would be impossible to deal with them all thoroughly in a single paper. Here we will try to survey the current situation in Canada and analyze some options for changes. We will start, in the next section, by looking at current patterns of savings plan participation and some of their implications for retirement income adequacy. In section III we then examine the rationale for having sheltered savings and multiple plan types. In section IV we will discuss how our various savings plans fit together, how they can be substituted for each other and for non-sheltered saving, and how they interact with taxes on public pension income (CPP/QPP/OAS 2) and claw-backs on OAS/GIS 3. Section V looks at the implications of savers' risk for plan design. Finally, section VI discusses possible impacts of savings plan design on the macroeconomy, productivity, labour mobility, labour supply and retirement age.
Information on the participation of Canadians in savings and pension plans comes principally from two sources: Statistics Canada surveys including the Survey of Financial Security (SFS), which is conducted periodically — most recently in 1999 and 2005, and Canada Revenue Agency (CRA) data, the most recent full version of which covers the 2006 tax year. The SFS data mainly cover stocks, while the CRA data cover flows. Also, the two data sources differ in terms of the unit covered — SFS data are for families whereas CRA data are for tax-filers. The two data sources complement each other.4 Here we refer to both sources in order to extract insights about the use of different savings instruments and the possible need/demand for new or enhanced instruments.
Table 1 shows summary information from the SFS for families by the age group of the major income recipient. Overall, 70.6% of Canadian families have either registered savings of some kind (mostly RRSPs) or employer pension plans (mainly RPPs). RRSPs may either be individual or group plans, which have become increasingly popular as a substitute for conventional employer-based pension plans.5 As people age their participation in RRSPs and RPPs goes up. While only 55.3% of families below age 35 have assets of these kinds, by age 55-64, 81.9% do so. In the 55-64 year age group 60.1% have employer pension plans and 69.4% have RRSP-type assets. In other words, a large majority of Canadian families have either RPPs or RRSPs by the time they retire. As the table shows, the mean amount of RRSP-type assets for those with such assets in the 55-64 year old group is $124,500 (the median is $60,000), indicating that the typical family on the verge of retirement has substantial investments in registered saving instruments.
The fact that by age 55 to 64 the great majority of Canadian families have RRSP or RPP assets is an important reality. It contrasts with, and puts into perspective, the observation that about half of all workers at a point in time are not covered by RRSPs or RPPs, which gives a very different, and potentially misleading, impression. The two kinds of observation are reconciled by the fact that the workforce as a whole includes younger workers and workers with temporarily low incomes, who are less likely to contribute to RRSPs for a time but may well make up for that later, and by the fact that many families include more than one earner.
Table 2 shows the asset and debt composition of portfolios for families divided into the less than 65 and 65+ age groups. Here I show mean amounts (across all families) and ratios of those means to after-tax income.6 A family over 65 with average net worth, at $485,148 has significant wealth. This includes a total of $174,437 in private savings plans and pension assets, of which $120,023 is in employer pension plans (mainly RPPs). Interestingly, almost as important are principal residences, worth an average of $151,664. These amounts are shown in an interesting light when we divide by mean after-tax income, which is $35,160 for the families aged 65+. Net worth for the retired group is 1379.8% of after-tax income, or about 14 times as great. For comparison, the ratio is 648.2% for families aged less than 65. Savings plans and private pension assets equal about five times mean income, and housing wealth is about four times income. These numbers indicate that retiree families with average characteristics have impressive private resources. Of course, these are just averages. As we know there is considerable variation, and some families reach retirement with quite small private resources (see e.g., LaRochelle-Côté et al., 2008).
Table 3 shows results based on unpublished Finance Canada calculations of the predicted pre-tax income replacement from public pensions (OAS/GIS and CPP/QPP) for earners at different levels using 2006 tax-filer data, and the saving rates needed to achieve different pre-tax replacement rates.7 These numbers are meant to be illustrative. They consider the case of a taxpayer without dependents and having single marital status who earns at a constant level through the working life. The first column shows the predicted retirement income replacement rate that would come from public pensions alone. We see that at the $20,000 earning level there is a 72% replacement rate — so that private savings or pension assets are not needed to get an adequate replacement rate. As earnings rise, however, the situation changes rapidly. At earnings of $40,000, for example, the replacement rate from public pensions is down to 40% and at $60,000 it is only 28%. This is a reflection of the claw-back of GIS and the restricted scale of Canada's contributory public pensions (CPP/QPP), which provide only 25% replacement up to about the average industrial wage.
The next three columns of Table 3 show the calculated RRSP/RPP contribution rates out of earnings that would be required to achieve the different target pre-tax replacement rates. A 3.5% real rate of return is assumed. The final column shows the average actual contribution rate. Comparison of the 70% contribution rate column with the actual contributions shows that, except at the highest income levels, average contributions are adequate to provide 70% income replacement. Recent work has suggested that 70% is an unnecessarily high pre-tax replacement rate for most people, however. This is because in retirement income taxes are lower, CPP/QPP and EI contributions are absent, people reduce expenditures by substituting time for goods, work or employment expenses are gone, mortgage payments are smaller or absent, and children have left. Putting this all together, U.S. studies have found that pre-tax replacement rates of 50% to 60% are typically sufficient to maintain living standards (see Brady, 2008, and Kotlikoff, 2008. Baker and Milligan (2009) argue that the situation is similar in Canada, and Ambachtsheer (2009) argues that 60% is an appropriate average replacement rate in Canada. We may take it that a 60% pre-tax replacement rate is likely sufficient for most people, but should keep in mind that since the reduction in income taxes on retirement is less significant at low income levels, a target of 70% replacement for the lowest income group in Table 3 is more judicious.
If the target replacement rate above the $20,000 income level is set at 60%, average contributions appear more than adequate up to earnings of $100,000, and only fall significantly below the required level at earnings of $150,000. But even the latter shortfall may not be problematic, as income-tax progressivity is sufficiently strong that a pre-tax income fall of even 50% on retirement for someone previously earning $150,000 implies a smaller proportional after-tax income drop than would be true for someone at an average earnings level. If the target replacement level is 50% average contributions are adequate even at the $150,000 earnings level.
Information about average replacement rates needs to be supplemented by some distributional analysis. Table 4 reports my calculations, using Finance Canada data, of the fraction of tax-filers at the different earnings levels who appear to be contributing too little in 2006 to achieve the different target replacement levels when only RRSP/RPP savings are considered. Results are shown for both the 25-44 year age group ("the young") and for those aged 45-64 ("the middle-aged"). It is important to note that tax-filers do not necessarily contribute at a steady rate throughout the working lifetime. Hence, some of those who apparently contributed "too little" in 2006 will compensate for that by contributing "too much" in another year.8 The tax system allows for this by giving an indefinite carry-forward of unused RRSP contribution room.
Table 4 has no entries at the $20,000 earnings level — no one at this level can be contributing "too little" since public pensions themselves provide an estimated 72% replacement rate (table 3). But the situation changes rapidly. At $40,000, 60% of the young do not contribute enough to get 70% replacement from RRSP/RPP saving alone, and even if the replacement goal is only 50%, 36% of these tax-filers would not achieve the desired target, according to these calculations, via RRSP/RPP saving alone. For the older group, fewer appear to be falling short, but at earnings of $40,000 we have 47% contributing less than would give 70% replacement and 32% too little for 50% replacement - - again taking only RRSP/RPP saving into account. In view of the evidence that a 70% replacement target is excessive for most people, it is best to focus on a lower replacement rate. Taking 60% replacement as the target, we see that 49% of the young and 36% of the middle-aged at earnings of $40,000 appear to need more than their RRSP/RPP saving to generate the desired retirement income. As earnings rise from $40,000, we at first find that the fraction in this category declines. It hits bottom at earnings of $80,000, with 37% of the young people and 26% of the older group having RRSP/RPP contributions that alone will not achieve the 60% replacement target. Then the fraction begins to rise quite steeply, to 66% for the young and 63% for the middle-aged at earnings of $150,000. Summing up, there would seem to be many people who need to supplement their RRSP/RPP contributions to achieve satisfactory retirement income. For many people, who may be investing in other financial assets plus housing, durables, real estate or business equity, this is likely not a problem. For a minority, however, there may well be under-saving.
It is also worth noting that t o some extent, variations in the apparent adequacy of RRSP/RPP contributions at different earnings levels may be the result of a high propensity to save out of transitory income. Average earnings in these data are around $40,000. As earnings rise above that level we likely get an increasing fraction of earners with positive transitory earnings, and therefore higher contributions. Above $80,000 this explanation does not work, of course, as the fraction with contributions insufficient to achieve replacement income begins to rise. Contribution rates do rise at these higher income levels, but not enough by themselves to fill the widening gap between the income replacement provided by public pensions and a 60% target replacement level.
Since $40,000 is close to average earnings, the effects of transitory earnings on average contributions are likely small at that earnings level. (The number of people with positive transitory income at this middle level is likely similar to the number with negative transitory income.) It is therefore interesting to focus on what the rates of lower contributions at this earnings level may tell us. Using the 60% replacement rate as a benchmark, 49% of the young appear to contribute too little to hit the retirement income target via RRSP/RPP saving alone. But the "flip side" of this is that 51%, or about the same number, are contributing too much. If people contribute at constant rates from year to year, it could be that some of them have a problem of insufficient contributions. But if they sometimes contribute too little and sometimes too much, there may be little wrong here. We know that some people are indeed contributing at a temporarily high rate in any given year. In 2006, of those at $40,000 earnings, 7% of the young and 13% of the old contributed more than 18%, which cannot be a permanent contribution rate since the maximum that can be contributed is 18% of (prior year) earnings. Thus, 49% is a generous upper bound on the fraction of young people at the $40,000 level who may be contributing too little, and likewise for the 36% number for the middle-aged. This point is strengthened when we consider that many people are also saving in the form of housing equity in these age ranges and some are making substantial non-registered investments e.g. in real estate, business equity and so on. There are people who really do contribute too little, but it would appear that they are in a minority.9
Table 5 shows the distribution of unused contribution room as a percent of earnings, at different earnings levels in 2006, again for the young and the middle-aged. This is important information since if, e.g., contribution limits to RRSPs, RPPs, or close substitutes are increased, people with unused contribution room should not respond — their savings will be unaffected. More subtly, concentrations of people who are "under using" RRSPs at certain ages or income levels may suggest something about the constituency for new or enhanced plans, like the TFSA or the new broad-based sheltered saving plans suggested e.g., by Joint Expert Panel on Pension Standards (2008) or Ambachtsheer (2008, 2009), that differ substantially from RRSPs/RPPs and may perhaps be of special interest to these groups.
We see in Table 5, first, that, overall, the % of tax-filers who have literally zero contribution room is small. It rises with income, but even at earnings of $100,000 this fraction is only 14% for the young and 22% for the middle-aged. On the other hand, there is a significant number of people who have "almost zero" unused contribution room. One can add those with exactly zero to those with 0% to 5% of earnings in unused contribution room, to get an idea of how many have are "effectively" constrained. This yields higher figures — 10% for the young and 19% for the middle-aged at a $60,000 earnings level and 27% and 40% for the respective groups at $100,000 in earnings.
The variation in unused contribution room with earnings shown in Table 5 is also striking. At earnings of $20,000 and $40,000, the incidence of zero unused contribution room is very small. Even applying the idea of an "effective" lack of contribution room, the highest incidence we see is 12% for middle-aged people at the $40,000 level. And the fractions with contribution room exceeding 100% of earnings are large — 55% and 61% of the young and middle-aged at the $20,000 earnings level for example. These figures suggest that there is unlikely to be much response at these earnings levels to higher contributions limits on RRSPs, RPPs or similar instruments. On the other hand, some have felt that there may be some interest here in TFSA type plans that would allow good returns on saving with zero impact on the GIS claw-back. Where there is an apparently important constituency for higher RRSP/RPP contribution limits it is at the higher income levels, although even there we see a significant fraction with substantial unused contribution room.
Rationale for Different Plan Types
Why should there be more than one kind of tax-sheltered savings or pension plan? One reason is that some people want to have an inescapable commitment to retirement saving. Saving regularly and adequately for a distant objective like retirement entirely on one's own is tough. It requires discipline and self-denial, as well as the cooperation of all those making significant spending decisions in the family. So there is a need and demand for pension plans, distinct from that for saving plans. Bolstering this are the advantages of the lower management fees and perhaps superior investment decisions in a group plan — although those are increasingly accessed through group RRSPs.
RPPs clearly do not exhaust the need for tax-sheltered saving vehicles. People need to save for a range of purposes other than retirement, including vacations, education, buying houses and consumer durables, health care, and long-term care. Many feel that such saving should be tax-sheltered, at least within limits. All these saving purposes could be accommodated by a sufficiently flexible type of plan, and the recently introduced TFSA is indeed such an instrument. However this is a somewhat unusual development in Canada. While early withdrawals can be made without penalty at any time from non-spousal RRSPs and after two years from spousal RRSPs (which contrasts with U.S. practice, where early withdrawals are subject to penalty), we have a history of creating special provisions within RRSPs or even special plans for particular purposes — for example Registered Home Ownership Savings Plans (RHOSPs), abolished in 1985, and the more long-lived Registered Education Saving Plans (RESPs) that are still with us. There is a desire to reward or encourage saving for particular purposes. But even if that were not the case, there would be a rationale for having more than one kind of sheltered saving plan.
Economists have long argued that, as long as personal income tax has different effective marginal tax rates at different income levels, two kinds of savings plan are needed for efficiency 10. One is the familiar RRSP-style tax-deferred plan under which contributions are tax deductible, investment returns accrue tax free, and withdrawals are taxable. But this type of plan, by itself, does not guarantee that the after-tax rate of return will be the same as the before-tax rate of return, and it does not allow full income averaging. If the marginal tax rate in retirement is expected to be higher than one's current marginal rate, for example, the after-tax return will be lower than the before-tax return, which causes a saving disincentive. A second type of plan, of the "tax prepaid" variety, is needed to allow undistorted saving, and full income averaging, in this kind of situation. This has been recognized in Canada through the creation of TFSAs, which allow non-deductible contributions of up to $5,000 per year to an account on which investment income and withdrawals are not taxable.11
There is feature of the tax treatment of public pensions in Canada that makes it useful to have tax-prepaid savings instruments like TFSAs in addition to RRSPs. This is that the GIS and OAS claw-backs raise the effective marginal tax rate (EMTR) on RRSP withdrawals and RPP-based pension income. In the case of the GIS claw-back, this EMTR is a minimum of 50%, which greatly reduces the return on RRSP saving, and can make it financially inadvisable for people with expected non-OAS/GIS retirement income up to about $15,000 (see e.g., Shillington, 1995, 1999). For OAS there is a 15% claw-back on net income over $66,335, which raises the EMTR accordingly until complete phase-out occurs at income of $107,692. For the many people affected by this phase-out, and also to an extent for those affected by the GIS claw-back, the TFSA should be helpful.
Differences Between Existing Plan Types
Below we will consider substitutability between different plan types. If employer-based pensions could be assumed to be of the RPP type and taken as a given, and if saving plans were of the two ideal types (tax-prepaid and tax-postponed) discussed above then substitution between plans would not be difficult to analyze. However, real-world plans are not of the ideal type — if only, as indicated above, because they are subject to contribution limits.
While TFSAs closely approximate the ideal tax-prepaid plan, aside from their contribution limits, the same does not hold for RRSPs. It is true that early withdrawals are allowed without penalty (except for spousal RRSPs in the sense that there is a two year wait before the spouse can withdraw). And the carry forward of contribution room reduces the distortions that would otherwise be caused by annual contribution limits considerably. These aspects lend flexibility to RRSPs and facilitate consumption smoothing over the working life plus saving for non-retirement purposes. However, if the saver's objective is to provide retirement income, or to save for even longer-run objectives — such as to self-insure against large health and long-term care expenses at the end of life, or to provide bequests, things are more restricted.
By age 71 RRSPs must be converted to a Registered Retirement Income Fund (RRIF) or used to purchase an annuity. A RRIF has a schedule of minimum annual payouts. Thus, while RRSP holders can take their funds out, and have them taxed early, they are not allowed to leave their funds in an RRSP indefinitely, as required in the ideal tax-postponed savings vehicle of economic theory. Until the advent of TFSAs there was no way that the compulsorily withdrawn funds could be deposited in another tax-sheltered savings instrument. Even with TFSAs the amount that can be sheltered in this way is quite limited. Hence there is a pro-consumption, anti-saving distortion created by the compulsory withdrawal of RRSP funds in retirement.12 This is inefficient from the economist's viewpoint. The policy intention is presumably to protect against too large a revenue loss from RRSPs, and perhaps also to help to ensure that RRSP funds retained into retirement are indeed used for retiree's consumption.13 Thus, these restrictive provisions may be effective in achieving policy objectives while at the same time being inefficient.
New Plan Types
As mentioned earlier, proposals for the creation of new kinds of pension and savings vehicles have recently been made in Canada. There are two broad types. First, some have called for an expansion of CPP/QPP. Most simply that could be done by increasing contribution and benefit rates. This has the advantage of ensuring higher public pensions for a very large portion of the population. However, it has the disadvantage of forcing increased contributions to a pay-as-you-go pension system that offers a relatively low effective rate of return and does not provide as tight a relationship between contributions and future benefits as in a DC pension plan or RRSP.
The other major suggestion is for the creation of a new broad-based public saving plan that would feature voluntary participation on an opt-out basis for all employees, and participation for the self-employed as well — perhaps on an opt-in basis (as in the "ABCP" plan advocated in Joint Expert Panel on Pension Standards, 2008). One might ask how such a plan would differ from existing RRSPs offered by financial institutions. A key part of the answer from the viewpoint of the advocates of these plans is that the opt-out provision is likely to ensure higher participation than one sees with RRSPs. Proponents also hope that administration fees would be smaller, partly because of economies of scale in administering large plans and partly because of other cost-savings including lower sales and service costs - - the latter due for example to a more restricted range of investment options leading to fewer transactions to discuss with, and implement for members. Other distinguishing features would depend on plan governance and management, or specific details of investment options — for example the Ambachtsheer (2008, 2009) "CSPP" proposal would allow gradual advance annuitization of a member's funds, generating more retirement income security.
This section discusses the substitutability of different kinds of savings plans, as well as the possibilities for substituting between non-sheltered and sheltered holdings. Why is substitution important? First, if substitution is too easy it can sometimes defeat the purpose of sheltered saving plans. People may merely transfer existing sheltered or non-sheltered assets into the new or enhanced plan. The net result can be a small or zero net increase in saving, along with reduced revenue. Second, we have different kinds of plans in order to accomplish different policy goals. These objectives may be jeopardized either by substitution being too easy (people save via the "wrong" plan) or too difficult (people cannot move their assets into the "right" plan).
There are two important preliminary points to make. One arises because the focus of this study is on retirement income adequacy, not saving. The standard life-cycle model, which is the first tool economists turn to in order to analyze the effects of saving incentives, says that their impact on saving is generally ambiguous. There is a substitution effect toward increased saving, but an income effect towards higher current consumption (and therefore lower saving). The net impact can only be established empirically.
The point that needs to be made here is that while the impact of saving incentives on net saving is theoretically ambiguous, the impact on retirement income is not. For retirement income, substitution and income effects are in the same direction. The substitution effect is toward higher retirement consumption and therefore higher retirement income, but so is the income effect. This means that, if the goal is to raise retirement income, providing savings incentives will definitely work, according to the life-cycle model.
The second preliminary point is that providing tax-deferred saving incentives causes a reduction in revenue in the near term, and it may be a long while before higher revenues resulting from increased investment accrue. Thus, increases in income or other tax rates will be required to offset the loss of revenue. Those tax increases will inevitably have negative income effects, and will tend to reduce planned retirement incomes. So, the largest beneficiaries of savings incentives may have their retirement incomes increased, but others may see a decrease. Whether this is judged desirable must depend on who gets the increase and who the decrease. Thus the distribution of impacts needs to be kept in mind.
Empirical Evidence on Substitution Between Non-Sheltered Saving and Sheltered Savings Plans
There has been considerable empirical work over the last 25 years, mostly in the United States but also to an extent in Canada, aimed at finding out what portion of plan contributions represent net savings. (See Bernheim, 2003, for a survey of the U.S. evidence.) The earliest contributions to this literature found large positive effects of IRAs on net private saving in the United States (Venti and Wise, 1986, 1990). This early work was criticized for not dealing adequately with econometric issues arising from the endogeneity of plan participation and heterogeneity in tastes. Papers quickly followed that attempted to deal with these issues in a variety of ways — by controlling for selection effects (Gale and Scholz, 1994); studying "differences in differences" in saving over time between households that were eligible or ineligible to participate in 401(k) plans (Poterba, Venti and Wise, 1995; Engen and Gale, 1997); by comparing earlier cohorts with little access to sheltered saving with later cohorts; and through the use of a statistical technique called propensity score matching (Benjamin 2003). While these studies sometimes found zero effects, typically they found small positive impacts. Those who believed the effects were larger fought back — admitting that there were difficulties with the early work but identifying remaining statistical problems in their critics' methods and trying to show through further statistical work and examination of saving trends over time that the introduction of 401(k)s and IRAs had a large positive effect on savings. After reviewing the studies on both sides of the debate for 401(k)s, Engen and Gale (2000) tried to correct for a range of remaining statistical issues and issues of interpretation. They concluded that about 30% or less of 401(k) contributions represented net saving. It appears that the majority of economists in the United States would now accept that the effects of sheltered savings plans on net private saving are likely small but positive and not negligible.
A further insight from the U.S. literature is that higher income groups are more likely to exhibit greater substitution of other assets into sheltered form than middle or lower income groups. This result has a reasonable theoretical explanation. Higher income people are more likely to be "limit contributors", and if so do not experience any impact on their marginal rate of return to saving as a result of the availability of the sheltered saving plan. Hence they are not subject to the same substitution effect acting to encourage increased saving as are non-limit contributors (see e.g., Bernheim, 2003).
Empirical work on the effect of saving plans in Canada has developed along with that in the United States. Venti and Wise (1995) found results using Canadian data similar to those in their U.S. studies. Considerable attention was paid on both sides of the border to a couple of other early studies using Canadian data, in part because in the 1980s contribution limits to sheltered savings plans were higher in Canada than the United States. Using time series data, Carroll and Summers (1987) seemed to find evidence that Canada's more generous savings plans had caused its personal saving rate to rise relative to the United States. And Engelhardt (1994) compared the impact of the 1985 cancellation of the Registered Home Ownership Savings Plan (RHOSP) on saving in 1986 compared with that in 1982 by RHOSP contributors and non-contributors. He found that saving by both groups declined in this period, but that the fall was larger for the former RHOSP contributors, suggesting that the plan had earlier buoyed their saving considerably. RHOSPs provided a greater incentive to save than RRSPs, however, since withdrawals were not taxed. This made it difficult to generalize from Engelhardt's results.
Since the early studies, there have been some papers that use more advanced techniques with Canadian data, trying to overcome the endogeneity and heterogeneity problems mentioned above. Milligan (1998a), for example, used propensity score matching, and found that between about 20% and 50% of RRSP contributions represented net saving, looking across three different Surveys of Family Expenditure and considering four different definitions of saving. There has been less work on the topic in Canada than the United States, however, in more recent years, perhaps due to less data availability.
Associated with the empirical work for the United States and Canada there has been discussion of why a positive effect on net saving is observed. The starting point for this discussion is that the theoretical impact of savings incentives in the life-cycle model is ambiguous. There has thus been a search for alternative explanations. Clearly, saving involves discipline and self-control. Merely making bank deposits or purchasing investment products does not guarantee saving will be high enough. One must also avoid over-spending, running up consumer debt and so on. Some economists and psychologists have studied how people try to commit themselves to saving. This can be done e.g., through contractual saving, but may also be assisted by psychological mechanisms such as "mental accounting". Savings plans may open a new mental account for some people. It has also been suggested that they may have a financial education effect. As pointed out by Milligan (1998), it has been found that people who contribute to RRSPs are very likely to continue doing so. Finally, if people are myopic they may weight the immediate tax deduction on RRSP contributions much more than the discounted future tax liability on withdrawals, and may therefore be more inclined to participate than one might otherwise expect (see Bernheim, 2003).
Simple tax-prepaid plans, such as the TFSA, have very similar characteristics to non-sheltered saving, but offer higher returns. One is therefore likely to see strong substitution effects between non-sheltered financial assets and TFSAs. Some of the people who take advantage of such plans will therefore merely be moving existing non-sheltered assets into sheltered form. This implies a potentially large loss of government revenue. This helps to explain why TFSAs were introduced with a relatively low contribution limit of $5,000.
In order to get some idea of the possible degree of substitution of non-sheltered assets into TFSAs we can look again at Table 5, which shows what fraction of working-age tax-filers were at their RRSP contribution limits in 2006. Including those with very small unused contribution room (less than 5% of earnings) there were 1.8 million tax-filers, or 10.8% of those aged 25-64, in this position. Most of them would have had at least some non-sheltered financial assets. If they all made use of TFSAs, the contributions generated would be very sizeable. To these numbers one can add possible contributions by the approximately 30% of tax-filers who are ill-advised to use RRSPs due to the GIS and other claw-backs. In total, the substitution possibilities, away from non-sheltered and RRSP assets appear considerable.
It is important to note that, over time, the mix of existing assets vs. fresh saving moving into TFSAs should change. When a vehicle like the TFSA is introduced, some people will have a stock of existing non-sheltered assets that they would like to move into the new vehicle. And some may have currently sheltered assets, e.g., in RRSPs, that they would like to move to TFSAs. But over time this desired substitution will be completed for an increasing number of people. Some people will only have $5,000 or less of existing assets that they wish to move into a TFSA. After the first year, any amounts they contribute will represent fresh saving. Others will take longer to move over their existing assets. It may be only a relatively small group of wealthier taxpayers who will never be able to exhaust the possibilities of substituting between existing assets and TFSAs.
There is another sheltered savings plan in Canada that falls roughly into the tax-prepaid category. This is the Registered Education Savings Plan (RESP). Taxpayers may contribute up to a cumulative total of $50,000 per child (there is no longer an annual limit). Contributions are not deductible, but investment income in the plans accrues tax free and if withdrawals are used for educational purposes they are taxed in the hands of the children, who are likely to be non-taxable or in a low tax bracket. This is fairly similar to tax-prepaid treatment. The major differences, compared with a TFSA, are (i) the plans are for a dedicated purpose and are therefore much less flexible, and (ii) contributions are subsidized — at a standard rate of 20% up to $500 per year and at a higher rate for low income people.
RESPs may seem far afield from retirement savings, but have some relevance here. First, the subsidy element demonstrates the use of incentives to encourage people in general, and low income people in particular, to save more for a defined purpose. This provides an interesting precedent for the possible use of incentives to encourage more retirement saving. Second, the amounts that have accumulated in RESPs are quite large, and may have significant effects on lifetime saving plans. The existence of these plans makes the contributors better off over the lifetime and would be predicted to increase their desired standard of living at all stages, including retirement. In terms of substitution, the existence of RESPs is likely to reduce the use of other savings instruments among parents of pre-college students, that is, among taxpayers concentrated roughly in the 25 to 45 year old age group.
There are other assets and debts that effectively received tax-prepaid treatment in Canada. Since interest on consumer and mortgage debt is not tax deductible, when people save by paying down such debt they are, in effect, making tax-prepaid "contributions" to a savings "plan". The interest rate earned by making these contributions can be very high, for example in the case of credit card debt. One would therefore expect most holders of such debt to pay it down first before making TFSA contributions. And while interest rates on lines of credit and mortgage debt are lower, they are still higher than those that can be earned, say, on GICs that might be held in a TFSA. Thus, paying down these debts may dominate TFSA contributions as a form of saving for many tax-filers, particularly the younger and early middle-aged, who have relatively high incidence, and amounts, of debt in these forms.
Paying down mortgage debt can be one way of building up equity in owner-occupied housing, which also effectively receives tax prepaid treatment since there is no tax on the imputed income or capital gains generated by this form of wealth. Recognizing that home equity has this tax treatment is important since, as we saw earlier, it is quantitatively almost as important as pensions and retirement savings in the life-cycle savings of Canadians. Once again, the availability of this form of tax-prepaid saving may reduce the interest in TFSA saving for many young and middle-aged taxpayers. It may also lead to sizeable withdrawals of TFSA funds when people purchase homes. Also note that the latter shift may be greater than, or may come over time to increasingly replace, the use of RRSP funds to aid in home purchase, since TFSAs and home equity, both tax-prepaid savings opportunities, likely have greater substitutability than RRSPs and home equity, which differ sharply in their tax treatment.
Implications of RPP Design
In thinking about substitution possibilities involving tax-deferred plans (mainly RPPs and RRSPs) it is important to recognize some key features of RPPs, and also the differences among them. The major divide, of course, is between Defined Benefit (DB) and Defined Contribution (DC) plans. These categories have major differences in risk characteristics, as discussed in the next section, but they differ in other ways too, for example in how an employee is affected when moving between employers.
When a DB RPP member moves to a new job he/she may suffer an effective loss of pension wealth. A typical DB plan would yield a pension equal to two percent for each year of contributions (up to some maximum, often 30 years) times the average of the worker's best five years' earnings. One way in which benefits may be lost when moving to a new employer is that the pension paid by the old employer, or its commuted value, is based on the earnings over the best five years to that point, which may be less than the best five years' earnings if the worker had stayed until retirement.14 Many older workers have passed through multiple pension plans at different employers, and will receive partial pensions from each. The net result may be inadequate total pension income. Since the incidence of life-long attachment to a particular employer appears to have been declining this point is of increasing importance. Over the lifetime many workers may receive inadequate total DB pension benefits in relation to the contributions they have made. This unfortunate outcome may have something to do with the gradual trend away from DB, and toward DC plans or group RRSPs, although the evidence is that this trend has taken place much more as a result of new firms adopting the latter plans rather than as a result of direct substitution (Luchak and Fang, 2005).
While there may be a risk of the pension plan collapsing and employees losing some part of their pension benefits, under a DB plan there is generally less uncertainty about the amount of the future pension than in a DC plan, contingent on the pension plan remaining in place. This may tend to differentiate the RPP more sharply from an RRSP from the viewpoint of the saver. In other words, a DB RPP may be less substitutable for an RRSP than is a DC plan. Many people are willing to take on more risk for the sake of higher expected returns over the lifetime. This means that, all else the same, the average DB member should perhaps, theoretically, contribute more to RRSPs and TFSAs than the typical DC plan member.
While the difference in substitutability between DB versus DC pension plans with private savings plans is worth keeping in mind, there are aspects that may reduce the impacts on behaviour. It may be, for example, that more risk averse people find their way into jobs with DB pension schemes — they become teachers rather than small business owners, say. Also, people who have experience of a DC pension plan, and who have perhaps had to make choices between different investment alternatives in their plan, may learn more about investment and also may become more willing to take on risk. Thus it is conceivable that the spread of DC pensions, rather than having reduced RRSP and TFSA contributions, may have had the opposite effect.
For DC plans we do not expect to see the same implications as for a DB plan when a member changes employers. A typical DC plan will have the employee contributing e.g., 7% to 10% of salary and the employer an additional amount, say 4% to 6%. New employees do not pay at a higher rate, so all else being equal there should be no loss of lifetime income due to pension effects when the employee changes employers. Similarly, group RRSP participants should be more mobile than DB plan members.
New Plan Types
The possibility of substitution between and among pension plans and sheltered savings plans is important when one considers the implications of adding new plan types to our current, already rather rich mix. If schemes like the ABCP or CSPP proposals were adopted, how much of the saving in these new forms would represent fresh saving and how much would be drawn from existing plans?
Our earlier analysis of the incidence and extent of unused RRSP contribution room suggested that merely increasing RRSP contribution limits would be unlikely to stimulate a large increase in contributions, except in higher income groups. The analysis of current retirement income adequacy by Baker and Milligan (2009), LaRochelle- Côté et al. (2008) and others also suggests that there is not a widespread large retirement saving shortfall. If one did not anticipate that the opt-out provision of the proposed new plans would have an impact, these aspects and findings would suggest that adding new tax-sheltered saving plans in Canada is unlikely to generate much new retirement saving. In this scenario, most of the funds contributed to the new plans would likely come from substitution away from other plans. The shift of funds would be motivated by higher promised returns and perhaps a perception of greater security.
Adoption of the opt-out feature in the proposed new savings schemes could lead to more take-up than one would otherwise expect. While there is likely more prospect of an actual increase in retirement saving in this case, it seems likely that there would again be strong substitution from existing plans.
While in the short run the substitution toward the new plan types would likely be mostly from RRSPs, in the long run it is possible that as existing pension plans are terminated, and as new firms enter the marketplace, there would be a gradual decline in the fraction of firms offering pension plans. Instead, workers would accomplish their retirement saving via the new public savings schemes. Substitution of this type would indeed likely be regarded as desirable by many proponents of the new savings schemes. Existing pension schemes are regarded as suffering from many administrative drawbacks, which have already led to substitution toward group RRSPs.
Until quite recently it was believed that the risk experienced by savers in DB plans in Canada was small. Typically in these plans the pension benefit is a fraction of the average best four or five years of earnings, generally the final years of the pre-retirement period. This structure is significant in helping to reduce inflation risk. There is some remaining inflation risk, however, since not all DB pension plans have indexed benefits. In addition to that risk, we should not neglect the possibility of a pension plan being wound up due to the failure of a firm and there being a resulting loss of benefits. While pension plan meltdowns have been prevented at the major automakers and in some other cases in the last year, there have also been some prominent wind-ups in which pensioners have lost substantial benefits.15
As emphasized e.g., by Ambachtsheer (2008, 2009) there are further, long run risks, to DB pension plan members arising from the adjustments to contributions and benefits that occur from time to time in response to the over- or under-funding of plans. When plans are overfunded employers may take contribution holidays, sometimes as a result of the Income Tax Act rule limiting pension surpluses (formerly 10 percent of assets but recently increased to 25 percent16), potentially leading to underfunding when markets decline in the future. But both benefit increases and member contribution reductions may be seen in this situation as well. On the other hand, when plans are underfunded benefits may be reduced and/or contributions may be increased. These changes and the process by which they occur are of course subject to legislative constraints, but they are not under the control of individual members and represent a significant source of risk for pension holders.
DC pension plans are less susceptible to the risk of default, but generally involve some risky investments. Plans differ considerably in the element of investment choice offered to participants. Some plans offer very little choice, while others have a standard-looking default option and a fairly wide range of choices — Canadian, U.S. and offshore equity, money market funds, "target dated funds", bonds and so on. Economists tend to think that people cannot be made worse off by having more choice, but the level of investment education varies greatly across individuals, and it is possible that having these choices may backfire for some people who do not understand the levels of risk in different funds, or who do not pay enough attention to what is happening with their DC plan investments.
The element of risk for DC plan members is somewhat reduced by the tendency for members to shift their plan assets into more secure form as they approach retirement. This is in line with the well-established tendency for older investors to move toward less risky portfolios as they age. On the other hand, most plans are missing a partial deferred annuitization option under which members could gradually move their assets into the secure form of annuities prior to retirement.
We have just been through a striking "natural experiment" illustrating the risks faced by holders of pension and saving plan assets, in the shape of the global financial crisis. Table 6 helps to put some dimensions on the scale of the problem. It shows the values of price indexes for a few assets of key importance for Canadians — houses, TSX stocks, and the S&P 500, at selected dates. The first date shown is June 2005, which is when the most recent Survey of Financial Security was conducted by Statistics Canada. The second date is May 2008, when the TSX reached its peak. Next is February 2009, when the TSX fell to its worst low. Finally we have August 2009, the most recent date for which all the indexes shown in the table are available.
The second panel of the table reflects, first, how wild the swings were in stock market prices over this period. (Note that these are monthly close data, so that the swings would be even larger if taken as of the peak and trough days.) Between June 2005 and May 2008 the TSX went up 48.6%, far outstripping the 7.2% increase in the CPI, and also the 17.6% rise in the S&P 500 index of U.S. stocks over the same period.17 Then came the crash, with the TSX falling 44.8% from the end of May 2008 to the end of February 2009. The S&P 500 performed similarly, with a 47.5% decline.18 Subsequently the markets have of course rebounded to an extent, with the TSX lying 33.8% above its trough at the end of August 2009, and the S&P 500 having risen 38.9%. Interestingly, the rebound in the S&P 500 has not had as large an impact for Canadian investors as U.S. investors, due to the rise in the 14.4% increase in the value of the Canadian dollar over this period. In Canadian dollars the S&P 500 is only up 21.5% from its trough.
Table 6 also shows an index of house prices — the Teranet-National Bank index, which is based on comparing the prices of the same houses sold at different points in time in six of Canada's largest cities.19 While house prices also outstripped inflation from 2005 to 2008, the real increase was about half as great as that of the TSX. And, after May 2008 the decline in house prices was relatively mild. As of August 2009, house prices stood only 4.0% below their May 2008 peak. Although housing wealth is relatively illiquid and price declines in local markets could considerably exceed the national average, the fact that overall house prices declined much les than the stock market is important in evaluating the vulnerability of Canadian households to asset price shocks. As we saw in Table 2, the mean value of principal residences is significantly greater than that of all savings plans and private pension assets for families aged less than 65, and the same is true for principal residences plus other real estate for those aged 65+. Further, on average people have substantial holdings of other non-financial assets, as well as of financial assets, like bank deposits, that are not subject to capital gains or losses. Thus while some individuals with large stock or mutual fund holdings, either inside or outside RRSPs, and DC pension plan participants, suffered devastating losses during the crisis, this is not true for the household sector as a whole.
Davies (2009) investigated the possible size of the loss in net worth for Canadian families resulting from the drop in real asset values from May 2008 to February 2009. The results are of interest in the present context, since they reflect a kind of worst case scenario in terms of risks people run on their private savings and pension assets. The calculations were made under the following assumptions:
- All stocks and mutual funds suffered the same % decline as the TSX.
- All RRSPs, RRIFs, other sheltered savings plans and DC employer pension plans were assumed to have 60% of their assets in stocks or mutual funds (following Table 7 of Statistics Canada, 2006).
- Capital gains and losses on all other financial assets, including bonds, were assumed equal to zero.
- Families kept the same assets and debts in place. Changes in values are entirely the result of price changes.
- Equity in employer pension plans was assumed to be 80% in DB plans and 20% in DC plans,20 following the division in terms of number of plan members observed in 2007 (see Statistics Canada's The Daily, July 4, 2008).
- Houses and other real estate changed price according to the Teranet-National Bank index.
Under the above assumptions, it can be calculated that the mean net worth of Canadian families would have risen 17.5% between June 2005 and May 2008, and then would have fallen by 17.5% from May 2008 to February 2009. On average, then Canadian families would have been "back where they started from" after the market crash. However, there would also have been distributional changes. Those who had more wealth than average in stocks, mutual funds, RRSPs or DC pension plans, ended up worse off than in June 2005, while those with more of their wealth in housing, DB pensions, and other assets with stable or rising values were better off
The simulated differences in impact across families are large. Some families, 14.5% of the population, would have seen no change in wealth due to asset price changes under the above assumptions. On the other hand, 0.7% would have lost half or more of their net worth, 3.7% lost more than 25% and 49.6% lost more than 10%.
The Davies (2009) calculations did not model losses in DB pension plans, and the average decline in pension assets was calculated to be just 7.1%. Making a further adjustment for losses in DB plans would likely add little to this decline. The DBRS study of August 2009 found an average funding ratio of 95% for 70 large private sector pension plans it reviewed (Financial Post, 2009). This represented only a small drop from the 98% funding seen in 2007. Further, it is not appropriate to assume that even this small decline represents a loss in the value of pension rights to DB plan-holders. As long as a firm remains viable and can be counted on to "make good" the underfunding of its pension plan, the value of pension benefits does not vary directly with the value of the funds in the pension plan. The firm remains liable for the promised benefits.
What can be done to reduce the risks to workers and savers involved in their pension plans and individual savings is a question deserving careful study. It can be argued that there should be more stringent standards for the funding of DB plans, but these have become a less important part of the overall savings/pension picture. Hybrid plan members, DC members, and RRSP holders — both group and individual, remain subject to risks that cannot so clearly be reduced through regulation. Members of these plans have some degree of control, of course, and can choose safer options than stocks and mutual funds. However, they have long been encouraged to take risks since it has been felt that this is the route to higher expected returns in the long run. Some of the pitfalls of this approach have now become apparent. They can be offset to some extent if the advantages of shifting portfolios toward more secure options are pointed out more clearly, particularly for those approaching retirement. Also, in the case of hybrid and DC pension plans help can be given by providing the option of partial deferred annuitization before retirement is reached.
As mentioned earlier, the design of pension and savings plans can potentially have an impact on macroeconomic stability. One normally thinks of savings as providing a cushion that may be called on when income is interrupted by unemployment or low business income. There is much evidence that people do indeed use their savings in this way, and it would be very surprising if they did not. The consumption smoothing achieved in this way likely not only helps the individual but also tends to reduce the amplitude of macroeconomic fluctuations. However, if savings are difficult to access, as a result of being locked up in savings or pension schemes that prohibit withdrawals or make them very difficult, this role of savings cannot be played. This is a further argument for the kind of flexibility we see with RRSPs in Canada, where withdrawals are allowed without penalty (except in the case of spousal RRSPs as noted earlier).
There is, however, another effect of private savings and pension plans on macroeconomic stability that is destabilizing. Economic theory predicts that people will respond to losses of wealth by decreasing consumption. As the economy moves into a downturn, the decline in stock values that typically occurs may tend to reduce consumption just at the time when it is most needed to put a brake on recession. If this effect is strong it could conceivably outweigh the helpful consumption-smoothing effect of savings referred to above.
How strong is the impact of asset values on consumption? There has been considerable empirical work on this topic, both in the context of housing prices (see e.g., Muellbauer, 2008, for a summary of the literature) and stock market prices (see Dynan and Maki, 2001, and Pichette and Tremblay, 2003, for summaries). There is fairly firm evidence, and consensus, that consumption does respond positively to house prices, but effects of stock market prices are less clear. Pichette and Tremblay (2003), for example find a distinct impact of house price on consumption in Canada, but only weak evidence of an impact of stock prices. In the United States traditionally econometric studies found that consumption would rise between about three and seven cents per dollar increase in stock market wealth. Recent studies appear to make this less certain. Poterba (2002), for example, reports that the increase is between one and two cents, and Starr-McCluer also finds little impact using data from the University of Michigan's SRC Survey of Consumers. On the other hand, using the Consumer Expenditure Survey Dynan and Maki (2001) find an impact of between 5 and 15 cents.
The strength of the stock market wealth effect on consumption depends in part on what saving instruments or pension vehicles people are using. CPP is now partially funded in Canada. When the value of its investments goes up, however, there is unlikely much if any direct effect on consumption. Promised benefits are unaffected, and while people often claim they are unsure that the plan will not go bust before they retire this does not seem to be a genuine concern. With safe DB RPPs the logic is similar. On the other hand, one could expect some impact on consumption from stock market fluctuations that affect the value of DC RPPs, group RRSPs, and individual savings plans. The impact on aggregate consumption should depend on the relative importance of these instruments compared with CPP and DB RPPs, but also on the mix of DC RPPs versus RRSPs and like plans. Due to borrowing constraints, some of those whose lifetime wealth rises as their DC RPPs are enriched by stock market growth will not be able to spend more in reaction to those gains. This means that when the markets are moving up and the economy is likely also expanding, this will be reinforced more the greater the importance of RRSPs relative to DC RPPs.
Labour Market Mobility and Productivity
As discussed earlier, under existing DB plans there is an incentive effect that encourages people to stay with their current employer, in order to avoid the effective loss of future pension benefits that typically comes with departing from a job with a DB plan. This tends to reduce labour market mobility, which could exert a negative effect on productivity. On the other hand, as emphasized by Luchak and Fang (2005), in a stable industry with mature firms the fact that DB plans encourage long-run attachment, and perhaps also loyalty to the firm and the development of firm-specific human capital, may tend to increase productivity. This illustrates the idea that what may be best for productivity is for the pension or savings plan used by a firm to be appropriate for its industry, market conditions, and business strategy. Emphasis on the need for greater cost containment and innovation in recent years may have fed the gradual decline in the number and membership of DB pension plans, and the rise of DC plans and group RRSPs, but this need not mean that the DB plans that continue to exist must put a brake on productivity.
Job mobility could also potentially be affected by lock-in provisions of pension plans. Lock-in regulations are set at the provincial level, and vary considerably. They generally allow lower lock-in for older workers. Assuming that the age test is passed, one province (Saskatchewan) allows 100% of pension value to be unlocked on employment termination, while other provinces require significant lock-in — 50% in Alberta for example and 75% in Ontario (see Joint Expert Panel on Pension Standards, 2008). It is generally believed that job mobility will tend to be encouraged more the greater the percentage of pension assets that are unlocked on employment termination. This is in part because many people would simply prefer to have the freedom to decide what to do with their assets. But, in addition, some people may suffer real hardship due to the limit on annual withdrawals from locked-in RRSPs and LIRAs. Encouraging mobility by allowing pension assets to be unlocked is likely a good thing for the sake of labour market efficiency and productivity, but there may bepolicy concerns motivating a significant lockin percentage. There may be a fear that some people will handle their unlocked funds badly and become unnecessarily dependent on the state for retirement income. And it could also be that people will leave employment for the "wrong" reasons — getting access to pension assets rather than the availability of a job in which they will be more productive.
It is important to note again, in this context, the increasing size and popularity of group RRSPs. These of course have no lock-in feature and therefore are neutral with respect to worker mobility.
Labour Supply and Retirement Age Effects
In principle pension and savings plans can affect both hours of labour supplied and the age of retirement. However, most observers believe that the most important margin is retirement age, and here we will focus on that for the most part.
Most people in Canada retire by the age of 65 — about 95% of those over 65 are completely out of the labour force. It is worth asking if our pension or saving schemes are inducing people to retire too early. There has been considerable research on this, for public pensions (see e.g., Milligan and Schirle, 2008, for a recent summary and the most recent evidence). It turns out that the retirement inducement effect of public pensions in Canada is mild, and weaker than found in most other Organisation for Economic Co-operation and Development (OECD) countries. Milligan and Schirle (2008) find that, as a fraction of current earnings, the impact on pension accrual of working an extra year for a single male in Canada is +3.6% at age 55 and -18 % at age 65. These rates are similar to those reported by others for the United States, but much lower than in e.g., France or Belgium.
One reason that public pensions in Canada should tend to have relatively little effect on retirement age is that the former CPP/QPP work test has been abolished for some time now. However, there are other effects that mean there is a mild pension rights "accrual effect" encouraging early retirement. First, CPP benefits can be taken starting at age 60, but monthly benefits have been 30% less than if one waited to age 65 to commence receiving the pension. Up to the present, for each month that one postpones starting the pension, between ages 60 and 70, benefits have risen by 0.5%.This rate of increase falls short of the actuarially fair rate, which has been recognized in plans to raise the rate of increase over the next few years to 0.6% from age 60 to 65 and 0.7% from age 65 to 70.21 Second, working longer raises CPP/QPP benefits, which will reduce eventual GIS payments for about the bottom third of earners. Third, working longer can reduce "the Allowance" (formerly termed the "spousal allowance") and past age 65 will directly reduce GIS benefits for low earners.
It might seem that these public pension effects do not have much implication for policy toward private savings and pension assets. However, there can clearly be impacts of changing the mix between public and private assets. TFSA income, for example, has no impact on GIS (or OAS) receipt, so the second and third accrual effects of CPP/QPP are not seen with TFSAs. And while RPPs or RRSPs do interact with GIS and OAS, for DC RPPs, RRSPs and RRSP-like assets there is no accrual effect from actuarial unfairness in the adjustment of pension benefits as one works longer.
Within the class of private assets there can be significant differences in retirement age effects. While DC RPPs and RRSPs may have little effect on retirement age, some DB RPPs can have a strong effect. A typical DB plan will give credit equal to 2% of the average of the best five years' earnings per year in the plan up to a limit — 30 years would be representative. This means that for the first 30 years the employee accrues pension benefits at a good rate, but then gets only small further pension growth (due to the possible rise of the best five years' earnings). Workers in this situation have achieved the right to a "full pension", and may retire earlier than might otherwise be expected, although this is constrained in many cases by a minimum retirement age — often age 55. The people who retire in this way may take new jobs, which may in some cases be higher paying. But a more typical case is likely to be a reduction in earnings, likely reflecting a decline in productivity.
Finally, one may ask what the pure effect of allowing sheltered retirement saving is on labour supply and retirement age. Assume there are no pension right accrual effects, that is that the pension and savings vehicles considered will simply allow expected pension benefits to be built up that have present value equal to that of contributions. Such neutral savings/pension vehicles (which are approximated by our RRSPs or DC RPPs) have counteracting effects on labour supply. Improved or more attractive savings and pension schemes increase the payoff from working, and act similarly to a wage subsidy. According to economic theory the results can be an increase or decrease in hours worked. If the effects are the same as those of a simple wage increase, the effect for prime-age male workers will be small and could be negative. For prime-age female workers the impact is more likely to be positive. For a complete analysis, however, one should also take into account that sheltering savings from tax requires the government's budget to be balanced by higher taxes, reduced expenditure or increased borrowing — all with possible labour supply effects whose strength and direction will depend on what measures are taken.
We have seen that there a rich range of issues of savings instrument design. These have been examined against a background of a high rate of participation by Canadians in RPPs, individual and group RRSPs, and some other sheltered savings plans (e.g., DPSPs) by the time retirement is reached. As noted here and as investigated by LaRochelle-Côté (2008) and Baker and Milligan (2009), a small minority appear to save too little in these forms to allow even 50% replacement of income in retirement. While the difference may be made up by saving and accumulating assets in non-sheltered assets, such as housing, it is no doubt true that there is under-saving by some groups. But it is important to recognize that lack of saving for retirement is not generally irrational behaviour for low income Canadians, who will receive a high replacement rate for their pre-retirement earnings from public pensions.
Our discussion of the rationale for private pension and saving plans indicates good reasons for the existence of these plans, and also for having multiple plans. As economists have long argued, a combination of tax-postponed plans, like RRSPs and DPSPs, with tax-prepaid plans, like TFSAs, is needed in order to allow tax averaging over time and to avoid savings distortions. These plans are also very likely to be effective in achieving the goal of enhanced retirement income for middle and upper income earners. Empirical evidence indicates that there is significant substitution of sheltered for non-sheltered assets, which reduces the net saving impact of sheltered plans. However, empirical evidence in the United States and Canada on balance indicates a non-negligible positive impact on saving, particularly for middle and upper-middle income groups. With rates of return much higher than could be achieved on non-sheltered saving the result must be significantly increased retirement income for plan participants, although there may be some opposite effect for those who suffer from increased taxes or reduced expenditures in other areas that are necessary to compensate for the revenue lost in sheltering savings.
We have also discussed substitution possibilities between different kinds of pension and savings plans. There are some interesting effects. For example, DC pension plans and group RRSPs appear to be good substitutes from the viewpoint of participants, and the latter may be superior from the viewpoint of both employers and employees in many cases due to greater flexibility and lower administrative costs. This is consistent with the rapid rise of group RRSP plans and the continuing decline of RPPs. There is likely also to be some substitution away from RRSPs toward TFSAs among savers who anticipate being in the GIS or OAS claw-back ranges for much of their retirement. And finally, substantial substitution is likely from existing plans towards the broad-based new public sheltered saving plans that have been proposed e.g., by the Joint Expert Panel on Pension Standards (2008) and Ambachtsheer (2008, 2009).
Our examination of the risk characteristics of different plans indicates a complex and interesting pattern. Both DB and DC pension plans have significant risks — with DB plans seemingly safer in normal times but having perhaps greater "catastrophic" risk. Looking at the numbers on asset price fluctuations since the time of the last Survey of Financial Security in June 2005 shows that there has likely been quite significant redistribution of private pension/saving assets in particular and net worth in general among Canadians. Participants in DC plans, DB plans that have been wound up, and risky RRSP investments have, in some cases, lost half or more of their net worth. The robustness of housing prices, and the popularity of less risky asset types, have cushioned these trends to an important extent however.
Finally, we have looked at a range of "other" effects — possible impacts on macroeconomic stability, productivity, retirement age, and labour market mobility. There is some reason for concern about effects on mobility and retirement age, stemming mainly from accrual effects and lock-in of pension assets when workers change jobs in the mobility case, and from effects including, again, interactions between pension benefits and the GIS claw-back in the retirement age case. Macroeconomic stability, on the other hand, may be enhanced by the large fraction of private savings that are in pension or savings plans in Canada. It is in general more difficult to access stock market gains for consumption purposes if the stocks are tied up in pensions or savings plans than if they were held directly, and the empirical evidence appears to indicate a small effect of changes in stock market wealth on consumption demand.
Ambachtsheer, Keith, 2008, "The Canada Supplementary Pension Plan: Towards an Adequate, Affordable Pension for All Canadians", Commentary No. 265, C.D. Howe Institute, Toronto..
Ambachtsheer, Keith, 2009, "Pension Reform: How Canada Can Lead the World", Benefactor's Lecture, C.D. Howe Institute, Toronto.
Auerbach, Alan J., 2006, "The Choice Between Income and Consumption Taxes: A Primer", UC Berkeley: Berkeley Program in Law and Economics. Retrieved from http://escholarship.org/uc/item/9q85f6qz
Baker, Michael and Kevin Milligan, 2009, "Government and Retirement Income in Canada", paper prepared for the Research Working Group on Retirement Income Adequacy.
Benjamin, Daniel J., 2003, "Does 401(k) Eligibility Increase National Savings? Evidence from Propensity Score Subclassification." Journal of Public Economics 87 (5-6): 1259-90.
Bernheim, B. Douglas, 2003, "Taxation and Saving", in A. Auerbach and M. Feldstein (eds.), Handbook of Public Economics, Vol. 3, North-Holland, Amsterdam.
Bradford, David and the U.S. Treasury Tax Policy Staff, 1977, Blueprints for Basic Tax Reform, U.S. Treasury; second edition, revised, 1984, Arlington VA, Tax Analysts.
Carroll, Chris and Lawrence H. Summers, 1987, "Why Have Private Saving Rates in the United States and Canada Diverged?", Journal of Monetary Economics 20: 249-79.
Davies, James B., 2009, "The Effects of Asset Price Changes on Economic Security in Canada, 2005-09", working paper presented at annual meetings of Canadian Economics Association, May.
Davies, James B. and France St-Hilaire, 1987, Reforming Capital Income Taxation in Canada: Efficiency and Distributional Effects of Alternative Options, Economic Council of Canada, Ottawa.
Dynan, Karen E., and Dean M. Maki, 2001, "Does stock market wealth matter for consumption?", Board of Governors of the Federal Reserve System, Finance and Economics Discussion Series: 2001-23.
Engelhardt, Gary V., 1994, "Tax subsidies to saving for home purchase: evidence from Canadian rhosps", National Tax Journal 47 (2): 363-88.
Engen, Eric M. and William G. Gale, 1997, "Debt, Taxes and the Effects of 401(k) Plans on Household Wealth Accumulation", working paper, Brookings Institution.
Engen, Eric M. and William G. Gale, 2000, "The Effects of 401(k) Plans on Household Wealth : Differences across Earnings Groups", working paper, Brookings Institution.
Financial Post, 2009, "Private Pensions OK, DBRS Says", The National Post, Thursday, Aug. 6.
Gale, William G., 2001, "The Effect of Pensions and 401(k) Plans on Household Saving and Wealth", in Gale et al. (2001): 102-122.
Gale, William G. and John Karl Scholz, 1994, "IRAs and Household Saving", American Economic Review 84: 1233-60.
Gale, William G., John B. Shoven and Mark J. Warshawsky (eds.), 2005, The Evolving Pension System, Trends, Effects and Proposals for Reform, Brookings Institution Press: Washington, D.C.
Globe and Mail, 2009, "Retirement Lost", October 17, p. A1.
Institute for Fiscal Studies, 1978, The Structure and Reform of Direct Taxation, (the "Meade Committee Report"), Allen and Unwin: London.
Jog, Vijay, 2009, "Investment Performance and Costs of Pensions and other Retirement Savings Funds in Canada: Implications on Wealth Accumulation and Retirement", paper prepared for the Research Working Group on Retirement Income Adequacy.
Joint Expert Panel on Pension Standards, 2008, Getting Our Acts Together, Alberta/British Columbia Pension Standards Review, November.
Kotlikoff, Laurence J., 2008, "Economics' Approach to Financial Planning", Journal of Financial Planning, March.
LaRochelle-Côté, John Myles and Garnett Picot, 2008, "Income Security and Stability During Retirement in Canada", Statistics Canada Catalogue No. 11F0019M - - No. 306.
Luchak, Andrew and Tony Fang, 2005, "Pensions or the Group RRSPs: Patterns of New Plan Adoptions, Establishing Secondary Plans, Terminations and Plan Substitutions", University of Alberta and University of Northern British Columbia, prepared for Conference on the Evolving Workplace, Toronto, Sept. 28-29.
Milligan, Kevin, 1998a, "Savings and Tax Incentives: Semiparametric Estimation of the Household Savings Impact of RRSPs", mimeo, July.
Milligan, Kevin, 1998b, "Does Tax-Sheltered Savings Crowd Out Other Savings? Implications, Evidence, and Explanations", mimeo, August.
Milligan, Kevin and Tammy Schirle, 2008, "Improving the Labour Market Incentives of Canada's Public Pensions", Canadian Public Policy 34 (3): 281-303.
Muellbauer, John, 2008, "Housing and Personal Wealth in a Global Context", in James B. Davies (ed.), Personal Wealth from a Global Perspective, Oxford University Press, Oxford: 293-311.
Pichette, Lise and Dominique Tremblay, 2003, "Are Wealth Effects Important for Canada?", Bank of Canada Working Paper 2003-30, October.
Poterba, James M., 2000, "Stock Market Wealth and Consumption", Journal of Economic Perspectives, 14 (2): 99-118.
Poterba, James M., Steven F. Venti, and David A. Wise, 1995, "Do 401(k) Contributions Crowd Out Other Personal Saving?", Journal of Public Economics 58 (1): 1-32.
Shillington, Richard, 1995, "Tax Treatment of RRSPs and RPPs", in Roundtable on Canada's Aging Society and Retirement Income System, Caledon Institute on Social Policy, June 5.
Shillington, Richard, 1999, "The Dark Side of Targeting: Retirement Saving for Low-Income Canadians", C.D.Howe Institute Commentary No. 130, September.
Starr-McCluer, Martha, 2002, "Stock Market Wealth and Consumer Spending", Economic Inquiry 40 (1): 69-79
Statistics Canada, 2006, "The Wealth of Canadians: An Overview of the Results of the Survey of Financial Security", Pensions and Wealth Surveys Section, Catalogue No. 13F0026MIE - - No. 001.
Venti, Steven F. and David A. Wise, 1986, "Tax-Deferred Accounts, Constrained Choice, and Estimation of Individual Saving," Review of Economic Studies 53: 579-601.
Venti, Steven F. and David A. Wise, 1990, "Have IRAs Increased U.S. Saving?: Evidence from Consumer Expenditure Surveys", Quarterly Journal of Economics 105: 661-98.
|Age||% of Family Units Holding…||Value of RRSPs, RRIFs,
LIRAs, other1 among holders($)
|35 to 44||72.9||63.3||48.0||22,500||49,100|
|45 to 54||76.7||68.1||51.6||40,000||90,300|
|55 to 64||81.9||69.4||60.1||60,000||124,500|
|65 and older||72.5||51.2||57.2||50,000||108,200|
|1 "Other" includes
DPSPs, annuities and other miscellaneous pension assets.
Source: Statistics Canada (2006) tables 6 and 12.
|Aged less than 65||Aged 65 or over|
|Mean ($)||% of After-tax Income||Mean ($)||% of After-tax Income|
|Stocks and Mutual Funds||15,741||30.3||34,511||98.1|
|Other Financial Assets||4,361||8.4||4,329||12.3|
|Total Financial Assets||35,573||68.3||91,332||259.8|
|Saving Plans and Pension Assets|
|Employer Pension Plans||68,344||131.2||120,023||341.4|
|Total Savings Plans and Pension Assets||108,540||208.4||174,437||496.1|
|Other Real Estate||36,581||70.2||32,682||93.0|
|Source: Author's calculations using Statistics Canada public use micro data file, 2005 Survey of Financial Security.|
|Earnings||Public Pension Replacement Rate (%)1||RRSP/RPP Contribution Rate (% of earnings) Required To Obtain Retirement Income Replacement Rate of…||Mean Actual RRSP/RPP Contribution Rate (%)|
|1Takes into account effect of RRSP and RPP income on GIS
and OAS payments; based on total replacement rate of 70% (except at the
$20,000 earnings level where the total replacement rate equals the public
pension replacement rate of 72%).
2Exceeds contribution limits by 2 % point.
Source: Columns 1 -3 and 6, Finance Canada. Columns 4 and 5, author's calculations extending Finance Canada results.
|Earnings||Age 25-44||Age 45-64|
|70% Replacement Rate||60% Replacement Rate||50% Replacement Rate||70% Replacement Rate||60% Replacement Rate||50% Replacement Rate|
|1Includes 10% of those at this earnings level who are
already at their contribution limit.
2Includes 13% of those at this earnings level who are already at their contribution limit.
Source: Author's calculations using data in Table 3 and data on the distribution of contribution rates by earnings level provided by Finance Canada.
|Earnings||Zero||0 – 5%||5 – 50%||50-100%||Above 100%|
|Earnings||Zero||0 – 5%||5 – 50%||50-100%||Above 100%|
|Source: Author's calculations using data provided by Finance Canada.|
|Canadian Dollar (in US $s)||S&P 500 in Canadian Dollars|
|I. Original index values|
|II. June 2005 Base|
|Source: Bank of Canada publications and Teranet-National Bank House Price index website.|
1 A number of different proposals of this type have been made. Prominent examples are the Alberta and British Columbia Pension (ABCP) scheme advocated in Joint Expert Panel on Pension Standards (2008), and the Canada Supplementary Pension Plan (CSPP) put forward by Ambachtsheer (2008).
2 The Old Age Security program.
3 The Guaranteed Income Supplement.
4 It would of course be preferable to have a single data source with both stock and flow information using consistent units of observation. Unfortunately, it is not realistic to expect that such a data source will soon be forthcoming. Tax-filers are not required to report the value of their holdings in RRSPs, RPPs etc. when they file their income tax returns. And adding questions on contributions, withdrawals, benefits and so on could make the SFS survey too burdensome for respondents.
5 In 2007, according to unpublished Finance Canada data, there were 12,535 RPP plans in Canada with assets of $38.2 billion. At the same time there were 23,524 group RRSPs with assets of $33.0 billion.
6 When looking at the composition of wealth it is more informative to look at means than medians, e.g., because medians for particular types of asset or debts are strongly influenced by the % holding that type of asset or debt. In some cases, e.g., stocks and mutual funds, other real estate or business equity, less than 50% of families hold the asset, leading to a median of zero. Also, means of the various assets and debts add up to the mean for total assets whereas medians do not have that adding-up property.
7 The numbers shown in the 60% and 50% replacement rate columns were calculated by adjusting the 70% replacement rate column , using the same assumptions as Finance Canada and taking into account impacts on OAS and GIS of lower RRSP/RPP saving.
8 This is one reason for the fact that in any given year a significant fraction of tax-filers contribute more than 18% of earnings. This may reflect unusually high savings resulting in the use of some contribution room from previous years. (Contributions exceeding 18% of current earnings may also result, however, from the fact that the contribution limit is set on the basis of prior-year earnings.) Among those aged 25-44, 5% of those with earnings at the $20,000 level had a contribution rate over 18% and this rose steadily with earnings up to a peak of 17% at $100,000. For those 45-64, 10% contributed more than 18% at $20,000 of earnings and once again this fraction rose with income – to a peak of 22% at earnings of $80,000.
9 This conclusion agrees with that of Baker and Milligan (2009) who review the evidence on the behaviour of consumption as people move into and through retirement. They conclude that there is not evidence of a systematic drop in consumption standards on retirement for the population as a whole, which would be observed if the majority of people were under-saving for retirement.
10 This was apparently first pointed out in the "Blueprints" report (Bradford et al., 1977) and the report of the Meade Committee (Institute for Fiscal Studies, 1978). The proposition has been restated many times in discussions of the consumption tax approach - - in Canada e.g. by Davies and St Hilaire (1987). See also Auerbach (2006). Having both plan types available is needed for averaging purposes.
11 In a pure consumption tax world there would be no limit on contributions to TFSA's. However, on the transition to such a world there would be large transitional losses of government revenue as a result of taxpayers transferring existing assets to TFSA's.
12 It has been suggested by some that this anti-saving distortion might not be large since there might not be much effect on the saving of people of working age - - say at age 40. This argument depends, to an extent, on myopia. Even if the claim were true, in retirement people are faced with a decision of whether to consume or save their forced withdrawals. The amounts involved in many cases will be sizeable, and greater than can be sheltered in TFSAs. The tax penalty on unsheltered saving may therefore represent a significant disincentive to save from withdrawals for many people.
13 There could perhaps be a fear that some retirees would experience poverty in consumption terms if withdrawals were not forced. This could occur, one can imagine, due to financially unsophisticated retirees not understanding how much they could safely withdraw for spending, or perhaps due to pressure by their heirs not to withdraw.
14 If the new employer is strongly motivated in its hiring effort it may compensate the new employee for this loss of pension wealth, at least in part. This can be very expensive, however, and is likely to the be the exception rather than the rule.
15 See e.g. Globe and Mail (2009), which reports losses of 30 – 40% of pension benefits at a number of bankrupt firms, including e.g. Nortel Networks. To put this in perspective, however,note the recent DBRS study which reviewed 70 of the largest private-sector DB plans and found that they were on average 95% funded, reflecting a relatively small drop from being 98% funded in 2007 (Financial Post, 2009).
16 On October 27, 2009 the Minister of Finance announced changes to the federal private pension legislative and regulatory framework including an increase in the pension surplus threshold from 10% to 25% of assets. See http://www.fin.gc.ca/n08/09-103-eng.asp
17 The timing of peaks and troughs is a little different in the U.S. than in Canada. The S&P 500 monthly closing peak came at the end of October 2007, when the index stood at 1549.4, or 30.1% above its June 2005 value.
18 From its peak at the end of October 2007 the S&P 500 fell 52.6% to the end of February 2009, which was its trough month, as for the TSX.
19 This is more appropriate than Statistics Canada's New House Price index, for present purposes. But it is interesting to note that the latter shows fairly similar price changes. From June 2005 to May 2008, for example, the Stat Can index rose 22.5%, compared with a 29.0% rise in the Teranet-National Bank index. In February 2009 the Stat Can index was 20.1% above June 2005, while the Teranet-National Bank index stood 21.2% above June 2005.
20 This follows the relative number of members in different plan types reported by Statistics Canada fairly closely (see Statistics Canada's The Daily, July 4, 2008). Of the 5,768,280 pension plan members in 2007, 80% were in DB plans, 16% in DC plans, 4% in combination (combined DB and DC) plans, and 1% in other (including hybrid) plans.