# Archived - Retirement Saving by Canadian Households

Archived information

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.

Report for the
Research Working Group on Retirement Income Adequacy

Keith Horner
December 1, 2009

I would like to acknowledge the helpful comments of referees Charles Beach and Ross Finnie as well as Jack Mintz and other participants at the Experts Day. I would also like to thank Pierre LeBlanc, Stephanie Andrews and Mark Christie of the Department of Finance for their excellent work in providing the data tabulations used in the study. The views expressed in the paper are my own, however, and should not be attributed to the Department of Finance or any of those who provided comments.

Summary
1. Introduction and Summary
2. Trends in Retirement Saving and Household Wealth 3. Benchmarks for Savings Adequacy 4. Household Retirement Savings Patterns in 2006 5. Evidence from Longitudinal Data
6. Savings Adequacy in the Future Annex A. Further Savings Adequacy Results
Annex B. Savings among the Self-Employed
Annex C: Savings Adequacy Results with a 2.5% Rate of Return
References

## Summary

#### Retirement Savings by Canadian Households -- by Keith Horner

Private provision of retirement income, through registered pension plans (RPPs), deferred profit sharing plans (DPSPs) and registered retirement savings plans (RRSPs) is a large and growing component of Canada's retirement income system. Unlike our public pension programs, this component of the system is highly diverse. It is also voluntary, relying on individual choice by workers and their employers. Its performance needs to be assessed from time to time.

This paper aims to contribute to such an assessment, examining savings levels among Canadian households and testing whether people seem to be saving enough to avoid serious drops in their living standards at retirement.

Section 2 begins by looking at the trend in the level of total contributions to RPPs and RRSPs and exploring the apparent disconnect between rising rates of RPP/RRSP saving and a personal sector savings rate that has fallen to very low levels since the early 1980s.

Section 3 sets the stage for a look at retirement savings patterns by developing benchmarks for savings adequacy. These benchmark savings rates, which vary by household or family type and across earnings levels, are derived from a simple model of consumption and savings over the lifecycle. The model is extremely simplified in some respects. The model is based on the hypothesis that people will seek to have the same level of consumption after retirement as before. In applying this idea, changes in family size are taken into account. Saving through homeownership is also taken into account.

The model yields savings rate benchmarks that tend to rise strongly with earnings, reflecting the diminishing role of public pension income as earnings rise, and that also vary quite widely according by household type (e.g., single vs. couple, homeowner vs. renter). The model also generates target replacement rates (i.e., ratios of total pre-tax retirement income to pre-tax earnings) that vary quite widely as well – from about 75% for single renters to about 50% for two-parent homeowner families.

Section 4 examines the level and pattern of retirement savings in Canada in relation to the benchmark rates. The analysis is based on a family file of taxation data from the Canada Revenue Agency for the year 2006.

Overall, it appears that about 69% of Canadian households saved in RPPs and RRSPs at rates sufficient to fully maintain their consumption levels in retirement (100% replacement rate). About 78% of households met a lower target consistent with a 10% reduction in consumption at retirement (90% replacement rate).

Section 5 considers how much these results might be distorted by the focus on saving in a single year. Evidence is provided that there is considerable regularity in retirement saving. People who belong to an RPP or contribute to and RRSP in one year tend to do so for a stretch of years.

Section 6 concludes with some thoughts regarding developments likely to influence the future performance of the private savings component of our retirement income system.

## 1. Introduction and Summary

Private provision of retirement income, through registered pension plans (RPPs), deferred profit sharing plans (DPSPs) and registered retirement savings plans (RRSPs) is a large and growing component of Canada's retirement income system.1 Unlike our public pension programs, this component of the system is highly diverse. It is also voluntary, relying on individual choice by workers and their employers. Its performance needs to be assessed from time to time.

This paper aims to contribute to such an assessment, examining savings levels among Canadian households and testing whether people seem to be saving enough to avoid significant drops in their living standards at retirement.

Section 2 begins by looking at the trend in the level of total contributions to RPPs and RRSPs and exploring the apparent disconnect between rising rates of RPP/RRSP saving and a personal sector savings rate that has fallen to very low levels since the early 1980s. One clarification is that RPP/RRSP contribution rates are not the same as savings rates because of the tax deductibility of contributions.

There are also problems with the National Accounts savings rate as a guide to wealth accumulation by households. It ignores wealth increases through capital gains, it ignores saving through the purchase of consumer durables, and it is distorted by inflation. When adjusted to overcome these problems, the personal sector savings rate, though volatile, is higher and has no significant downward trend. This is consistent with the view obtained from personal sector asset and debt trends. They show that, despite strongly growing levels of consumer debt, the net worth of households has grown appreciably over the long term. Even after last year's market crash, the ratio of total net worth to total earnings remains considerably higher than it was in the early 1990s or any time before that.

Section 3 sets the stage for a look at retirement savings patterns by developing benchmarks for savings adequacy. These benchmark savings rates, which vary by household or family type and across earnings levels, are derived from a simple model of consumption and savings over the lifecycle. The model is extremely simplified in some respects. For example, a 55-year adult lifespan is collapsed into two periods – before and after retirement at age 65. This simplification allows the model to be very detailed when it comes to representing the tax/transfer system and its effect in shaping savings needs and incentives.

The model is based on the hypothesis that people will seek to have the same level of consumption after retirement as before. In applying this idea, changes in family size are taken into account. Saving through homeownership is also taken into account.

The model yields savings rate benchmarks that tend to rise strongly with earnings, reflecting the diminishing role of public pension income as earnings rise, and that also vary quite widely according by household type (e.g., single vs. couple, homeowner vs. renter). The model also generates target replacement rates (i.e., ratios of total pre-tax retirement income to pre-tax earnings) that vary quite widely as well – from about 75% for single renters to about 50% for two-parent homeowner families.

Section 4 examines the level and pattern of retirement savings in Canada in relation to the benchmark rates. The analysis is based on a family file of taxation data from the Canada Revenue Agency for the year 2006.

Overall, it appears that about 69% of Canadian households saved in RPPs and RRSPs at rates sufficient to fully maintain their consumption levels in retirement. About 78% of households met a lower target consistent with a 10% reduction in consumption at retirement (90% replacement rate).

At low earnings levels, the targets were met with little or no saving. Savings shortfalls were greater at modest and middle earnings levels and are greatest for high-income earners. The latter group is likely affected by the contribution limits but is also the most likely to have retirement savings outside of RPPs and RRSPs. Shortfalls from the benchmark rates were found particularly among those with the highest savings benchmarks – single individuals and two-earner couples without children.

Those in the modest and middle earnings groups are the main focus of policies relating to retirement saving. In these groups, 28% (modest earners) and 29% (middle earners) did not save enough to meet the 90% consumption replacement target.

Those saving for retirement in RRSPs alone were much more likely to fall short of the savings rate benchmarks than those belonging to RPPs. This is evident in comparisons of RPP-only and RRSP-only savers. It is exacerbated by the fact that RPP members were almost as likely as non-members to contribute to RRSPs.

Savings in financial assets held outside of RPPs and RRSPs are important among high earners, but taking them into account would have little effect on the estimated incidence of savings shortfalls among modest and middle earners.

Households with earnings (over $8,839) from self-employment have RPP/RRSP savings rates that are surprisingly similar to those of employed workers. Thus, low savings by this group (which accounts for about 8% of households at modest/middle earnings levels) contributes little to the savings shortfall incidence estimates. The value of business assets is substantial, but these assets are very unevenly distributed. Taking them into account would raise average wealth levels considerably, but would have a much smaller effect on the incidence of savings shortfalls among modest and middle earners. Because of slow labour force growth, the rate of return on saving may be lower over the coming decades than in the past. A sensitivity analysis provides savings adequacy rates assuming a 2.5% rate of return. Under this hypothesis, the proportions of modest and middle earners falling short of the 90% consumption replacement targets would rise from 28% to 30% for modest earners and from 29% to 34% for middle earners. The effects are perhaps smaller than expected because the higher savings levels needed with a lower rate of return depress pre-retirement consumption levels and thus reduce the post-retirement consumption levels needed to maintain living standards. Section 5 considers how much these results might be distorted by the focus on saving in a single year. Evidence is provided that there is considerable regularity in retirement saving. People who belong to an RPP or contribute to and RRSP in one year tend to do so for a stretch of years. A study by LaRochelle-Côté, Myles and Picot (2008) of the time path of after-tax income levels as people enter retirement provides information consistent with the savings adequacy results of Section 4. Another study, by Ostrovsky and Schellenberg (2009), demonstrates that the savings gap between RPP and RRSP-only savers found in the 2006 data does not show up in the same way in retirement income levels. Even though RPP and RRSP saving may be reasonably consistent from year to year, the categories of RPP member and RRSP-only saver blur over a full career as people change jobs. Many RRSP-only savers have RPP wealth from earlier employment. Section 6 concludes with some thoughts regarding developments likely to influence the future performance of the private savings component of our retirement income system. Last year's market crash and the economic recession are too recent to be evaluated but will undoubtedly have longer-term effects. If the recession deepens, forced early retirements and other employment losses may be more important effects than the current financial losses. A continuation of the long-term decline in pension coverage would obviously have negative consequences, particularly given the evidence that RPP members save substantially more than RRSP-only savers. The introduction of Tax Free Savings Accounts (TFSAs), on the other hand, will make it easier for Canadians to maintain their consumption levels in retirement. TFSAs will not stimulate additional saving but will make retirement saving more effective by reducing the tax burden on retirement income. TFSAs are likely to be attractive to about 40% of households, those earning up to about$75,000 per year. The advantages of TFSA saving to households are reflected in some level of net fiscal cost to governments, mainly in the form of higher rates of GIS eligibility.

## 2. Trends in Retirement Saving and Household Wealth

In 2007, Canadians and their employers contributed $76 billion to RPPs and RRSPs. This represented almost 8.5% of their disposable income and about 10.8% of earnings reported for tax purposes. Moreover, contribution rates to retirement savings plans have risen steadily though unevenly for the past several decades (Figure 2.1).2 Figure 2.1 - Contribution Rates to RPPs and RRSPs Sources: Cansim table 2800026 for RPP contributions; Canada Revenue Agency Income Statistics for earnings and RRSP contributions. In contrast, the personal savings rate was only 2.5% in 2007, and as shown in Figure 2.2, this savings rate has been in steady and pronounced decline since the early 1980s. Figure 2.2 Personal Savings Rate Sources: Cansim tables 3800019 and, for 2009 data, 3780009 These dramatically different pictures of the state of household saving raise questions. Which measure should we believe. Or, might they both be correct with the difference explained by massive dissaving, through growing use of consumer credit for example, offsetting increases in contributions to retirement plans? ### 2.1 Contribution and savings rates Limitations of both measures go far to explain the differences. As introduction, consider that saving is both a setting aside of income for future use and an accretion of wealth. The growth of wealth may be described by the accounting identity in which Wt is wealth in year t, i is the rate of return on savings, Et is earnings, Trt is income from government transfer programs, Tt is taxes and Ct is consumption. By rearranging terms, the level of savings, St, may be expressed both as the increase in wealth and the shortfall of consumption from disposable (i.e., after-tax) income. The first point to note is that the contribution rate to RPPs and RRSPs is not the same as a savings rate. Consider an RPP or RRSP contribution of$5,000, for example. Because the contribution is tax deductible, the net cost to the contributor, and the net effect on consumption, is considerably less – $3,450, say, if the contributor faces a marginal tax rate of 31%. For someone with income of$60,000, this implies a retirement savings rate of 5.8% rather than 8.3%.

The personal savings rate in the National Accounts also has problems when used as a measure of changes in the wealth position (or saving) of Canadian households.

1. Consumer durable expenditures are treated purely as consumption. In economic terms, though, they involve the purchase of assets that will provide consumption services over periods of several, or many, years and so should be treated mainly as saving rather than consumption.
2. Because the income and expenditure accounts are aimed at measuring and classifying economic activity in a year, their definition of income excludes capital gains – an important component of household wealth increases.
3. The National Accounts savings rate is distorted by inflation. Investment returns include an inflation premium. This premium increases measured savings and income despite the fact that it serves only to maintain the existing level of real wealth. Since its effect on the numerator of the savings rate (disposable income less consumption) is proportionally much greater than its effect on the denominator (disposable income), the inflation premium produces an upward bias in the measured savings rate. This source of bias is always present but has been particularly important during episodes of high inflation such as the early 1980s.

Using National Accounts balance sheet data for the personal sector, it is possible to derive a measure of saving that corrects for these three problems.

The first step is to deflate all the relevant flows and wealth levels by changes in the CPI to express them in constant dollars. This does not affect the savings rate, but it is important for the other adjustments. The relevant flows and levels are: savings (as per the income accounts), personal disposable income, net worth, and levels of consumer durables.

Second, annual levels of net saving in consumer durables are obtained by calculating the year-over-year changes in the stock of those assets.

Third, a simple measure of annual capital gains is created by subtracting from the annual change in net worth (measured on a market value basis) the sum of measured savings and consumer durable savings. These annual capital gains amounts are then added to disposable income to create an adjusted income measure ('full disposable income').

Finally, the savings rate is calculated as savings (equal to the change in real net worth) divided by full disposable income. This method follows that set out in Finance Canada (2000 and 2002).
The result is shown in Figure 2.3. The Balance Sheet savings rate is much more volatile than the National Accounts one, reflecting swings in asset values from year to year. However, over the longer term, the Balance Sheet rate is higher than the National accounts rate and exhibits no discernable trend (at least so long as 2008 is treated as an outlier).

Figure 2.3 Balance Sheet Savings Rate

Source: Author's calculations from data in Cansim tables 3780004 and 3780009.

### 2.2 Trends in household wealth

A similar picture is obtained by looking directly at how personal sector assets and debts have evolved over time. This is done in Figure 2.4 which shows the trends in (a) total assets, (b) total liabilities and (c) net worth [= (a) less (b)].

To correct for inflation, real wage growth and population growth, the asset and debt levels are expressed as ratios of earnings, here defined as wages, salaries and supplemental labour income plus net income from unincorporated business. Earnings are used here instead of disposable income since they are the relevant base for retirement saving. They exclude components such as investment income that do not cease at retirement and so do not need to be replaced to allow living standards to be maintained in retirement. Basing the ratios on disposable income rather than earnings, however, would have little effect on the trends.

Figure 2.4 shows that, despite the decline in the personal savings rate since it peak in 1982, the net worth of Canadian households has continued to grow. Moreover, the growth has occurred mainly since the savings rate started to decline.

The bursting of the high technology bubble is reflected in a short-term decline in net worth from 6.24 in 1999 to 5.73 in 2002, and the effects of last year's market crash are plainly evident, with the June 2009 ratio of 6.10 remaining over 8% below the 2007 peak of 6.66.

Another point of interest concerns household debt. Total household liabilities (mainly mortgage debt and consumer credit) have indeed risen dramatically – from about 1.0 times earnings in 1980 and 1990 to over 1.5 times earnings today. But, because household indebtedness is dwarfed by household assets, its strong growth has had relatively little impact on the growth net worth.

Figure 2.4 Growth of Household Assets and Debts
Ratio to Household Earnings

Sources: Cansim tables 3780009 and, for pre-1990 data, 3780004.
Note: the 2009 ratio is based on second quarter data.

Figure 2.5 shows how different asset classes contributed to the asset growth documented in Figure 2.4. In the top two lines of Figure 2.5, total assets are divided between non-financial assets (mainly residential structures and land) and financial assets. It is evident that financial assets contributed much more to the growth in net worth than did non-financial assets – until the collapse of the dot-com boom at least, when real assets began to appreciate faster than paper.

Within total financial assets, the two biggest categories are (a) life insurance and pensions and (b) shares. Deposits at banks and other financial institutions are a third important category, but these have grown only at the same pace as earnings. It is not possible in these data to identify assets held in RRSPs. However, RRSPs likely account for a majority of the holdings of shares. One observation is that the growth in pension assets has been considerably stronger than the growth in contributions portrayed in Figure 2.1. This reflects the increasing role of investment income, relative to contributions, as these plans have matured.

Figure 2.5 Growth in Selected Asset Categories
Ratio to Household Earnings

Sources: Cansim tables 3780009 and, for pre-1990 data, 3780004.
Note: the 2009 ratio is based on second quarter data.

The positive picture of household wealth accumulation in Figures 2.4 and 2.5 needs qualification in two respects. First, because wealth levels rise with age (at least up to retirement), population aging helps to produce an increasing wealth-to-earnings ratio. Second, it must be recognized that wealth is notoriously concentrated. Households with strong asset growth may not be the same ones that are becoming increasingly indebted.

It is possible to roughly simulate the effect of population aging on aggregate wealth by assuming that age-specific wealth levels are held constant and only the age structure changes. For example, in 1999, those in the age groups 45-54, 55-64 and 65+ had average levels of net worth that were 36%, 83% and 34% above the average for all households, and the combined share of these group in the total adult population rose from 37.5% in 1971 to 49.8% in 2008.3 In Figure 2.6, the trend in the ratio of net worth to earnings is adjusted by subtracting the amount of net worth growth calculated to be due to age structure changes.4

Figure 2.6 Trend in the Net Worth-to-Earnings Ratio with an Age Structure Adjustment
Ratio to Household Earnings

Source: Figure 2.4 and calculations by the author.

These calculations suggest that population aging accounts for a good part of the increase in household wealth, relative to earnings, over the past 30 years. The adjusted wealth-to-earnings ratio (after last year's asset value declines) is currently higher than it was in the 1970s or 1980s but has changed little from the early 1990s.

Regarding the distribution of household wealth, Morissette and Zhang (2006) have shown that, since the mid-1980s, it has become more unequal, with a growing share belonging to high-income and high-wealth households. This trend is evident in the period between the asset and debt surveys of 1984 and 1999 and it has continued up to 2005, the year of the latest survey.5 The main causes of the trend appear to be (a) historically high returns on existing wealth and (b) growing inequality in the earnings that support wealth accumulation.

Morissette and Zhang also showed that wealth ownership has become more concentrated among those in or nearing retirement. This trend was particularly strong in the 1984-1999 period but may have weakened since. Between 1999 and 2005, the growth in median and average net worth was strongest among those in the 35-44 age group.6

## 3. Benchmarks for Savings Adequacy

Because retirement savings needs vary dramatically across households according to their earnings level and family type, it makes no sense to explore savings patterns without some benchmarks indicating adequate savings levels.

The object of retirement saving is to maintain living standards in retirement by providing sufficient retirement income to replace earnings that cease at retirement. This is reflected in the second objective of government policy regarding retirement income security: to enable Canadians to avoid serious disruption of their living standards upon retirement (Finance Canada, 1989).7 It also conforms to the findings of economic lifecycle models of consumption and saving. In these models, the diminishing marginal utility of consumption in any year drives utility maximizers to smooth the time path of their consumption by saving in high-income years and dissaving in others.

To achieve this earnings replacement goal, private pension income need not equal pre-retirement earnings because:

• Several charges against income are lower in retirement than before. Work-related expenses (e.g., work clothes, tools, transport, some meals away from home) are eliminated. Payroll taxes cease. Income taxes, in a progressive tax system that includes special tax credits for seniors, are lower after retirement. And, income no longer needs to be set aside for retirement;
• Public pensions (OAS/GIS and the CPP/QPP) provide a base level of retirement income; and
• For families with children, parents' pre-retirement consumption levels are reduced by the income devoted to the children's consumption.8

Canada's public pension plans provide progressively less earnings replacement as earnings rise. OAS provides a flat rate benefit (reduced for those with high retirement incomes). The GIS is income tested. And, for those with earnings above the maximum covered (the YMPE, about 1.08 times the average wage), CPP/QPP benefits replace a declining percentage of earnings. Consequently, retirement savings levels needed to maintain living standards are zero for those at low earnings levels but rise very sharply as earnings increase.

### 3.1 A stylized model of household saving

As a guide to the exploration of savings levels among Canadian households, benchmark savings levels are developed here based on a stylized lifecycle model of household consumption and saving.
In this model:

(a) The accumulation period is 35 years, from age 30 to 64, and the retirement period is 20 years, from age 65 to 84;
(b) Contributions are made and benefits received at year-end. Inflation is 2%, the real rate of return on saving is 3.5%, and each saver's real wage grows at the same rate as the average wage – 1% per annum;
(c) Retirement income from RPPs, RRSPs or TFSAs is always received as a price-indexed life annuity.
(d) The savings rate is constant over the accumulation period;
(e) All saving is treated as RRSP saving that provides a tax deduction to the wage earner, and dollar limits on RRSP contributions are ignored;9
(f) The tax system is implausibly but conveniently wage indexed before retirement and price-indexed thereafter; and
(g) In cases where homeownership is considered, the value of the home (at age 64) is taken to be three times household earnings, the purchase is financed through 'constant-percentage-of-earnings' payments from age 30 to 64, and the full value of the home is realized in the form of 'housing services' (or imputed rent) contributions to consumption spread over the pre- and post-retirement periods.

These assumptions are designed to allow the analysis to be based on two sets of tax-transfer parameters – pre- and post retirement. For example, a worker whose income (net of deductible RRSP contributions) puts her halfway into the second tax bracket is assumed to remain at that point for the full accumulation period. The assumption of retirement at age 65 also ensures that measures like OAS/GIS and the age credit apply for the full retirement period and none of the accumulation period.10

For a single individual, for example, consumption at 64 =

Earnings
Less: retirement savings
Less: homeownership savings (for homeowners)
Plus: imputed rent from homeownership (for homeowners)
Less: work-related expenses (taken as = $300 + 3% of earnings) Less: payroll taxes ( CPP/QPP and EI) Less: federal and provincial income taxes Plus: GST credit and provincial credits. Consumption at 65 = Private pension income Plus: CPP/QPP benefits Plus: OAS benefits, net of the claw-back Plus: GIS benefits Plus: imputed rent from homeownership (for homeowners) Less: federal and provincial income taxes Plus: GST credit and provincial tax credits. Private pension income is determined as the retirement savings level times a factor that summarizes the 35-year accumulation and conversion to a 20-year annuity. For the purposes of provincial income taxes and tax credits, an Ontario resident is assumed. Although the model is applied to a sample of households for 2006, important changes tax-transfer changes introduced since then are taken into account. These include: increases in the GIS benefit maximums (beyond normal indexation) in 2006 and 2007; reduction in the first federal tax rate to 15%; introduction of the Universal Child Care Benefit (UCCB) and the$2,000 child tax credit; increases in the federal basic, spousal and age credits; the introduction of pension income splitting between spouses; the reduction (proposed for 2010) of the first Ontario tax rate to 5.05%, changes to surtax thresholds, and the enrichment of the Ontario sales and property tax credits. The levels of benefits and credits increased or introduced since 2006 are deflated to 2006 dollars.

Additional assumptions are required in the treatment of other family types.

For two-earner couples, a 60-40 split of earnings is assumed, but all saving is assumed to be done by the higher-income spouse.

As was the case for single parents vs. singles, benchmark savings rates are considerably lower for two-parent families than for childless couples. At household earnings of $100,000, for example, the rates are 5.6% for a two-earner family with children and 9.2% for a two-earner childless couple. Again, the benchmark savings rates are higher for one-earner than two-earner couples. This analysis also demonstrates that while the earnings replacement rates consistent with consumption continuity are fairly uniform across earnings levels (above a modest earnings level), they vary widely by family type and circumstance – from a high level of over 75% for a single renter to 67% for a two-earner homeowner couple, 55% for a single-parent renter and 53% for a two-parent homeowner family. ### 3.3 Comments on the model and parameters Other choices could have been made in the design of this model and its parameterization. Comments are provided here on the rationale for some of the choices made and the effects on the benchmark savings rates of variations in some parameters. 100% Consumption Replacement (1). The idea of utility maximization that underlies lifecycle models could have been pursued, leading to retirement consumption targets that different from 100% of pre-retirement consumption. This approach was followed in the savings adequacy study by Scholz et al (2004) and also in my analysis of the interaction between retirement savings and the cost of the OAS/GIS programs (Horner 2008). In the solution of a lifecycle model with utility maximization, the relative size of optimal C65 and C64 are determined by the Euler condition which embodies trade-offs between (a) the consumer's preference for consumption now over consumption in the future, (b) the ability to increase total lifetime consumption through saving (or consumption deferral) at a positive after-tax rate of interest, and (c) the desire by risk averse consumers to avoid years of relatively low consumption. Factors (a) and (b) tend to offset each other except, for example, where high effective marginal tax rates produce negative rates of return on retirement saving. Depending on how risk averse consumers are, factor (c) tends to diminish the effect of any imbalance between (a) and (b). A fairly typical set of assumptions might have a 3% rate of time preference, a 3.5% after-tax real rate of return on saving and a risk aversion parameter of 2. With these parameters, the optimal ratio C65/C64 would be 1.0024 implying a level of C65 that is one-quarter of 1% higher than C64. The 100% replacement target was chosen on the basis of simplicity, clarity and convenience. 100% Consumption Replacement (2). The target of full consumption continuity may be questioned for other reasons. If the lifecycle model were broadened to include leisure in the utility function, then the increased availability of leisure after retirement might allow people to smooth their marginal utilities over time while consuming somewhat less over retirement. If consumption and leisure were seen as complementary goods, on the other hand, then utility maximizers would want to increase their consumption levels after retirement. A variant of this idea is that consumption is a joint product involving time as well as money. With more time, retirees can shop more carefully and effectively, thus meeting consumption needs at reduced expenditure levels. They can also substitute home production in meal preparation and household maintenance for expenditures on food and other goods and services. These ideas have been examined as contributing solutions to what is known as the 'retirement consumption puzzle', the observation that expenditures (adjusted for changes in family-size) on the consumption of food and other non-durables follow a humped age profile with a decline around the age of retirement. As well as the substitutability of time for consumption expenditure, several other explanations have been put forward for this consumption profile. They include: a lack of foresight leading to failure to anticipate the size of income drops at retirement, a lack of commitment to saving even when income drops are correctly anticipated, and unplanned early retirement due to health problems. Recent studies based on U.S. data (Hurst, 2008, Aguiar and Hurst, 2008, and Hurd and Rohwedder, 2008 – the former two studies based on synthetic panel data and the latter on true panel data), provide several interesting results regarding consumption drops at retirement: • The drops found in the true panel data particularly are smaller than previous results, in the range of 1% to 6% on average; • The biggest drops are found in the lowest quartile of the wealth distribution and are related to health shocks; • The decline in work-related expenses is a significant factor. Expenditure items showing declines included nondurable transport and clothing and personal care – items that are inputs to market work; • Food expenditures also declined slightly, giving some support to the substitutability of time for spending on meals; • Spending on housing services did not decline; and • Spending on entertainment increased, suggesting that time and spending are in fact complementary in certain cases. This evidence casts some doubt on the materiality of the consumption puzzle and suggests that the inclusion of work-related expenses in the model may be sufficient to accord with the empirical findings. This allowance, set at$300 plus 3% of earnings, represents in most cases a drop of about 6% in observed spending at retirement in the model solution.

While 100% consumption replacement is taken as the base assumption in applying the model, results are also provided in all cases for a lower replacement target of 90%.

Retirement at Age 65. The assumption of retirement at age 65 suffers from the obvious problem that it does not accord with the average age of retirement (although that age has been rising since the mid-1990s). As noted earlier, assuming retirement at age 65 is virtually necessary to make the model tractable, given that it aims to capture most of the complexity of Canada's tax-transfer system. For example, assuming retirement at 65 allows OAS/GIS and the age and pension credits to apply only in retirement.

However, there is another important reason for assuming retirement at age 65. Public policy discussions in many countries share the conclusion that, given increasing longevity, it makes no sense to encourage early retirement and may be necessary to actively encourage later retirement. In this context, benchmarks devised to guide policy should not be based on retirement before age 65.

Savings from Age 30. The assumption of a 35-year accumulation period, from age 30 to 64, is clearly arbitrary. It may exaggerate savings accumulations and earnings replacement rates if people start saving later or if they save at lower than average rates early in their careers.

One offsetting factor is that some workers do belong to RPPs and/or contribute to RRSPs before age 30. RPP/RRSP contributions by those under age 30 (with RPP contributions measured by pension adjustment (PA) amounts) accounted for 7% of total RPP/RRSP contributions made by those under age 65 in 2006. Another point is that some households may focus on savings in home equity first and shift more toward RPP/RRSP saving later. Assuming that both savings processes are spread evenly over the 35-year period would involve offsetting errors in this case.

For a two-earner homeowner couple, Figure 3.11 shows the increase in savings rates necessary to compensate for a shorter accumulation period of 30 years. For a couple with combined earnings of $100,000, for example, the savings rate needed for consumption continuity would increase from 9.2% to 10.9%. Increased Longevity. Similarly, the assumption of a 20-year retirement period, while consistent with current average male and female life expectancy at age 65, may understate future life expectancy. For example, a study by the Office of the Chief Actuary of Canada (July 2009) projects an increase in the duration of CPP benefit receipt by 2050 of 3.3 years for males and 2.3 years for females. Figure 3.11 shows the effect of a 5-year increase in the duration of the retirement period on benchmark savings rates for the two-earner homeowner couple. The effects are somewhat less than those of a 5-year shorter accumulation period – at earnings of$100,00, an increase in the savings rate of 1.1 percentage points (from 9.2% to 10.3%) as compared to 1.7 percentage points.

Figure 3.11 Effect on Benchmark Savings Rates of a Shorter Accumulation Period and a Longer Retirement Period
Two-earner homeowner couple RPP/RRSP Contribution Rate as a per cent of Earnings

Lower Return on Savings. The assumed real rate of return of 3.5% derives from an assessment of pension plan returns over a long period. See Carty and Pressman (1979). It is reasonable in relation to the long-term real rate of return of 4.2% assumed for CPP assets. Diverse holdings of retirement savings should not be expected to generate the same rate of return as a very large fund like that of the CPP.

In the 1980s and 1990s, pension plan assets earned better than a 3.5% return, but there is a risk that returns will be lower over the coming decades as slow labour force growth leads to a rising capital-labour ratio that depresses the return to capital (Hviding and Mérette, 1998, Horner, 2009).

There is also evidence that rates of return can vary quite widely between pension plans and individual savings plans and even between larger and smaller employer-sponsored plans. Explanatory factors include economies of scale, investment fee differentials and differences in investment skill. In the US, Munnell et al (2006) found average returns on saving in individual retirement accounts (IRAs) of 3.8% per annum compared with rates of 6.6% and 5.6% for DB pension plans and 401(k) plans. Estimates for Quebec using a methodology similar to Munnell's suggest that the average annual rate of return on RRSPs from 1999 to 2005 was about 1.5% per annum as compared to a reported return of 6.7% for RPPs (Quebec, Régie des rentes, 2008). Sinha and Jog (2007) also found a 2 percentage point difference between the rates of return reported by equity mutual funds and those actually realized by investors trading in them.

Figure 3.12 shows, for the two-earner homemaker couple, the increase in the benchmark savings rates that result if a real rate of return of 2.5% per annum is assumed rather than 3.5%.

Figure 3.12 Effect on Benchmark Savings Rates of a 1% Lower Real Rate of Return on Savings
Two-earner homeowner couple RPP/RRSP Contribution Rate as a per cent of Earnings

A lower rate of return on savings substantially increases the savings rates required to achieve consumption continuity. The size of the increase rises with earnings simply because high earners need larger pools of savings.

While the savings rate increase is substantial – from 9.2% to 11.0% at earnings of $100,000, for example, it is perhaps not as great as one would expect when considering the difference between asset accumulations at 2.5% and 3.5% after a lengthy holding period. The explanation is that the higher savings needed for consumption replacement with a lower rate of return depresses pre-retirement consumption and so reduces the target level of private pension income. In the 2.5% return case, C64 and C65 are equated at a value of$63,730, about 3% lower than the base case value of $65,558. In Section 4, the adequacy of household savings levels in 2006 is assessed with a 2.5% rate of return as well as the base case assumption of 3.5%. Declining Role of OAS. The analysis has assumed that OAS, GIS and the CPP/QPP will maintain their earnings replacement effectiveness in the future. Since the model assumes real wage growth of 1% per annum, the implicit assumption is that benefit levels under these programs will increase at the same rate as the average wage, at least up to retirement age. This is a fair assumption for CPP/QPP benefits as the YMPE is indexed to the average wage. The future of benefit levels under OAS and GIS is less clear. In law, these benefit levels are indexed to prices, which would imply under our assumptions a 26% reduction in the benefit levels relative to earnings over the next 30 years. A look at the history of the programs over the last 40 years, however, shows that: • ad hoc benefit increases have served to prevent the OAS/GIS minimum income guarantee from failing to keep pace with real wage growth (of which, admittedly, there has been little), and • the adjustments have been made entirely through increases in the GIS maximums rather than OAS. It seems reasonable to project this political outcome rather than a continuing decline in the anti-poverty effectiveness of the programs. Under this hypothesis, the benchmark savings rates may understate true savings needs for all those who do not expect to receive GIS benefits in retirement. Status Changes Before Retirement. Tractability of the model requires that earnings grow at the same rate as the tax brackets and credits are indexed and that people do not shift among family types over time.11 These two simplifications introduce offsetting errors. Many earners, particularly at higher education levels will have age profiles of wages that embody faster growth than the 1% real growth in average wages assumed in the model. This means that for any given earnings level at age 64, they will have lower past earnings, lower past savings and lower future levels of private pension income than assumed in the model. This sort of effect has been found in unpublished simulations with Statistics Canada's LifePaths model that showed that many workers with pre-retirement earnings levels at or slightly above the YMPE would not qualify for full CPP/QPP benefits because of lower earnings levels when younger, or interrupted earnings histories. On the other hand, many of those found in a given year in the populations of singles or childless couples will either have had children when younger or will go on to have children when older. As the savings rates required for consumption continuity are relatively low for those with children this biases the savings targets upwards. ## 4. Household Retirement Savings in 2006 The foregoing framework is now applied to examine the level and pattern of household retirement saving in Canada. The analysis is based on saving in a single year. Possible biases due to this approach are considered in Section 5. The nature of the data is first described, then general information on the household population and its retirement savings is provided. Finally, savings levels are assessed in comparison to the benchmark levels developed in the last section. ### 4.1 The data The data are taken from the T1 family file prepared by the Canada Revenue Agency (CRA). This is based on a large stratified sample of tax returns with the added feature that information for both spouses in a couple is provided. The T1 file is the same as that used by the CRA in preparing its annual set of Income Statistics tables. It should be noted that the concept of family available in these data differs from the Census family or Economic Family definitions used by Statistics Canada. For example, a person living with an adult child reports for tax purposes as a single person. Information on the presence of children comes only from parental claims of child benefits or child care expenses. As in Section 3, six family types are distinguished: singles, one-parent families, one- and two-earner childless couples, and one- and two-earner two-parent families. In assessing adequacy, the savings benchmarks developed for two-child families are applied to all families with children. Earnings include employment income and net income from self-employment. For the purposes of this study, individuals (single or spouses) are treated as 'earners' only if they have earnings in excess of the federal basic personal amount ($8,839 in 2006).

There is no information on homeownership in the tax file. In applying the savings benchmarks to measure savings adequacy, populations by household type and broad earnings groups (as defined in Table 4.2 below) were divided into renters and homeowners based on data for the age 45-64 population in the 2005 Survey of Financial Security (SFS). The proportions of homeowners are set out in Table 4.1.

 Single 1PF Couple 2PF Earnings (%) Low 40.3 46.6 72.5 78.7 Modest 58.0 75.6 90.2 93.0 Middle 69.9 89.0 95.8 100.0 High 42.0 84.5 99.9 100.0

Source: Tabulation from the Survey of Financial Security; population age 45-64.
1PF = single-parent famil. ; Coupl. = couple without children; 2PF = two-parent family

Retirement saving is the sum of RPP and RRSP saving. RPP saving is measured as the Pension Adjustment (PA) amount, and RRSP saving as the RRSP contribution amount less any RRSP withdrawal. As it applies to defined benefit RPPs, the PA amount is an artificial measure of the combined employer-plus-employee contribution needed to fund benefits accrued in the year, calculated from pensionable earnings and the benefit formula. There is an argument that PAs tend to overstate RPP saving in the private sector where ancillary benefits are less generous than in the public sector and where workers who change jobs or become unemployed before retirement age more frequently lose benefits. The taxation data does not permit public and private sector employees to be distinguished, however, so no adjustments have been made to PA amounts.12

It should be noted that, for 2006 and other recent years, the PA measure of saving provides a much lower estimate of total RPP saving than actual aggregate contributions. The total of all PAs in 2006 amounted to $27.6 billion, less than two-thirds of total employer and employee contributions of$43.8 billion. The high level of contributions may be explained in good part by above-normal employer contributions needed to respond to plan deficits.

### 4.2 A profile of household saving

An advantage of this data source is that it permits savings patterns to be examined on a household or family basis, something that has not often been done in Canada.13 Combining income and savings information for single individuals and single parents, on one hand, and couples and two-parent families on the other, does present a challenge, though, as their income distributions differ appreciably. To permit this information to be summarized compactly, separate earnings groupings are used for two types of household. They are defined in Table 4.2.

Table 4.2
Earnings Classifications
Earnings Groups Singles and
1-Parent Families
Couples and
2-Parent Families

($000) Low 0 - 25 0 - 40 Modest 25 - 60 40 – 100 Middle 60 - 100 100 – 166.7 High 100 + 166.7 + This classification is less random than it might seem. Reference to Figures 3.1, 3.3, 3.5 and 3.7 shows that in general, private retirement savings needs (for consumption continuity) are low or zero in the Low earnings range. Savings needs rise over the 'Modest' earnings range and plateau over the top two earnings ranges. Also, the top earnings range contains households who may be affected by the dollar limits on RPP and RRSP saving. The RRSP dollar limit was$18,000 in 2006, which, based on the 18%-of-earnings limit, covered earnings of $100,000 for a single earner. For a two-earner couple, with a 60/40 earnings split, it covered earnings of$166,667. As Low earners have little need to save and High earners are likely to have reasonably high levels of retirement income, the focus of a savings adequacy analysis should be on households in the Modest and Middle earnings ranges.

Table 4.3 shows how Canada's population of households was distributed by family type and earnings range. This population is limited to those (for couples, the older spouse) between ages 30 and 64.

Single 1PF Couple 2PF 1-Earner 2-Earner 1-Earner Number (000) 3,131 603 1,118 1,228 957 1,708 % age 45-64 55.4 26.5 82.9 81.0 36.7 47.0 Distribution (%) Low 39.7 48.6 57.7 8.1 50.7 6.8 Modest 41.8 41.0 33.5 52.4 40.2 53.3 Middle 14.7 9.0 5.8 29.2 6.5 30.3 High 3.8 1.5 3.0 10.2 2.6 9.6 Total 100.0 100.0 100.0 100.0 100.0 100.0 Av Earnings ($) 39,309 30,627 51,711 107,514 53,944 106,309 1PF = single-parent famil. ; Coupl. = couple without children; 2PF = two-parent family The population of households headed by someone age 30-64 numbers 8.7 million. Singles account for about 36% of the total, two-parent families 30%, couples 27% and single-parent families 7%. About one-half of couples, and 60% of two-parent families, have two earners. Singles are divided fairly evenly between the younger and older age groups (age 30-44 vs. 45-64) as are 2-earner families with children. Single-parents and 1-earner two-parent families tend to be younger, while childless couples are concentrated in the older age group. Overall, about one-third of households are in the Low earnings group that has little need to save to obtain consumption continuity. This is consistent with the fact that GIS benefits continue to be paid to over one-third of Canadian seniors. About 5.4% are in the High earnings group whose RPP/RRSP contribution rates may be constrained by the dollar contribution and pension benefit limits. The Modest and Middle earnings groups account for about 62% of this population of households. Looking at the household types separately, we see that savings needs are lowest (i.e., the Low earner population biggest) among one-earner couples, one-earner two-parent families and single parents, and highest among two-earner families. Earnings Group Low Modest Middle High Total Single (%) Non-saver 76.0 28.4 9.8 9.1 43.8 RPP only 8.2 23.2 23.2 14.3 16.9 RRSP only 12.9 23.9 21.4 41.3 19.8 RPP + RRSP 2.8 24.5 45.6 35.2 19.4 Total 100.0 100.0 100.0 100.0 100.0 Single Parent Non-saver 78.5 28.2 8.1 9.3 50.5 RPP only 11.8 30.1 31.1 14.7 21.1 RRSP only 7.5 19.7 17.3 38.2 13.8 RPP + RRSP 2.2 22.0 43.5 37.8 14.5 Total 100.0 100.0 100.0 100.0 100.0 1-Earner Couple Non-saver 67.0 16.3 9.3 12.1 45.0 RPP only 10.7 26.6 20.9 20.1 16.9 RRSP only 15.6 21.7 25.6 40.9 19.0 RPP + RRSP 6.7 35.5 44.1 26.9 19.1 Total 100.0 100.0 100.0 100.0 100.0 2-Earner Couple Non-saver 53.9 15.0 3.5 3.9 13.7 RPP only 13.1 20.8 14.3 6.8 16.8 RRSP only 23.9 23.3 17.1 33.7 22.6 RPP + RRSP 9.2 41.0 65.1 55.7 46.9 Total 100.0 100.0 100.0 100.0 100.0 1-Earner 2PF Non-saver 68.1 20.0 8.8 6.9 43.3 RPP only 12.4 26.3 17.9 12.9 18.4 RRSP only 13.9 22.5 30.2 47.6 19.3 RPP + RRSP 5.5 31.2 43.1 32.6 19.0 Total 100.0 100.0 100.0 100.0 100.0 2-Earner 2PF Non-saver 58.0 15.9 3.7 3.5 13.9 RPP only 13.6 23.1 17.4 7.2 19.2 RRSP only 20.4 21.3 15.9 32.7 20.7 RPP + RRSP 8.0 39.7 63.1 56.6 46.2 Total 100.0 100.0 100.0 100.0 100.0 All Households Non-saver 71.4 21.2 6.1 5.9 34.3 RPP only 10.2 23.9 19.1 10.2 17.8 RRSP only 13.8 22.6 18.9 36.6 19.8 RPP + RRSP 4.5 32.4 55.9 47.3 28.1 Total 100.0 100.0 100.0 100.0 100.0 Table 4.4 provides the first information on savings patterns. For the six household types and four earnings ranges, it documents household participation in registered plan saving through RPP coverage and RRSP contributions. Households are classified as either: non-savers, RPP members only (i.e., having no RRSP contribution in the year), RRSP contributors only, or savers in both RPPs and RRSPs in a year. (For two-earner families, no information is provided on whether more than one spouse is an RPP member or RRSP contributor.) Looking at all household types together, the first observation that one might make is that RPP/RRSP participation seems to accord quite well with the savings needs framework developed in Section 3. Across the Low earnings group with low savings needs, 71% are non-savers. In contrast, nearly 80% of the Modest earnings group, and 94% of the Middle and High earner groups save in RPPs and/or RRSPs. That being said, the level of non-participation remains fairly high among those in the Modest earner group. In this earnings range, it is not surprising that non-saver rates are highest among singles (28.4%) and single parents (28.2%). Among savers in the Low earnings groups, RRSP saving is more prevalent than RPP saving. This may reflect the fact that voluntary RRSP contributions are not locked in to retirement age and so may be accessed in case of need. Furthermore, some RRSP contributions may be made expressly for early withdrawal under the Home Buyers Plan or the Lifelong Learning Plan. In the Modest and Middle earnings ranges, it is striking that quite large fractions of the household populations saved in both RPPs and RRSPs in 2006 – 32% among the Modest earners, 56% among the Middle earners.Not surprisingly, these proportions were highest for couples and two-parent families where there are two possible participants. ### 4.3 Assessment of savings adequacy Table 4.5 summarizes the evidence on savings adequacy provided by retirement savings patterns in 2006. It reports the percentages of households age 30-64, by family type and earnings range, whose saving rates exceed the 100% and 90% consumption continuity benchmarks defined in Section 3. For the total population of households, 69% meet the 100% target (C65 = C64), while 78% meet the less strict 90% benchmark (C65 = 0.9 x C64). These summary statistics, however, are not very informative. For one thing, most of those in the Low earnings range can achieve the targets without any need to save for retirement. On the other hand, many of those in the High earner group are constrained by the dollar limits on RPP/RRSP saving. They may well be achieving consumption continuity, but not by means of RPP/RRSP saving alone. Thus, it is best to focus on the results for households in the Modest and Middle earnings ranges.In these earnings ranges, the aggregate results are similar, with 60% of Modest-earnings households (63% for Middle) exceeding the 100% benchmark and 72% (71%) exceeding the 90% benchmark. Overall, this analysis suggests that in the modest/middle earnings range about 40% of households are not saving enough for full maintenance of their pre-retirement consumption, and about 28% face significant shortfalls in their living standards.  Single 1PF 1E C 2E C 1E 2PF 2E 2PF All % saving enough for 100% consumption replacement Low 83 100 100 100 100 100 91 Modest 44 77 72 52 83 78 60 Middle 56 78 64 56 66 71 63 High 24 56 37 57 51 74 61 All 61 88 79 56 82 74 69 % saving enough for 90% consumption replacement Low 95 100 100 100 100 100 97 Modest 52 91 86 71 94 96 72 Middle 65 82 71 66 72 78 71 High 35 63 45 67 57 79 68 All 71 94 86 68 88 82 78 1PF = single parent family, 1E C = 1-earner couple, 1E 2PF = 1-earner two-parent family Among family types in these middle earnings ranges, savings shortfalls from the 90% target are highest for singles (especially in the Modest earnings group) and, to a lesser extent, for two-earner couples. These results are not unexpected since the benchmark savings rates are considerably higher for singles and couples than the other family types. Also, older couples may be parents of children who have left home. By the logic of the consumption replacement model, they should be assigned the lower savings rate benchmarks of the two-parent family in recognition of the child-related expenses that have reduced their 'own consumption' consumption levels before retirement. As well as having the highest rate of savings shortfalls, singles and two-earner couples make up the bulk of the population in the Modest and Middle earnings groups that do not meet the lower 90% consumption replacement target. As shown in Figure 4.1, households of these two types together account for nearly three-quarters of those with significant savings shortfalls. Figure 4.1 Distribution by Household Type of Those Not Meeting the 90% Savings Rate Benchmark Population in the Modest and Middle Earnings Groups Another issue of policy interest concerns savings adequacy among RRSP-only savers compared to those who belong to RPPs. Table 4.6 provides the savings adequacy report card for the total population of households broken down into the categories: non-saver, RPP-only saver, RRSP-only saver and saver in 'Both' (RPP and RRSP). For the Modest and Middle earnings groups, Tables A.1 and A.2 in Annex A provide a breakdown of the results by family type. Table 4.6 shows that, as must be expected, RPP members who also contribute to RRSPs are the most likely group to achieve consumption continuity. For all household types together, 82% of them exceed the 100% savings rate benchmark as compared to 72% for those saving in RPPs alone and 61% for those saving in RRSPs alone. The 11 percentage-point difference between the proportions of RPP-only and RRSP-only savers that meet the 100% savings benchmark is also important. (For the 90% benchmark, the gap is 12 percentage points.. Tables A.1 and A.2 in Annex A show that among Modest earners, the gaps are biggest for singles and single parents, while among Middle earners the gaps are biggest for couples and two-parent families.  Non-Saver RPP only RRSP only Both All % saving enough for 100% consumption replacement Low 91 96 95 96 91 Modest 29 75 62 87 60 Middle 0 64 49 84 63 High 0 52 41 76 61 All 63 72 61 82 69 % saving enough for 90% consumption replacement Low 98 99 98 98 97 Modest 46 86 74 92 72 Middle 0 77 59 90 71 High 0 63 50 83 68 All 73 82 70 88 78 This difference in savings levels between RPP-only and RRSP-only savers does not tell the whole story, because it ignores the role of RPP-plus-RRSP savers. Do workers without RPP coverage compensate by contributing more frequently, and at higher levels, than those who belong to pension plans. The answer is mostly 'No'. From the incidence figures provided in the 'All Households' panel of Table 4.4, one can calculate that, among those in the Modest earnings range, the likelihood of being an RRSP contributor in 2006 was 57.5% for RPP members but only 51.2% for those without RPPs. Among those in the Middle earnings range, the RRSP-contributor proportions were almost as high for RPP members as for non-members – 73.6% for members and 75.6% for non-members. RRSP contributors who do not belong to RPPs do contribute more on average to RRSPs than do contributors who are RPP members. The difference is not great enough, however, to support the idea that RRSP-only contributors are achieving savings levels similar to those of RPP members. Table 12 in the CRA's Income Statistics for 2006 provides a basic indicator.14 The average level of RPP savings (as measured by the PA) in 2006 was$5,188. The average RRSP contribution level among those without RPPs was $5,790. Among RPP members with RRSP contributions, the average RRSP contribution level was$4,157.

Figure 4.2 shows how those in the Modest and Middle earnings groups who do not meet the 90% consumption replacement target are distributed by saver category. Half of those with significant savings shortfalls had no RPP/RRSP savings in 2006; another quarter saved in RRSPs only.

Figure 4.2 Distribution by Saver Category of Those Not Meeting the 90% Savings Rate Benchmark
Population in the Modest and Middle Earnings Groups

This analysis lends support to the idea that RPP and RRSP saving are not pure substitutes, and that we cannot expect RRSP saving to fully fill the gap that will be left if RPP coverage continues to decline. This conclusion is consistent with old ideas, such as the hypothesis of Cagan (1965) that the provision of a base level of retirement income under a pension plan encourages supplementary saving by making pension income targets seem more attainable, as well as more recent suggestions that weaknesses in savings choices by individuals who rely on voluntary plans provide a potential role for 'soft paternalism' in retirement savings systems. Soft paternalism refers to ideas such as automatic enrolment in savings plans with default contribution levels and default investment options. See, for example, the seminal paper by Choi et al (2001).

The preceding analysis ignores three factors that may demand qualification of the results. First, households may have savings outside of registered plans and apart from homeownership that reduce their need for RPP/RRSP saving. Second, the relatively low savings levels among RRSP-only savers may be misleading because the group includes self-employed workers who are generally ineligible for RPP membership and who may focus their saving on their business investments. Third, as noted earlier there are risks that future rates of return, on all retirement savings or on RRSP saving in particular, could fall below the assumed rate of 3.5%.

### 4.4 Other Assets

Figure 4.3 provides an idea of the relative importance of assets outside registered plans for those in the four broad earnings groups. Since the focus is on earnings replacement, the asset levels are expressed as ratios to earnings. The ratios are based on total asset amounts and total earnings in each group regardless of the distribution of assets within the group. The population is limited to those without business assets. The asset classes are as follows:

• RPP and RRSP assets,
• Home = value of principle residence less mortgage on it,
• Durables = 'other real assets',
• Other real estate = value of property less 'other mortgage', and
• Other financial = deposits plus other financial assets less non-mortgage debt.

Figure 4.3 Ratio of Assets to Earnings – Various Net Assets
Households without Self-Employment Income

Source: Tabulation from the 2005 Survey of Financial Security

In the Modest and Middle earnings groups, the picture is quite similar. Assets sheltered from taxation – RPP/RRSP assets, home equity and durables – account for over 90% of net worth. Other financial assets, net of consumer debt, account for 3%-4% of net worth. It may also be noted that, while deposits are very widely held, other financial assets are held by about 40% of Modest earners and 54% of Middle earners. And, judging by the high ratios of average to median amounts, these assets are concentrated in relatively few hands.15

In the High earnings group, other financial assets are considerably more important, accounting for 12.5% of net worth. The dollar limits on RPP/RRSP saving provide one explanation for this.

These observations on the level and distribution of unsheltered financial assets suggest that taking them into account would have only a minor effect on the picture of whether Modest/Middle earners are saving enough for retirement. They do play a bigger role, though, among higher income earners affected by the RPP/RRSP limits.

### 4.5 Savings among the self-employed

If self-employed workers are more likely to invest in their businesses than save in RPPs or RRSPs, then their presence in the population may lead to a downward bias in the estimated proportions of households meeting the savings rate benchmarks. Also, since the self-employed are not permitted to sponsor RPPs for themselves, a preference for investment in their businesses over RRSP saving could distort the comparison of savings adequacy between RRSP-only savers and those with RPPs.

Annex B provides a profile of RPP/RRSP saving among the self-employed in 2006. From 2005 SFS data, it also provides evidence of the importance of 'other assets', including net business equity, parallel to that provided for employee households in Figure 4.3 above.

For the purposes of the RPP/RRSP saving profile, the self-employed are identified in the T1 data file as households with more than $8,839 (the basic personal amount in 2006) of net income from self-employment. Self-employment income includes professional, business, commission, farming and fishing income. In the age group 30-64, the self-employed, so defined, made up 9.4% of households. The earnings of the self-employed are less evenly distributed than those of employees, so the self-employed accounted for only 8.0% of households in the Modest and Middle earnings groups. Table B.1 shows that, while RPP membership is substantially lower among the 'self-employed' households than those without self-employment income. For Modest earners, 31.9% (aggregating over the 'RPP-only' and 'Both' categories) of self-employed households saved in RPPs as compared to 58.4% of employee households. For Middle earners, the proportions were 43.5% and 77.6%. While the gaps are significant, it is the incidence of RPP membership among self-employed households that may appear surprising.There are two explanations. First, among couples, one spouse may be self-employed while the other is employed. Second, even among singles, many self-employed workers also have employment income. Overall, among self-employed households, self-employment income accounts for only 56.0% of total earned income in the Modest earnings group 52.4% in the Middle earnings group. The incidence of RRSP saving was slightly higher among self-employed than employee households – 59.0% vs. 54.6% for Modest earners and 79.6% vs. 74.4% for Middle earners. Despite the partial compensation of higher RRSP participation, the proportions of non-savers (in RPPs and RRSPs) were higher for the self-employed than the employed – 30.3% vs. 20.4% and 12.5% vs. 5.6% for the two earnings groups. Table B.2 compares RPP/RRSP savings rates for savers in self-employed and employee households for the four categories of saver. The last column of the Table summarizes the results, comparing the average savings rates across savers and non-savers together for the two populations. The differences are surprisingly small. Among Modest earners, the average savings rate is 8.5% for the self-employed and 8.8% for employee households; among Middle earners the rates are 11.8% and 12.3%. One reason why the differences are small is that contribution rates among RRSP-only savers are considerably higher for self-employed than employee households – 11.6% vs. 7.4% for Modest earners and 12.8% vs. 10.0% for Middle earners. This profile has two implications for the savings adequacy estimates presented in Section 4.3. First, because the self-employed population is relatively small, because the incidence of RPP/RRSP saving is not that much lower among the self-employed than employees, and because the gap in contribution rates between the self-employed and employed is small, the presence of self-employed households is unlikely to have a material affect on the overall estimates of proportion of household meeting the savings rate benchmarks. Second, among RRSP-only savers, self-employed contributors have considerably higher savings rates than employees. Thus, the presence of the self-employed does not explain the relatively low proportions of RRSP-only savers that meet the savings rate benchmarks. Annex Figure B.1 shows asset levels, as ratios of income, by earnings group for self-employed households. For these households, it shows that business assets are a very important component of total wealth, accounting for about 40% of it in both the Modest and Middle earnings groups. It is also evident that total wealth levels are much higher among self-employed than employee households. Even after subtracting business assets, self-employed households have higher net worth, on average, than employee households. This information does suggest that, unlike the case of net financial assets among employee households, not taking into account other assets may understate the retirement preparedness of those with business assets. Another point here is that the proportion of households reporting business assets is over twice as high as the proportion of households with self-employment income (above$8,839) in the taxation data. However, care is needed in assessing the weight of this information because wealth, and business equity in particular, is extremely concentrated. For example, while average business equity values in 2005 were $197,000 and$402,000 for self-employed households in the Modest and Middle earnings ranges, the median values were $20,000 in both cases. Thus, most households reporting business equity had insignificant amounts of it. ### 4.6 The effect of a 2.5% rate of return As noted in Section 3.3, the slow labour force growth projected for most OECD countries over the next two or three decades could put downward pressure on the return to capital and thus on the returns available on pension funds and RRSP savings. Another suggestion is that lower rates of return should be assumed for savings in individual plans than for savings in larger-scale professionally invested RPPs. Replicating the savings adequacy analysis of Section 4.3 with an assumed real rate of return of 2.5%, rather than 3.5%, permits the consequences of such outcomes to be considered. The results are summarized here. Detailed results parallel to those provided in Tables 4.5, 4.6, A.1 and A.2 are provided in Annex C. (No estimation has been carried out with lower returns on RRSP saving only, but the results of such a case may be inferred from the results presented by category of saver.) In considering the results, note that the lower rate of return is assumed to apply to all saving, including that through homeownership. This explains what would seem to be a puzzling result – namely, that for some households in the Low or Modest earnings groups, the drop in the rate of return leads to a reduction in the benchmark savings rates and an improvement in estimated savings adequacy. What happens is that, for homeowners, the cost of financing the assumed home purchase rises. The extra cost reduces pre-retirement consumption and so reduces the target level of post-retirement consumption. Consequently, some households with RPP/RRSP savings needs at a rate of return of 3.5% no longer need to save in RPPs or RRSPs at the 2.5% rate. Table 4.7 provides a summary comparison across earnings groups of the savings adequacy results with 3.5% and 2.5% rates of return. 100% 90% 3.5% 2.5% Change 3.5% 2.5% Low 91 91 0 97 97 0 Modest 60 57 -3 72 70 -2 Middle 63 55 -8 71 66 -5 High 61 51 -10 68 60 -8 All 69 65 -4 78 75 -3 These results show that, as with the effect on benchmark savings rates portrayed in Figure 3.11, the effects of a lower rate of return increase markedly as the required savings rate increases – with the level of earnings and with the level of the post-retirement consumption target (i.e., 100% vs. 90% of C64). In the Modest earnings range, the impact on the savings adequacy results is not as large as might have been expected. This is partly the due to the increased homeownership savings assumed for a substantial portion of the group and partly due to the fact that, by reducing pre-retirement consumption, any increase in savings reduces the level of post-retirement consumption required for consumption continuity. The effects of a lower rate of return do become quite significant, however, among those in the higher earnings ranges. As RRSP savings account for somewhat over half of total RPP and RRSP saving, a rate of return reduction applying to RRSP saving only would have slightly more than half the effects shown in Table 4.7. ## 5. Evidence from Longitudinal Data The preceding analysis has the key limitation that it uses savings patterns in a single year as the basis for judgments about pension accumulation and payout processes that last a full adult lifetime. Clearly, many people do not belong to an RPP for their full career or contribute a constant rate of earnings to their RRSPs. No data exists that would allow savings rates to be tracked over a full career, but there is beginning to be some evidence from longitudinal data on multi-year patterns in RPP membership and RRSP saving. Regarding pension coverage, Drolet (2009) provides evidence from the Longitudinal Administrative Data base (LAD) maintained by Statistics Canada that of RPP members who reported a PA in both 2001 and 2002, 88.9% reported PAs in at least five of the preceding ten years, and 67.3% reported PAs in all ten years. For those who belonged to an RPP in either, but not both, of the years 2001 and 2002, the previous decade's participation rates were somewhat lower – 79.7% for at least five years and 58.4% for the full ten years. The preceding evidence suggests that among RPP members, membership status is reasonably stable over time. On the other hand, Drolet also reports evidence that job tenure in general has been declining, particularly among men. Between 1987 and 2007, for example, the percentage of males age 35-44 who had been in the same job for over ten years declined from 40.5% to 30.9%. With regard to the consistency of RRSP contributions over time, Table 5.1 presents data similar to Drolet's RPP data, though only for a five-year period. Across earnings ranges, it shows the percentages of tax filers who contributed to RRSPs at least once during the five-year period, 2002-2006, and at least three times during those years. The final column gives the probability of contributing for at least three years conditional on having contributed at least once in the period. RRSP Contribution Probabilities, 2002 - 2006 Earnings ($000) 1+ Year(s) 3+ Years 15 - 25 61.8 39.1 63.3 35 - 45 71.0 51.2 72.1 45 - 55 76.5 58.5 76.5 55 - 65 79.8 63.4 79.4 65 -75 82.9 67.8 81.8 75 - 85 85.3 71.3 83.6 85 -95 87.9 74.8 85.0

Source: Tabulation, provided by the Department of Finance, of data from the longitudinal T1 file.

These statistics show, to begin with, that it is an exaggeration to assume that workers contribute every year to their RRSPs. On the other hand, it also shows that there is a considerable degree of consistency in RRSP usage. In the $65,000-$75,000 earnings range, for example, 81.8% of those who made at least one contribution in the five year period contributed for three years or more. Moreover, it should be noted that these statistics include those whose RRSP contributions are supplemental to saving in RPPs as well as those saving in RRSPs only. Finally, note as well that, because workers don't make RRSP contributions every year, some of the households shown as non-savers in the data analyzed in Section 4 will have contributed to RRSPs in other years.

LaRochelle-Côté, Myles and Picot (LMP, 2008) have used longitudinal tax data to provide evidence directly on the earnings replacement rates achieved by Canadians who have moved from the workforce into retirement over the past two decades. In the results they present, they focus on the cohort of individuals who were in the age group 54-56 in 1983, and they follow their income histories through to 2005. They base their replacement rates on after-tax family income, and, as they can track changes in family size over time, they use family-size adjusted income (equal to family income divided by the square root of family size).

This is a very valuable study, and it is worth reporting some results that are most relevant to this paper. Table 5.2 extracts information from Table 6 of the paper. It shows the distribution of replacement rates obtained by those age 69-71 (in 1998) who were in the bottom, middle and top earnings quintiles (Q1, Q3 and Q5) when they were age 54-56 in 1983.

Distribution Across RR Ranges (%) Replacement Rate Range Q1 Q3 < 40% 0.1 1.0 7.5 40% - 60% 1.4 23.3 28.7 60% - 80% 19.4 38.8 34.6 80% - 100% 28.1 21.4 14.8 100% - 150% 35.0 12.9 10.2 > 150% 16.0 2.6 4.2 Total 100.0 100.0 100.0

Source. LaRochelle-Côté, Myles and Picot (2008), Table 6, p. 29.

In interpreting these results, it must be recognized that the replacement rate concept used is intermediate between the standard concept, based on pre-tax incomes, and the consumption replacement targets used in the analysis in Section 4 of this paper. Table 5.3 presents calculations from the stylized lifecycle model that show how the three measures compare for those in the Modest earnings group, a group that compares closely with earnings quintile 3 in LMP. The calculations are averaged across the six household types, and across owners and renters within each household type.

Table 5.3
Comparison of Replacement Rates Modest earners - average for all household types
Consumption After-tax Income Before-tax Income

(%)
100.0 75.2 63.6
90.0 69.2 56.8

The LMP results presented in Table 5.2 appear reasonably congruent with, but less positive than, the results of the savings adequacy analysis in this paper. For quintile 3 earners, a linear interpolation within the 60%-80% replacement rate range (into below 69.2%, between 69.2% and 75.2%, and above 75.2%) provides a rough estimate of the proportions not meeting the 100% and 90% consumption replacement targets. The results suggest that about 54% have not met the 100% benchmark and 42% are not meeting the 90% target. The corresponding figures for Modest earners in Section 4.3 of this paper are 40% and 28%.

For those in the bottom earnings quintile, most have achieved an 80% after-tax earnings replacement rate. Among those in Q5, replacement rates are lower, with 70.8% of families below 80% and 36.2% below 60%.

In another study based on the LAD file, Ostrovsky and Schellenberg (2009) compare the income levels at age 70-72 of individuals who either belonged, or did not belong, to an RPP when they were age 55-57 in 1991. They make this comparison separately for men and women by earnings quintile and supplement their tabular results with regression analysis to control for the effects of earnings levels, age of retirement, changes in marital status and other variables.

They present two very interesting summary findings. First, those without RPP coverage in their mid-50s were significantly more likely to still be working at age 70-72. Second, among those retired by age 70, pre-tax income replacement rates did not differ significantly between those who were, or were not, RPP members in their mid-50s.

Obviously, the first of these findings qualifies the second to some extent. Non-members of RPPs were less likely to feel that they had sufficient resources to retire by age 70.16 Nevertheless, the second finding is an important one as it is at variance with the pattern of savings adequacy – lower among RRSP-only savers than among those with RPPs – found in the present analysis of savings in 2006.

The main explanation for this result must be that, despite Drolet's evidence of a degree of consistency in RPP membership from year to year, over a full career many workers are RPP members in some years and RRSP-only savers in others. The two categories become more blurred as the period of observation lengthens.17 A second explanation is that, over the years 1980 to 2006, rates of return on typical savings portfolios were well above their longer-term average. This has favoured saving in defined contribution plans and RRSPs over that in defined benefit RPPs. A third factor that may have contributed to the result is that many RPP members, particularly in the private sector, often lose significant benefits when they change jobs or become unemployed.18

These explanations notwithstanding, some caution is required in interpreting the results. As noted above, RPP non-members are more likely than members to still be working at age 70. Also, the market incomes (i.e., incomes net of OAS/GIS, CPP/QPP and other government benefits) of retirees at age 70 who were not in RPPs at age 55-57 are found to be almost the same as those of former RPP members partly because the non-members have higher levels of investment income from other financial assets. But, the distribution of these assets is very skewed, so the comparison of average income levels may mask a result that more non-members than members end up with low incomes in retirement. In fact, the Ostrovsky and Schellenberg results provide evidence of this. Looking at the males in earnings quintiles two and three in their Table 2, for example, the average level of GIS benefits (a good indicator of low-income status) is about twice as high for non-members as for members of RPPs.

## 6.Savings Adequacy in the Future

Assessments of earnings replacement or savings adequacy generally focus mainly on those at or nearing retirement age. There is simply less information available to judge how well younger workers are likely to fare in the future in building adequate retirement incomes.

There are some trends or factors in the current environment that are likely to affect retirement income outcomes in the future. They deserve consideration and assessment. Three examples are (a) the effects of the 2008 stock market crash and economic recession, (b) the continuing decline in pension coverage, and (c) the introduction of the new savings vehicle, the TFSA .

### 6.1 Stock market crash and recession

It is too early to assess the longer-term effects of these shocks. As seen in the data on personal sector assets, there seems to have been a strong bounce-back in many asset values. It is clear, though, that the stock market crash has done serious and permanent damage to the retirement savings of many people. It has also made many employer-sponsored pension plans less viable, and so increased the likelihood of plan changes such as benefit cuts, conversion from defined benefit to defined contribution, or plan termination.

The economic shocks following the financial markets crash are still playing out, and it is difficult to determine how serious and long lasting the effects on employment may be. While our unemployment rate has jumped from 6.0% in 2007 to 8.6% in October, 2009, the labour market consequences of the crash have, so far, been less serious in Canada than in many other countries.Coile and Levine (2009) predict that forced early retirement, rather than workers' adjustments to financial asset losses (e.g., through delayed retirement), are likely to be the strongest effect of the current economic crisis.

### 6.2 Declining pension coverage

There has been a longstanding trend to lower pension coverage in Canada, at least among men. Drolet (2009) finds that declines in the rate of union membership in the labour force is the most important factor explaining the decline. Changes in employment composition towards sectors like 'other services' with low pension coverage are another important cause. There is no strong evidence to suggest that the trend to declining RPP coverage will not continue in the future. The effects of the current economic turmoil could accelerate it.19

The savings adequacy analysis of Section 4 showed significant differences in retirement savings rates between members of RPPs and those saving in RRSPs only. Thus, if the population of households used in that analysis were re-weighted to simulate the effects of a further decline in RPP coverage, the effect would be to lower the predicted proportion of households meeting the savings adequacy targets.

### 6.3 Introduction of the TFSA

The introduction of the new savings vehicle, the TFSA, is a more positive development for savings adequacy in the future.

The consumption continuity model used in Section 3 to develop savings adequacy benchmarks can be applied to investigate which households (which family types at what earnings levels) would be better off saving through TFSAs than RPPs or RRSPs. These results can be applied in turn to examine the potential effects of the new vehicle on the achievement of adequate savings rates, on aggregate levels of saving, and on future levels of public pension costs and tax revenues.

The key result of the lifecycle model analysis is that households of all types, over a certain range of earnings levels, would be better off saving through TFSAs than RPPs or RRSPs. Where individuals or families are better off saving with a TFSA, they can achieve a higher level of after-tax retirement income with the same level of saving as for RPPs/RRSPs. However, in the lifecycle model where the aim is to equalize pre- and post-retirement consumption, the result is that they will reduce their savings levels slightly, attaining higher consumption levels than with RPP/RRSP saving both before and after retirement.

The reason that TFSA saving is superior to RPP/RRSP saving for these households is that the retirement income is both tax-free and excluded from the income used to determine eligibility for GIS and income-tested tax credits. In fact, in the model solutions, those that are better off saving with TFSAs all share the characteristics that they have low or zero income tax liabilities in retirement and that they receive substantial benefit levels under the GIS. The net gain for the households from this result is offset to a considerable degree, though, by the fact that, since TFSA contributions are not tax-deductible, they pay higher taxes while working. Table 6.1 illustrates these points by comparing savings, consumption, income tax, GIS and refundable credit levels under the lifecycle model for TFSA and RRSP saving for a single renter with earnings of $70,000.  TFSA RRSP ($) Contribution 6,207 9,520 Net saving (after tax deduction) 6,207 6,409 Consumption (C65 = C64) 43,447 43,245 Fed + Prov income tax (age 30-64) 15,305 12,194 Fed + Prov income tax (age 65+) 224 9,621 GIS 2,268 0 GST credit and provincial credits 1,547 265

The key points from Table 6.1 are that, for the single renter earning $70,000, TFSA saving provides, as compared to RPP or RRSP saving: • An increase of$202 per year, or 0.5%, in annual consumption;
• An increase in income tax while working of $3,111 per year; • A drop in income tax after age 65 of$9,397;
• An increase in GIS benefits of $2,268: and • An increase in other refundable tax credits of$1,282.

Note, though, that the tax increases experienced while working under the TFSA option extend over a longer period than the lower taxes and higher benefits in retirement. Also, they come earlier and so carry more weight when the fiscal effects are evaluated on a present value basis.

The households found to be better off saving with TFSAs are those with earnings up to about $62,000 for two-earner couples (with or without children) and$75,000 for singles, single parents and one-earner couples. Together, they account for 3.6 million households or about 40% of the population of households with heads age 30-64. While full take-up of TFSAs by those who would benefit from using them is unlikely, it is safe to predict that the introduction of the new vehicle will significantly alter savings patterns and have important effects on government revenues and program costs.

It is also likely that the attractiveness of TFSA tax treatment to a large segment of the population will result in pressure to allow TFSA tax treatment of some RPPs, or for the replacement of some pension plans by group TFSAs.

## Annex A Further Savings Adequacy Results

 Non-Saver RPP only RRSP only Both All Single (%) Modest 0 62 42 82 44 Middle 0 48 41 80 56 1PF Modest 42 95 81 95 77 Middle 0 90 56 94 78 1E Couple Modest 39 88 81 92 72 Middle 0 67 49 85 64 2E Couple Modest 22 62 64 80 52 Middle 0 47 46 76 56 1E 2PF Modest 69 92 87 94 83 Middle 0 79 49 87 66 2E 2PF Modest 62 87 82 95 78 Middle 0 76 56 90 71 All Modest 29 75 62 87 60 Middle 0 64 49 84 63

1PF = single parent family, 1E C = 1-earner couple, 1E 2PF = 1-earner two-parent family

 Non-Saver RPP only RRSP only Both All Single (%) Modest 0 77 54 89 52 Middle 0 66 48 87 65 1PF Modest 79 98 92 97 91 Middle 0 94 65 95 82 1E Couple Modest 71 95 90 96 86 Middle 0 78 57 90 71 2E Couple Modest 51 80 79 88 71 Middle 0 64 55 85 66 1E 2PF Modest 90 97 96 98 94 Middle 0 86 60 92 72 2E 2PF Modest 92 98 97 99 96 Middle 0 86 70 94 78 All Modest 46 86 74 92 72 Middle 0 77 59 90 71

1PF = single parent family, 1E C = 1-earner couple, 1E 2PF = 1-earner two-parent family

## Annex B RPP/RRSP Savings among the Self-Employed

 Non-Saver RPP only RRSP only Both All Self-Employed (% Distribution) Low 71.3 3.0 23.7 2.0 100.0 Modest 30.3 10.7 37.8 21.2 100.0 Middle 12.5 7.9 44.1 35.5 100.0 High 6.9 3.3 59.1 30.7 100.0 All 39.5 6.7 36.3 17.6 100.0 Employed Low 71.4 11.1 12.7 4.8 100.0 Modest 20.4 25.0 21.2 33.4 100.0 Middle 5.6 20.0 16.8 57.6 100.0 High 5.6 12.1 30.6 51.7 100.0 All 33.3 19.1 18.5 29.2 100.0

 Non-Saver RPP only RRSP only Both All Self-Employed (% of Earnings) Low 0.0 6.8 13.8 18.4 3.8 Modest 0.0 7.2 11.6 16.0 8.5 Middle 0.0 7.6 12.8 15.5 11.8 High 0.0 4.4 7.2 8.6 7.0 All 0.0 6.0 12.3 18.1 8.1 Employed Low 0.0 14.1 12.8 29.5 4.6 Modest 0.0 9.0 7.4 15.1 8.8 Middle 0.0 9.9 10.0 15.0 12.3 High 0.0 6.3 7.9 9.5 8.1 All 0.0 9.5 9.7 20.0 9.4

Note. Percentages are based on estimated earnings in the saver category. E.g., the 6.8% savings rate of RPP-only savers in the Low earnings self-employed group is based on the earnings of those in the group that belong to RPPs but do not contribute to RRSPs.

Figure B.1 Ratio of Assets to Earnings – Various Net Assets
Households with Self-Employment Income

Source: Tabulation from the 2005 Survey of Financial Security

## Annex C Savings Adequacy Results with a 2.5% Rate of Return

Single 1PF 1E C 2E C 1E 2PF 2E 2PF All Per cent saving enough for 100% consumption replacement Low 82 100 100 100 100 100 91 Modest 37 78 74 50 87 78 57 Middle 39 75 57 47 61 68 55 High 11 47 26 43 42 67 51 All 54 88 78 47 82 70 65 Low 95 100 100 100 100 100 97 Modest 48 90 85 70 94 96 70 Middle 54 79 66 60 69 76 66 High 20 36 36 57 50 74 60 All 66 84 84 62 87 80 75

1PF = single parent family, 1E C = 1-earner couple, 1E 2PF = 1-earner two-parent family

Non-Saver RPP only RRSP only Both All Per cent saving enough for 100% consumption replacement Low 91 96 95 96 91 Modest 34 66 58 81 57 Middle 0 50 42 76 55 High 0 39 32 66 51 All 65 61 56 74 65 Low 98 99 99 99 97 Modest 46 82 71 90 70 Middle 0 68 54 85 66 High 0 53 41 75 60 All 73 76 66 83 75

Note: in the 100% consumption replacement case, the percentage of Modest-earner non-savers that meet the savings targets at a 2.5% rate of return on saving (34%) exceeds the corresponding percentage (29%) when the rate of return is 3.5%. The reason for this is that in the 2.5% rate of return case, homeowners have higher level of homeownership savings. This translates into lower pre-retirement consumption levels and lower savings targets. According, a higher proportion of households meets their targets without any need for RPP/RRSP saving.

 Non-Saver RPP only RRSP only Both All Single (%) Modest 0 44 35 73 37 Middle 0 16 31 62 39 1PF Modest 57 93 75 93 79 Middle 0 85 48 93 75 1E Couple Modest 57 83 77 89 74 Middle 0 56 43 79 57 2E Couple Modest 23 57 61 75 50 Middle 0 33 40 66 47 1E 2PF Modest 79 92 87 95 87 Middle 0 73 43 83 61 2E 2PF Modest 62 86 82 94 78 Middle 0 70 52 88 68 All Modest 34 66 58 81 57 Middle 0 50 42 76 55

1PF = single parent family, 1E C = 1-earner couple, 1E 2PF = 1-earner two-parent family

 Non-Saver RPP only RRSP only Both All Single (%) Modest 0 68 49 84 48 Middle 0 44 40 78 54 1PF Modest 79 98 89 96 90 Middle 0 91 57 94 79 1E Couple Modest 71 92 86 95 85 Middle 0 70 51 86 66 2E Couple Modest 51 78 77 86 70 Middle 0 54 50 80 60 1E 2PF Modest 90 97 96 98 94 Middle 0 83 56 90 69 2E 2PF Modest 92 98 98 99 96 Middle 0 84 68 93 76 All Modest 46 82 71 90 70 Middle 0 68 54 85 66

1PF = single parent family, 1E C = 1-earner couple, 1E 2PF = 1-earner two-parent family

## References

Aguiar, Mark and Erik Hurst, "Deconstructing Lifecycle Expenditures", National Bureau of Economic Research, Working Paper 13893, March 2008.

Cagan, D., The Effect of Pension Plans on Aggregate Saving, National Bureau of Economic Research, 1965.

Carty, E. Bower and Irwin Pressman, "Methodology and Sources for Calculating Rates of Return", Appendix 10 of The Retirement Income System in Canada: Problems and Alternative Policies for Reform, Government of Canada, Task Force on Retirement Income Policy, Volume 2, 1979.

Choi, James J., David Laibson, Brigitte C. Madrian and Andrew Metrick, "Defined Contribution Pensions: Plan Rules, Participant Decisions, and the Path of Least Resistance", National Bureau of Economic Research, Working Paper 8655, December 2001.

Drolet, Marie, "Pension coverage in Canada", Statistics Canada, mimeo, October 1, 2009.

Coile, Courtney and Phillip B. Levine, "The Market Crash and Mass Layoffs: How the Current Economic Crisis May Affect Retirement", National Bureau of Economic Research, Working Paper 15395, October 2009.

Engen, E. M., W. G. Gale and C. E. Uccello, "The Adequacy of Household Saving",Brookings Papers on Economic Activity, 2:1999.

Finance Canada, Pension Reform: Improvements in Tax Assistance for Retirement Saving, a guide to pension reform tabled in Parliament by the Minister of Finance, December 1989.

Finance Canada, "The Financial Position of the Personal Sector in Canada", Economic Analysis and Forecasting Division, November 2000.

Finance Canada, "The Financial Situation of Canadian Households" Julie Turcotte and Peter Robinson, Economic Analysis and Forecasting Division, October 2002.

Horner, Keith, "Savings Incentives and OAS/GIS Costs", Canadian Public Policy, XXXIV Supplement, November 2008.

Horner, Keith, "Approaches to Strengthening Canada's Retirement Income System", Canadian Tax Journal, Vol. 57, No. 3, November, 2009.

Hurd, Michael D., and Susan Rohwedder, "The Retirement Consumption Puzzle: Actual Spending Changes in Panel Data", National Bureau of Economic Research, Working Paper 13929, April 2008.

Hurst, Erik, "The Retirement of a Consumption Puzzle", National Bureau of Economic Research, Working Paper 13789, February 2008.

Hviding, Ketil and Marcel Mérette, "Macroeconomic Effects of Pension Reform in the Context of Aging: OLG Simulations for Seven OECD Countries", OECD Working Paper No. 201, June 1998.

LaRochelle-Côté, Sébastien, John F. Myles and Garnett Picot, "Income Security and Stability During Retirement in Canada" Statistics Canada, Analytical Studies Branch Research Paper Series, Catalogue no. 11F0019 – No. 306, 2008.

Milligan, Kevin, "Life-cycle asset accumulation and allocation in Canada", Canadian Journal of Economics, Vol.38, No. 3, August 2005.

Morissette, René and Xuelin Zhang, "Revisiting wealth inequality", Perspectives, Statistics Canada, Catalogue 75-001, December 2006.

Morissette, René and Yuri Ostrovsky, "Pensions and retirement savings of families", Perspectives, Statistics Canada, Catalogue 75-001, November 2007.

Munnell, Alicia H., Mauricio Soto, Jerrilyn Libby and John Prinzivalli, "Investment returns: Defined Benefit vs. 401(k) Plans", Center for Retirement Research at Boston College, Brief Number 52, September 2006.

Office of the Chief Actuary of Canada, Actuarial Report (23rd) on the Canada Pension Plan as at 31 December 2006, October 2007.

Office of the Chief Actuary of Canada, Canada Pension Plan Mortality Study, Actuarial Study No. 7, July 2009.

Ostrovsky, Yuri and Grant Schellenberg, "Pension coverage, retirement status and replacement rates among a cohort of Canadian seniors", Statistics Canada, mimeo, October 5, 2009.

Quebec, Régie des Rentes, "Revenus de retraite au Québec : une projection sur 30 ans", presentation to the 9th Global Conference on Ageing, International Federation of Ageing, 7 September 2008.

Scholz, J. K., A. Seshadri and S. Khitatrakan, "Are Americans Saving "Optimally" for Retirement?", National Bureau of Economic Research, Working Paper 10260, January 2004.

Sinha, Rajeeva and Vijay Jog, "Fund Flows and Performance: A Study of Canadian Equity Funds", Canadian Investment Review, Summer 2007.

Statistics Canada, The Assets and Debts of Canadians: Focus on private pension savings, Cat. 13-596, December 2001.

1 In this paper, references to RPPs should be read to include DPSPs.

2 The decline in the RRSP contribution rate from 1997 to 2003 appears to have two main causes: (1) a shift to saving for children's education in registered educational savings plans (RESPs) following the enrichment of the tax advantages for such saving with the introduction of the Canada Education Savings Grant (CESG) in 1998, and (2) a shift in saving away from financial assets toward real estate following the bursting of the high technology bubble in 2000 (see Figure 2.4 below). Rising RPP contribution rates after 2001 likely reflect catch-up contributions made by sponsors of plans in deficit positions.

3 The net worth averages by age are taken from Table 7.2 of Statistics Canada Catalogue 13-596 (2001).

4 The age structure adjustment is made as follows. The 1999 age profile of net worth levels is multiplied by corresponding age group population weights for each year to obtain an aggregate net worth level for each year. An index is then created by expressing these net worth averages as ratios of the 1971 average. For example, the average for 2008 is 1.20 times the 1971 average. Finally, the 1971 ratio times the difference in the yearly index from 1.00 is subtracted from each year's base net worth-to-earnings ratio. For 2008, for example, the base net worth ratio is 6.01, up from 4.01 in 1971, and the corrected ratio is 5.21 [= 6.01 - 4.01 x 0.20].

5 The asset and debt survey was replaced by the Survey of Financial Security beginning in 1999.

6 As in the earlier period, median and average net worth declined among those under age 35. Stagnant real wage levels among young workers appear to be a leading cause of this.

7 The first objective, to guarantee a basic minimum income for all seniors, relates principally to public pension provision rather than retirement saving.

8 Couples must also expect family size to shrink on the death of a spouse. This unpleasant fact is ignored in the stylized model presented here.

9 In Section 6.4, however, the lifecycle model is used to examine which households would be better off saving with TFSAs and what the implications of the new savings vehicle might be for savings levels, retirement income adequacy and future public pension costs.

10 As summarized in Engen, Gale and Uccello (EGU, 1999), there has been considerable discussion in the literature concerning whether and to what extent housing wealth should be taken into account in measuring savings adequacy. In their analysis, EGU provided cases with all, half or none of housing wealth included. The method adopted here takes all housing wealth into account but spreads its consumption over the whole lifetime – as if it were all received in the form of imputed rent or, alternatively, as if a reverse mortgage on the full home value were bought at age 30. As an alternative, it could be assumed that only part of the value of the home is spread over the full lifetime (as imputed rent), and a residual portion is received as a reverse mortgage over the retirement period alone. This would lower benchmark savings rates somewhat. This effect would be diminished, however, if a bequest motive were assumed so that only part of the residual portion is consumed over the lifetime. The current approach is simpler than these alternatives.

11 As noted earlier, the part-career duration of child-related benefits and expenses is taken into account by spreading prorated amounts across the 35-year pre-retirement period.

12 The T1 file has in the past had an industry identifier that could be used to separate public and private sector workers. This information is no longer considered sufficiently reliable to include on the file.

13 A notable exception is the article by Morissette and Ostrovsky (2007) which examines how rising RPP coverage among women has mitigated the opposite trend among their spouses resulting in a more modest coverage decline when looked at on a family basis. It also looks at trends in the combined levels of RPP/RRSP saving by husbands and wives.

15 The means and medians for holders of other financial assets (before netting out debts) are: for Modest earners, $15,900 and$4,500, and for Middle earners, $39,000 and$13,000.

16 Of course, some who continue working past age 70 may do so for reasons other than financial need.

17 Further evidence of this comes from Statistics Canada's assessment of savings adequacy based on the 1999 SFS (Statistics Canada, 2001). They found, for example, that many self-employed people had significant pension assets derived from earlier periods of employment.

18 This is likely reflected in another finding of the Statistics Canada savings adequacy study cited above – that savings adequacy was surprisingly high among manufacturing sector employees, despite the relatively high incidence of pension coverage in the sector.

19 RPP coverage rates have tended to rise temporarily during recessions, however, because workers without pensions are the most likely to become unemployed.