ArticlePDF Available

How Accurately Does 70% Final Employment Earnings Replacement Measure Retirement Income (In)Adequacy? Introducing the Living Standards Replacement Rate (lsrr)

Authors:

Abstract and Figures

Will 70% of a worker's final annual employment earnings sustain living standards after retirement? Despite increasing skepticism, the most dominant measure of retirement income adequacy by financial planners, pensions plan ad-visors, academics and public policy makers is the " final employment earnings replacement rate " , where 70% is considered the right target to ensure living standards remain at approximately the same level after retirement. Using Statistics Canada's LifePaths dynamic population micro-simulation model, this paper asks whether those individuals from the 1951–1958 Canadian birth cohort who attain roughly a 70% final employment earnings replacement rate (as conventionally measured) at retirement do, in fact, achieve approximate continuity in their living standards. We find that the conventional final earnings replacement rate measure has little predictive value for living standards continuity between working-life and retirement. The primary reason is that employment earnings in a single year is not a reliable representation of a worker's standard of living — it relies on an inadequate pre-retirement measurement period, does not incorporate important components of consumption sources (such as home equity), and ignores household size (particularly children). As a result, we find that the correlation between the conventional earnings replacement rate and actual living standards continuity is relatively low (0.11). The paper therefore suggests an alternative metric for assessing how well a worker's living standard is maintained after retirement — i.e., the Living Standards Replacement Rate, or the LSRR. The LSRR provides a more accurate, understandable and consistent measure of retirement income adequacy. KEYWORDS Retirement income adequacy, 70% earnings replacement rate, micro-simulation, standard of living, Living Standards Replacement Rate, LSRR, retirement planning , employer pension plan design, public pension plan design
Content may be subject to copyright.
HOW ACCURATELY DOES 70% FINAL EMPLOYMENT EARNINGS
REPLACEMENT MEASURE RETIREMENT INCOME (IN)ADEQUACY?
INTRODUCING THE LIVING STANDARDS REPLACEMENT RATE (LSRR)
BY
BONNIE-JEANNE MACDONALD,LARS OSBERG AND KEVIN D. MOORE
ABSTRACT
Will 70% of a worker’s nal annual employment earnings sustain living stan-
dards after retirement? Despite increasing skepticism, the most dominant mea-
sure of retirement income adequacy by nancial planners, pensions plan ad-
visors, academics and public policy makers is the “nal employment earnings
replacement rate”, where 70% is considered the right target to ensure living
standards remain at approximately the same level after retirement. Using Statis-
tics Canada’s LifePaths dynamic population micro-simulation model, this paper
asks whether those individuals from the 1951–1958 Canadian birth cohort who
attain roughly a 70% nal employment earnings replacement rate (as conven-
tionally measured) at retirement do, in fact, achieve approximate continuity in
their living standards. We nd that the conventional nal earnings replacement
rate measure has little predictive value for living standards continuity between
working-life and retirement. The primary reason is that employment earnings
in a single year is not a reliable representation of a worker’s standard of living —
it relies on an inadequate pre-retirement measurement period, does not incor-
porate important components of consumption sources (such as home equity),
and ignores household size (particularly children). As a result, we nd that the
correlation between the conventional earnings replacement rate and actual liv-
ing standards continuity is relatively low (0.11). The paper therefore suggests an
alternative metric for assessing how well a worker’s living standard is maintained
after retirement — i.e., the Living Standards Replacement Rate, or the LSRR.
The LSRR provides a more accurate, understandable and consistent measure
of retirement income adequacy.
KEYWORDS
Retirement income adequacy, 70% earnings replacement rate, micro-simulation,
standard of living, Living Standards Replacement Rate, LSRR, retirement plan-
ning, employer pension plan design, public pension plan design
Astin Bulletin 46(3), 627-676. doi: 10.1017/asb.2016.20 C
2016 by Astin Bulletin. All rights reserved. This is an Open
Access article, distributed under the terms of the Creative Commons Attribution-NonCommercial-ShareAlike licence
(http://creativecommons.org/licenses/by-nc-sa/4.0/), which permits non-commercial re-use, distribution, and reproduction in any
medium, provided the same Creative Commons licence is included and the original work is properly cited. The written permission
of Cambridge University Press must be obtained for commercial re-use.
available at http:/www.cambridge.org/core/terms. http://dx.doi.org/10.1017/asb.2016.20
Downloaded from http:/www.cambridge.org/core. IP address: 142.167.14.111, on 01 Oct 2016 at 18:18:06, subject to the Cambridge Core terms of use,
628 B.-J. MACDONALD, L. OSBERG AND K.D. MOORE
1. INTRODUCTION
The capacity of retirement income to sustain the standard of living of future
seniors is a growing concern in Canada and elsewhere, owing to (1) low inter-
est rates, (2) high investment fees, (3) longer life expectancies, (4) rising divorce
rates among seniors (with likely negative nancial implications), (5) an aging
population, (6) an increasing reliance on paid services for the potentially costly
expenses associated with chronic health conditions and (7) less secure sources of
retirement income.1For example, the shift among employer pension plans from
dened benet (DB) to dened contribution (DC) in many countries around the
world largely shifts the market risk of benet payments to individual retirees.
A common response to these concerns is to encourage individuals to save
more. But saving too much for retirement has its disadvantages. Very high lev-
els of saving could be a rational strategy for highly risk-averse individuals who
place a high value on contingency income, for individuals anticipating higher
expenses after retirement (such as home repairs and covering medical expenses
not covered by the state health care system), and for individuals wishing to leave
bequests or improve their living standards in retirement. For others, however,
over-saving could mean unnecessary “scrimping and saving” during young and
healthy years when the welfare of children is possibly involved, and could pro-
duce excess wealth that is not necessarily enjoyed much, at the margin, during
more advanced ages. For instance, Smith et al. (2009) found that retirees con-
sume their personal savings at a very slow rate (the personal wealth of the upper
income quintile actually continued to accumulate after retirement).
The most widespread metric for evaluating the adequacy of retirement sav-
ings has historically been the “nal earnings replacement rate”, where 70% has
been advocated as the right target. Financial planners widely use the earnings
replacement rate benchmark, as do pension plan advisors, academics and public
policy makers. It also underlies much of the research that has predicted that a
large proportion of Canadian and American workers will be nancially unpre-
pared for an adequate retirement. In Canada, for example, workers have often
been told to aim for a 70% replacement rate of nal employment earnings in or-
der to maintain their working-life standard of living after retirement. The Bald-
win Report (2009: iv) stated, “The common approach to determining whether
the elderly are maintaining their standard of living is to compare the income
of the elderly population with their pre-retirement earnings. This comparison is
called the replacement rate and it is usually expressed as a percentage. Actual
replacement rates are compared with a benchmark replacement rate – usually
in the range of 70 to 75 per cent of gross pre-retirement earnings – to decide
whether people are maintaining their pre-retirement standard of living”. Liu et
al. (2013: 8) similarly wrote: “One approach taken in the literature is to assess the
adequacy of retirement income by focusing on the proportion of pre-retirement
income that has to be replaced during retirement in order to maintain living
standards at the pre-retirement level. A rule of thumb is that post-retirement
income should replace at least 70% of pre-retirement income”. The same
available at http:/www.cambridge.org/core/terms. http://dx.doi.org/10.1017/asb.2016.20
Downloaded from http:/www.cambridge.org/core. IP address: 142.167.14.111, on 01 Oct 2016 at 18:18:06, subject to the Cambridge Core terms of use,
THE LIVING STANDARDS REPLACEMENT RATE (LSRR) 629
target measure is conventionally advocated in other countries — for example,
the U.S. Social Security Administration (2008:7) wrote: “(w)hile Social Security
replaces about 40 percent of the average worker’s pre-retirement earnings, most
nancial advisors say that you will need 70 percent or more of pre-retirement
earnings to live comfortably”. Moreover, the assumption that 70% nal earnings
replacement after 35 years of service is the appropriate retirement income target
underlies the design of most employer DB pension plans. At the individual level,
it is currently the “staple of web-based nancial planning products” (Scholz and
Seshadri, 2009).
Studies skeptical about whether target earnings replacement rates provide
an adequate benchmark for assessing the adequacy of an individual’s retire-
ment income include Vanderhei (2006) and Scholz and Seshadri (2009). This
paper continues that tradition by testing the earnings replacement rate measure
empirically — asking how well living standards are maintained by simulated in-
dividuals who actually achieve the often-advocated 70% target. The paper then
goes on to suggest a more accurate metric of retirement income adequacy.
We examine Canadians who attain approximately a 70% nal employment
earnings replacement target at retirement, and ask how well each such persons
working-life living standard is maintained after retirement. More specically,
for individuals projected to retire at age 61 (the median retirement age in
Canada2) between years 2012 and 2019, we calculate the nal employment
earnings replacement rate for each person and select those for whom the nal
earnings replacement rate lies between 65% and 75%. For each such individual,
we then compare average post-retirement living standards to average pre-
retirement living standards, to evaluate whether there is, in fact, approximate
continuity in his/her living standards.
To estimate living standards over the life course, we utilize Statistics
Canada’s LifePaths dynamic micro-simulation model of the Canadian popula-
tion,3which provides the level of detail necessary for evaluation of living stan-
dards before and after retirement. LifePaths is a long-standing model whose de-
velopment began in the early 1990s. Although models like LifePaths can never
be perfect substitutes for actual data, large-scale, complex, dynamic micro-
simulation models are increasingly becoming the international tool of choice
for assessing retirement income adequacy, as well as a variety of other analyses.
In addition to testing the accuracy of the conventional earnings replacement
rate measure and the 70% target, this paper presents a framework for a statistic
that more accurately assesses how well a worker’s living standards are main-
tained after retirement using income and wealth data — the Living Standards
Replacement Rate (LSRR). The goal of the LSRR is to capture a worker’s living
standards continuity after retirement, by comparing how much money a worker
has available to support personal consumption of goods and services before and
after retirement. Its purpose is to serve as a retirement income adequacy statistic
that is more accurate, understandable and consistent in its application to anal-
ysis compared to the conventional earnings replacement rate measure. We hope
that the LSRR can provide a better metric for public policy analysts, pension
available at http:/www.cambridge.org/core/terms. http://dx.doi.org/10.1017/asb.2016.20
Downloaded from http:/www.cambridge.org/core. IP address: 142.167.14.111, on 01 Oct 2016 at 18:18:06, subject to the Cambridge Core terms of use,
630 B.-J. MACDONALD, L. OSBERG AND K.D. MOORE
plan advisors, nancial analysts and academics when measuring retirement in-
come adequacy.
Section 2 of this paper reviews the conventional nal employment earnings
replacement rate measure; Section 3 presents the LSRR and outlines our meth-
ods; Section 4 analyzes our results; Section 5 discusses future work in terms of
practical considerations when adopting the LSRR; and Section 6 concludes.
2. REPLACEMENT RATES AND REPLACEMENT RATE TARGETS —REVIEW
AND DISCUSSION
The common criterion for retirement income adequacy is whether working-life
living standards are sustained after retirement and the conventional statistic for
evaluating an individual’s likely living standards in retirement relative to work-
ing life is the nal employment earnings “replacement rate” — i.e., the fraction
of a worker’s annual nal employment earnings replaced by annual retirement
income. Across academia, industry, and government public policy research, the
earnings replacement rate is the most prevalent statistic for measuring retire-
ment income adequacy, and many analyses rely on it exclusively.4
The apparent simplicity of the earnings replacement rate approach has been
a primary reason for its popularity. There exist, however, major inconsistencies
in the analysis of earnings replacement rates owing to differences in the concep-
tual framing of retirement income adequacy5as well as analytical purpose,6but
most of all owing to data constraints.
From the perspectives of the policy analyst, academic, nancial planner, em-
ployer and the individuals themselves, the most easily obtainable data is current-
year before-tax earnings data. For example, current year earnings are commonly
collected in public surveys; a client who visits his/her nancial planner typically
brings his/her most recent income tax return; and employers who sponsor a pen-
sion plan have the current year payroll for their employees. Having this limited
data readily at hand, retirement income adequacy has historically been mea-
sured using the “nal employment earnings replacement rate”, which is con-
ventionally calculated as
Conventional nal employment earnings replacement rate
=gross (i.e., before-tax)income in rst year of retirement
gross pre-retirement nal year employment earnings .(1)
The conventional nal employment earnings replacement rate is sometimes
also referred to as the “gross” employment earnings replacement rate since it is
calculated using pre-tax employment earnings and pre-tax retirement income.
The “retirement income” in the conventional earnings replacement rate formula
consists of income reported for income tax purposes (and therefore routinely
does not include savings whose withdrawals are not taxed). (Note that Equa-
tion (1) is intended to represent the most conventional earnings replacement
available at http:/www.cambridge.org/core/terms. http://dx.doi.org/10.1017/asb.2016.20
Downloaded from http:/www.cambridge.org/core. IP address: 142.167.14.111, on 01 Oct 2016 at 18:18:06, subject to the Cambridge Core terms of use,
THE LIVING STANDARDS REPLACEMENT RATE (LSRR) 631
rate formula, although the authors were unable to locate an original source that
prescribes this measure.)
The “rule-of-thumb” belief is that 70% nal employment earnings replace-
ment will sustain a worker’s standard of living after retirement — i.e., provide
100% replacement of pre-retirement living standards. This expectation is based
on the idea that retired individuals will generally pay lower taxes, not be sav-
ing for retirement, typically have paid off their mortgage and no longer need
to support children and/or pay work-related expenses. It also stems from the
perspective that, since retirement is the end of employment, only employment
earnings should be replaced.
Unfortunately, many components of living standards are either poorly dealt
with, or omitted from, Equation (1):
1. household-level differences in individual consumption due to family size,
particularly dependent children (and changes over time in household size
and composition);
2. the return on house ownership, or “imputed rent”;
3. taxes (specically the differentials in taxation year-by-year, pre- and post-
retirement);
4. government transfers (e.g., unemployment insurance, child benets and
social assistance);
5. the accumulation and drawdown of non-traditional forms of savings (e.g.,
non-tax-sheltered nancial wealth, debt, and home-ownership equity);
6. earnings volatility (discussed more below);
7. retirement income volatility;
8. pre- and post-retirement risks (e.g., poor nancial market returns,
death/divorce of a spouse, longevity, expensive medical conditions, ex-
tended care needs and ination uncertainty);
9. phased retirement and continuing employment income after retirement;
10. individual preferences (e.g., risk aversion, the value of leisure and bequest
motives);
11. changes in expenses over the life course.
By way of example, consider #6 (the omission of earnings volatility). When
calculating the denominator of Equation (1), the choice between using earnings
in the nal year of employment (such as age 64) versus a longer measurement
period (such as averaged from age 60–64) can produce substantial discrepan-
cies across the results, due to short- and long-term earnings volatility across
workers’ careers (see Finnie, 1999; Morissette et al., 2007; Finnie and Gray,
2011). Biggs and Springstead (2008), Munnell and Soto (2005) and Boskin and
Shoven (1987) are examples of studies that observed an enormous impact of the
measurement period choice on the resulting replacement rate, leading to quite
different conclusions about replacement adequacy.
Owing to these omissions, some literature has questioned the validity of tar-
get earnings replacement rates as a benchmark for assessing retirement income
adequacy. Engen et al. (1999) pointed out that the existence of risk necessarily
available at http:/www.cambridge.org/core/terms. http://dx.doi.org/10.1017/asb.2016.20
Downloaded from http:/www.cambridge.org/core. IP address: 142.167.14.111, on 01 Oct 2016 at 18:18:06, subject to the Cambridge Core terms of use,
632 B.-J. MACDONALD, L. OSBERG AND K.D. MOORE
creates a distribution of target earnings replacement rates whose mean or me-
dian can only be interpreted as such, and not as a minimum single target. In
more recent literature, using longitudinal data from the Health and Retirement
Study and a life-cycle consumption and savings model, Scholz and Seshadri
(2009) found that optimal gross earnings replacement rates covered a very wide
range — concluding, “What is clear from this discussion is that the substan-
tial variation in optimal target replacement rates presents a challenge for de-
veloping sensible replacement rate rules of thumb. Conventional advice may
overstate optimal targets by a factor of two, or understate retirement consump-
tion needs by a factor of three depending on the idiosyncratic experiences of
households” (21). VanDerhei (2006) also found that “a simple one-size-ts-all
replacement rate will not work for most Americans” (5). Simulating 1,000 life-
paths of stylized 65-year-old retirees and testing the adequacy of earnings re-
placement rate targets, he concluded that “the huge variation in the range of
replacement rate targets—depending on the individual’s income, degree of an-
nuitization for initial retirement wealth, and the asset allocation of the post-
retirement investments—call into question whether the use of a single rule-of-
thumb measure is realistic to use in the retirement planning process” (5).
Owing to the generally understood gap between the objective of the earnings
replacement rate benchmark (that is, to measure the continuity of a worker’s
living standards after retirement) and its limited formulation (being simply the
ratio of gross retirement income divided by gross nal employment earnings),
analysts seeking to improve on its methodological weaknesses have applied it in
a wide variety of ways. The renements made by analysts are often driven by
data availability. Consequently, the replacement rate calculation has been very
inconsistent across analyses, which has led to varying and often conicting re-
sults regarding retirement income adequacy (as we discuss further in Sections 4
and 5). Appendix C discusses the methodological weaknesses listed above, and
provides examples of the varied approaches taken in literature.
3. CONCEPTUAL AND METHODOLOGICAL FRAMEWORK FOR RETIREMENT
INCOME ADEQUACY
3.1. Why Microsimulation?
As discussed in Section 2, analysts often tweak the earnings replacement rate
measure by incorporating some improvements to the conventional formula, and
the availability of data often shapes these improvements. This is particularly true
in analyses at the population-level. For example, available longitudinal popula-
tion data rarely reports on all components of income, savings, dissavings and
wealth, making it difcult for analysts to investigate the adequacy of a coun-
try’s pension system for future retirees. Although more comprehensive data can
sometimes be found in cross-sectional surveys reporting on a particular year,
having only a single year of data often constrains analysts to projections using
available at http:/www.cambridge.org/core/terms. http://dx.doi.org/10.1017/asb.2016.20
Downloaded from http:/www.cambridge.org/core. IP address: 142.167.14.111, on 01 Oct 2016 at 18:18:06, subject to the Cambridge Core terms of use,
THE LIVING STANDARDS REPLACEMENT RATE (LSRR) 633
highly stylized assumptions that do not capture the realistic variability across
and among the life courses of individuals within a population. Examples in-
clude Palmer (1988), Mitchell and Moore (1998), Munnell et al. (2006) and
Palmer (2008). In comparison, replacement rate studies that employed longitu-
dinal data include Boskin and Shoven (1987), Smith (2003), Munnell and Soto
(2005), LaRochelle-Cote et al. (2008), Ostrovsky and Schellenberg (2009) and
Denton et al. (2011).
As a result, large-scale, complex, dynamic micro-simulation models are in-
creasingly being used to assess retirement income adequacy, because they have
the potential to
integrate and extend existing data sources to give the most comprehensive
picture of consumption sources before and after retirement;
model individuals’ interactions with and accruals under retirement income
programs throughout their entire lifetime, simultaneously enabling full exi-
bility in analysis;
generate results that reect the realistic complexity and diversity within life-
courses, and across individuals;
model the likely impact of current socio-economic trends on future outcomes,
and facilitate explicit evaluations of the uncertainty of the future (including
the post-retirement risks that people face).
For example, micro-simulation enables nearly full exibility to choose alterna-
tive measurement periods before and after retirement rather than depend exclu-
sively on employment earnings in the year before retirement (as noted earlier, the
measurement period choice plays a critical role in the results of replacement rate
analyses). For example, using the U.S. Social Security Administration’s Model-
ing Income in the Near Term (MINT) micro-simulation model, Butrica et al.
(2003) measured and averaged pre-retirement income from ages 22 to 62. Also
using the MINT micro-simulation model, Biggs and Springstead (2008) com-
pared the different results that follow from using various measurement periods
for earnings, including a variety of career-average and nal average earnings
measures.
As Klevmarken explained nearly two decades ago “The micro-simulation
approach is thus primarily designed for studies of the distributional effects of
economic policy, and one of its main advantages is that it permits assumptions
of heterogeneous behavior” (Klevmarken 1997: 2). In other words, it allows
people to act differently, rather than assume that everyone behaves like the “av-
erage”. “This, as a matter of fact, widens the scope of micro-simulation beyond
that of conventional econometric modeling. When economic relations are highly
nonlinear, when tax laws and rules of transfer programs introduce censoring and
truncation and when sub-populations differ in behavior, then models of average
behavior become inadequate to evaluate the average impact of policy changes,
while a micro-simulation model can be used also for this purpose” (ibid). Ow-
ing to their advantages, dynamic micro-simulation models are ever more the
tool of choice by policy makers throughout the industrialized world for analysis
available at http:/www.cambridge.org/core/terms. http://dx.doi.org/10.1017/asb.2016.20
Downloaded from http:/www.cambridge.org/core. IP address: 142.167.14.111, on 01 Oct 2016 at 18:18:06, subject to the Cambridge Core terms of use,
634 B.-J. MACDONALD, L. OSBERG AND K.D. MOORE
relating to the ageing of the population and future retirement income system
outcomes — in addition to Statistics Canada’s LifePaths, examples include
APPSIM in Australia (Harding, 2007), CBOLT in the United States (O’Harra
et al., 2004) and MIDAS in Belgium (Dekkers and Belloni, 2009). Interested
readers are directed to Li and O’Donoghue (2013) for a recent survey of dynamic
micro-simulation models internationally, including their uses, model structure
and methodology. ESPlanner (Bernheim et al., 2000; Kotlikoff, 2006) in the
United States and Ruthen (Avery and Morrison, 2009) in Canada are examples
of personal dynamic micro-simulation models for individual nancial planning
purposes.
3.2. Tool of analysis: LifePaths
To benet from the advantages of dynamic micro-simulation modeling identi-
ed in Section 3.1, the present paper evaluates living standards continuity by
building on Statistics Canada’s LifePaths micro-simulation model (Spielauer,
2013).
LifePaths is a dynamic micro-simulation model of the Canadian population
that simulates individual life-courses (birth, education, employment, income,
taxes, marriage, child-bearing, retirement, etc) of synthetic individuals that
are representative of the Canadian population. Using behavioral equations,
it simulates each “life-path” year-by-year and case-by-case, while striving
to incorporate the diversity between individuals and over each persons life-
course. LifePaths summarizes, incorporates and integrates an enormous range
of Canadian data in order to simulate statistically representative data samples
of the history of the Canadian population. Examples of studies that employed
LifePaths in the context of projecting the retirement income adequacy of
Canadians include Moore et al. (2010), MacDonald et al. (2011) and Wolfson
(2011). Appendix D provides a further description of LifePaths.
Using LifePaths’ simulations, we compute estimates for individual living
standards across the life-course to evaluate how well living standards are
maintained by people who approximately hit a 70% earnings replacement
rate target. We use a comprehensive denition of income that includes non-
traditional working and retirement income sources, and we go beyond the
single year before and after retirement to look over the individual’s entire
lifetime (year-by-year at the family level, making appropriate adjustments for
family size). We stochastically model nancial market returns and mortality
(including the death of family members). Individuals exit the workforce in a
realistic manner that matches Canadian labor force data.
3.3. Living Standards Replacement Rate (LSRR)
This section suggests a metric that more comprehensively and accurately de-
termines how well a person’s working-life living standards are replaced in
available at http:/www.cambridge.org/core/terms. http://dx.doi.org/10.1017/asb.2016.20
Downloaded from http:/www.cambridge.org/core. IP address: 142.167.14.111, on 01 Oct 2016 at 18:18:06, subject to the Cambridge Core terms of use,
THE LIVING STANDARDS REPLACEMENT RATE (LSRR) 635
retirement — the LSRR. The goal of the LSRR is to capture a worker’s living
standards continuity after retirement, by calculating how much money a worker
has available to support his/her personal consumption of goods and services
before and after retirement.
(After presenting the LSRR in this section, we next use the LSRR in Section
4 to measure the continuity of living standards for workers who attain a 65%–
75% conventional earnings replacement rate. Section 5 discusses the value of
adopting the LSRR as a retirement income adequacy statistic, as well as some
practical considerations when doing so.)
The LSRR measures living standards by calculating the ow of income avail-
able to support the individual’s standard of living pre- and post-retirement (in
other words, the income available to support the consumption of goods and
services). During the working-years, the LSRR assumes that the income avail-
able for individual consumption equals his/her family’s disposable income (gross
income after taxes and transfers) less net savings, and then adjusted for family
size. During retirement, it assumes that the income available for individual con-
sumption would equal disposable income plus the drawdown from accumulated
savings, which would also be calculated at the family level and adjusted for fam-
ily size. In short, the LSRR captures living standards by determining the amount
of money that the individual has available to spend before and after retirement to
support his/her personal level of consumption.
Figure 1 presents the LSRR framework for estimating an individual’s living
standards each year while working, and his/her potential living standards in
retirement.
Once an individual’s living standards are estimated for each year of life, then
the LSRR is simply the average estimated retirement living standards divided
by the average estimated working-life living standards. More precisely, with an
accounting period of one scal year, the LSRR is calculated as
Living Standards Replacement Rate (LSRR)
=average annual retirement living standards
average annual working-life living standards
=average real annual retirement income for potential individual consumption expenditure
trimmed average real annual working income for individual consumption expenditure .
(2)
“Real” or “constant dollar” income is calculated using the All-Items Con-
sumer Price Index for Canada. We use a trimmed average of income in the work-
ing life: Using the 30 years leading up to retirement, we remove the incomes in
the lowest and highest ve years, and average the real income of the remain-
ing middle 20 years. We chose to trim our average, as trimming reduces the
inuence of outlier years — e.g., both abnormally low (possibly negative) and
unusually high employment earning years — which is particularly important
available at http:/www.cambridge.org/core/terms. http://dx.doi.org/10.1017/asb.2016.20
Downloaded from http:/www.cambridge.org/core. IP address: 142.167.14.111, on 01 Oct 2016 at 18:18:06, subject to the Cambridge Core terms of use,
636 B.-J. MACDONALD, L. OSBERG AND K.D. MOORE
+
+
Family Member Working? Family Member Retired?
- +
-
= Repeat for each family member
LIS equivalence scale = (Number of Family Members)0.5
Subject Individual Working? Subject Individual Retired?
=
Market Income =
Employment income (employee income and net-income from self-employment)
+Investment Income (dividends, interest and other investment income such as real estate)*
+Imputed return on owner occupied housing (= imputed rent repairs/taxes/mortgage interest)
Taxes, Deductions, Non-refundable Credits, Transfers and Work Expenses Paid =
Taxes on income and wealth (other than owner-occupied home)
+
+Cash transfers paid to other households and institutions
+Work expenses (travel and child care payments)**
**
Transfers Received =
Benefits from employers (e.g. DB pension plan benefits)
+Social insurance benefits (e.g. EI, OAS, GIS, C/QPP benefits, death and survivor benefits)
+Social-assistance benefits (e.g. social assistance payments, child benefits, provincial
supplements and tax credits)
+Cash transfers received from other households and institutions (e.g. alimony, child support,
support from other persons not in the household, inheritance**, etc.)
Net Savings to:
Employer pension plan (DB and
DC pension plan contributions )
RRSP savings (contributions and
withdrawals)
Mortgage principal reduction***
Other savings****
Notional potential dissavings from net worth
(assets minus liabilities): individual payment from
inflation-indexed life annuity purchased at
retirement with
RRSP/RRIF wealth (including DC pension
plan wealth)
Housing equity**
Other savings****
f
und
f
f
a
b
C
r
e
d
it
s
,
Cumulative Income for Family Consumption Expenditure
Working Income for Individual
Consumption Expenditure (proxy for fiscal
Working Life Living Standards)
Retirement Income for Potential
Consumption Expenditure (proxy
Retirement Living Standards)
Repeat
f
or eac
h
f
am
il
y mem
b
er
0
5
S
u
g?
f
o
f
f
r
I
n
d
Annual
for fiscal
Individual
Annual
FIGURE 1: LSRR’s conceptual framework to measure an individual’s scal year’s “living standards” (using
census family-level data of subject individual).
Notes: * If any portion of investment income is saved rather than consumed, that portion would then be added
to savings and hence subtracted (having zero net effect, aside from any possible tax implications). ** The
italicized items are those not included in this paper. We do not model work expenses and we assume that
seniors do not draw down their housing equity to support consumption. We do not explicitly model the
receipt of inheritance, although past inheritances are recorded in wealth and therefore are implicitly treated as
past savings. LifePaths does not model cash transfers paid to other households, although it does loosely model
cash transfers received. *** Retiring homeowners with a mortgage may also continue to make principal
payments into retirement depending of the analyst’s treatment of housing equity. For instance, this paper does
not assume that the retiree downsizes at retirement by selling his/her home, and therefore he/she continues to
benet from imputed rent and may make mortgage payments. Note that the interest portion of mortgage
payments is incorporated as a component of the imputed return on owner occupied housing. **** The “other
savings” concept used is marketable wealth (or net worth) other than primary housing. These include the sum
of non-registered nancial assets (chequing accounts, GICs, trusts, etc), real estate assets (other than primary
housing), and business equity, less non-mortgage debt (credit card, lines of credit, car loans, etc).
Acronyms in Figure 1: The Canadian public pension system consists of the universal Old Age Security
program (OAS), the income-tested Guaranteed Income Supplement program (GIS) and the earnings-based
contributory state plan — the Canadian/Quebec Pension Plans (C/QPP). Registered Retirement Savings Plans
(RRSPs) and Registered Retirement Income Funds (RRIFs) are tax-sheltered retirement savings.
Employment Insurance (EI) provides temporary nancial aid to Canadians who have lost their jobs.
available at http:/www.cambridge.org/core/terms. http://dx.doi.org/10.1017/asb.2016.20
Downloaded from http:/www.cambridge.org/core. IP address: 142.167.14.111, on 01 Oct 2016 at 18:18:06, subject to the Cambridge Core terms of use,
THE LIVING STANDARDS REPLACEMENT RATE (LSRR) 637
among the self-employed. Note that the results were not largely impacted had
we used a conventional average (for 92% of the sample, the trimmed average lies
between 90% and 110% of the full 30-year average). Regardless of the method of
averaging, it is crucial to use a representative number of years in the calcu-
lation (so as to smooth out year-to-year volatility). (Section 5 discusses this
further).
Our estimate for annual retirement living standards (the annual “Retirement
Income for Potential Individual Consumption Expenditure” in Figure 1) is av-
eraged from retirement until death. We do not trim the average since the stability
of retirement income sources is particularly high relative to employment earn-
ings volatility (see Finnie, 1999; Morissette et al., 2007; Finnie and Gray, 2011),
especially given that we assume the full annuitization of savings (as we discuss
below).
Since our focus is retirement income adequacy, we infer what individuals
could potentially consume in retirement. As is customary in this literature, we
convert the stock of wealth into a ow of potential annual consumption by as-
suming that wealth (net worth =assets liabilities) is annuitized at retirement.
This is a simplifying assumption, although not a realistic one — we discuss the
issues surrounding this assumption at the end of this section. The payments of
this notional annuity are added to other retirement income ows. The “potential
dissavings from net worth” portion in Figure 1 is, therefore, the annual poten-
tial payout from an ination-indexed life annuity purchased at retirement with:
registered and non-registered wealth,7real estate investments, and business eq-
uity, minus debt. To compute individual equivalent income, we measure income
ows at the census-family level, correct for ination, and adjust for family size
using the Luxembourg Income Study (LIS) equivalence scale (the square root
of household size8).
For our analysis in Section 4, we use 80% <LSRR <120% as the range
of outcomes compatible with living standards continuity (i.e., working life “liv-
ing standards” plus or minus 20%). Although these ranges are subjective, some
support can be drawn from Binswanger and Schunk (2012: 217), which found
that, “a large majority of individuals aims to achieve a spending prole where,
under normal circumstances, old-age spending exceeds 80 percent of working-
life spending” using individually tailored internet surveys in the United States
and the Netherlands.9
The differences between the conventional earnings replacement rate from
Equation (1) and the LSRR from Equation (2) are as follows: (1) The LSRR
uses a much broader measurement period for pre- and post-retirement, (2)
the LSRR measures income at the family level rather than at the level of the
individual and (3) the LSRR includes a much more comprehensive denition
of income. To illustrate, we next consider these differences in turn between
Equation (1) and Equation (2) (assuming a retirement age of 61). Note that
these formulas use an accounting period of a full scal year (not included are
the partial year of death as well as the rst year of retirement, when the subject
is working a part of the year and retired the other part).
available at http:/www.cambridge.org/core/terms. http://dx.doi.org/10.1017/asb.2016.20
Downloaded from http:/www.cambridge.org/core. IP address: 142.167.14.111, on 01 Oct 2016 at 18:18:06, subject to the Cambridge Core terms of use,
638 B.-J. MACDONALD, L. OSBERG AND K.D. MOORE
Integrated data that covers all aspects of Figure 1 are not available in
Canada. This study is able to include most components (Figure 1 italicizes the
missing elements) by building on Statistics Canada’s LifePaths model.
In constructing the LSRR for the purpose of this study, we do not model
differences in individual preferences or changes in expenses over the life course.
These considerations generally fall outside the basic earnings replacement rate
framework when examining populations, although they can be critical to per-
sonal retirement nancial planning. They can be incorporated in the LSRR
by including the changing expenses in its numerator and/or denominator —
this is illustrated with “work expenses” in Figure 1, for example, although we
do not model this changing expense in our analysis (see Appendices C.6 and
C.7 for further discussion on these considerations). Additional components
of consumption not incorporated in Figure 1 that could be relevant to retire-
ment income adequacy include (1) income in-kind — that is, goods produced
by households for their own consumption, such as housekeeping, child-rearing
and cooking (for a discussion on the value of home food-production for seniors,
see Aguiar and Hurst, 2005; Brzozowski and Lu, 2010); (2) the ow of services
from durable goods (other than just owner-occupied housing) purchased while
working but consumed after retirement; (3) irregular inter-household transfers
in kind (gifts) and (4) social transfers in-kind — i.e., goods provided by the
public sector such as hospital and medical care services, police, parks and roads.
There are a number of difculties when attempting to include any of these four
additional considerations, and these difculties are sufciently widespread that
available at http:/www.cambridge.org/core/terms. http://dx.doi.org/10.1017/asb.2016.20
Downloaded from http:/www.cambridge.org/core. IP address: 142.167.14.111, on 01 Oct 2016 at 18:18:06, subject to the Cambridge Core terms of use,
THE LIVING STANDARDS REPLACEMENT RATE (LSRR) 639
these components are conventionally not incorporated, although analysts may
wish to include them depending on the purpose of their investigation and the
availability of data. The Canberra Group (2001), an international expert group
on household income statistics, concluded that “it is difcult to nd a metric by
which it would be possible to say by how much greater expenditure on defense
or on road-building increased the well-being of the inhabitants. Because of this
difculty, it is not usual to include the level of government collective services
in income comparisons” (22) and “(t)he treatment of irregular inter-household
transfers in kind – gifts – is another issue raised but not resolved” (110), in its
report that prescribes best practices for the production of household income
statistics (primarily for the measurement of economic well-being).
Figure 1 includes some items that our study precludes on conceptual
grounds, although analysts building the LSRR may wish to include them. This
includes the assumption that seniors will not draw down housing equity to sup-
port consumption. Canadian seniors are much less likely to move from (and sell)
their home than any other age group (Clark, 2005), and the take-up of nancial
instruments that draw on housing equity (such as reverse mortgages) is rare in
Canada (Chiuri and Jappelli, 2010). Similarly in the United States, American
seniors typically do not spend down the housing asset. For the most part, the
elderly are unlikely to move or draw on housing equity to support consumption
after retirement, except after major life transitions (such as the death of a spouse
or illness), despite the fact that housing equity is most often a retirees’ largest
asset (other than Social Security and sometimes employer pension plan) (Venti
and Wise, 2001, 2004). This suggests that housing equity is not generally viewed
as a source of consumption, but as contingency savings to be used in the event
of large and unforeseen expenses.
We also do not include “post-retirement” employment earnings since the
very purpose of a “replacement rate” concept is to evaluate the extent that em-
ployment living standards have been replaced after retirement (where retirement
is the cessation of employment) (see Appendix C for further discussion). Clearly,
on an individual nancial planning level, an analyst building a personalized
LSRR would want to include post-retirement employment earnings to account
for any work that the individual subject plans to do in retirement, health per-
mitting, as well as the drawdown of home equity if the individual plans to sell
his/her home. As already noted, moreover, the LSRR can be further tailored
to t an individual’s personal circumstances/preferences by including other an-
ticipated income sources (i.e., large gifts) and changes in expenses, as well as
aiming for a higher/lower standard of living after retirement by setting the tar-
get at more/less than 100%. The analyst can also include intended bequests (by
removing the bequest amount from net worth) as well as continued savings af-
ter retirement. Section 5 discusses these practical considerations further, and
Appendix C reviews these various issues when building replacement rates.
We convert wealth stocks into an income-ow at retirement by assum-
ing a life-only ination-indexed annuity for all individuals (whether single or
non-single).10 The replacement rate calculation is simplied by assuming that
available at http:/www.cambridge.org/core/terms. http://dx.doi.org/10.1017/asb.2016.20
Downloaded from http:/www.cambridge.org/core. IP address: 142.167.14.111, on 01 Oct 2016 at 18:18:06, subject to the Cambridge Core terms of use,
640 B.-J. MACDONALD, L. OSBERG AND K.D. MOORE
savings are annuitized at retirement, and this simplifying assumption is the norm
in replacement rate literature (for example, see Gustman and Steinmeier, 1998;
Munnell et al., 2006, 2007a, 2007b). If retirement income ows are taken only
in the rst year of retirement (see Equation (1)), for example, assuming that
personal savings are annuitized enables a straightforward comparison between
employees with DB pension plans and those with DC accounts. Using ination-
indexed annuities keeps the income ows from wealth on par with those re-
tirement income ows that are also ination-adjusted (C/QPP, OAS, GIS and
ination-adjusted employer DB pension benets).
In actual fact, however, voluntary annuitization is extremely rare — see
Milevsky and Young (2007) and Brown (2009) for U.S. evidence, and James
et al. (2001) for international — and ination-indexed annutization is nearly
non-existent (indeed the authors were unable to locate a Canadian insurer who
offered this product). Since people generally do not voluntarily annuitize, the
mortality premium underlying annuity pricing will, on the whole, overstate the
observed ow of gross income from wealth. The assumption of annuitization
understates the post-retirement nancial risks that retirees actually face — such
as ination rate uncertainty, nancial market risks and longevity risk, thereby
attenuating our results. As noted in Vanderhei (2004) and Scholz and Seshadri
(2009), the shift from DB to DC pension plans among employers has compli-
cated the replacement rate measurement since calculating an “income” from a
DC account requires behavioral drawdown assumptions (while the income from
a DB pension benet is prescribed, and does not carry the risks associated with
the DC income streams). The increasing proportion of DC pension plans leaves
retirees more exposed to nancial market volatility, both in capital value and
returns to assets owned, leading to a large range of possible future outcomes.
A challenge for future researchers is to decide how to handle these various risks,
either within or outside the replacement rate framework (Appendix C.4 dis-
cusses the pre- and post-retirement risks not well captured in the replacement
rate framework, and possible approaches for incorporating them in the study of
replacement rate adequacy). Given these issues, an analyst building the LSRR
may choose to assume another drawdown strategy, depending on the purpose
of the measure and prior knowledge on the individual subject’s drawdown pref-
erences.
3.4. Sample
The conventional earnings replacement rate framework norm assumes the com-
plete cessation of employment at a single age of retirement. Since our purpose is
to understand whether the traditional 70% target does in fact maintain working-
life living standards after retirement without continued employment, this anal-
ysis does not incorporate the growing trend to partial retirement and part-time
work among older workers — for example, Quinn (1999) estimates that 33%–
50% of older Americans use bridge jobs between a full-time career and complete
retirement. (Note that, unlike the conventional earnings replacement rate, the
available at http:/www.cambridge.org/core/terms. http://dx.doi.org/10.1017/asb.2016.20
Downloaded from http:/www.cambridge.org/core. IP address: 142.167.14.111, on 01 Oct 2016 at 18:18:06, subject to the Cambridge Core terms of use,
THE LIVING STANDARDS REPLACEMENT RATE (LSRR) 641
LSRR accomodates partial retirement and part-time employment of seniors in
its calculation, as Section 3.3 discussed.). We therefore include in our sample
only those working individuals who cease employment after retirement, which
we assume is employment of 25% or less of any scal year. We dene retirement
as the transition from working at least 75% to less than 25% of the year.11 Den-
ing retirement by reduction of hours worked is a commonly used approach in
literature (see Denton and Spencer (2009) for an international review on retire-
ment concepts and measures).
With this denition of retirement, we examine the continuity of living stan-
dards using the LSRR (given in Equation (2) and explained in Figure 1) for
individuals with a 65%–75% earnings replacement rate (given in Equation (1)
and repeated here):
replacement rate =gross (i.e., before-tax)retirement income in the rst scal year of retirement
gross pre-retirement nal employment earnings in the last scal year before retirement,
where employment earnings are made up of wages and self-employment gross
income, and gross retirement income consists of
the retirement benets from
the Canadian federal pension system (Canada/Quebec Pension Plan,
Guaranteed Income Supplement, and Old Age Security);
any occupational DB pension plan(s);
notional annuity income from registered savings — single-life ination-
indexed annuity, purchased at retirement with any registered personal savings
(Registered Retirement Savings Plans) and occupational DC pension plan
wealth.
Since retirement generally occurs partway through a nancial year (the most
popular month being June in Canada (Schwartz, 2010)), incomes are calculated
for the full year preceding and following the age of retirement.
Our sample consists of simulated individuals born between 1951 and 1958,
who reside in Canada from at least age 35 onward, and who retire at age 61 (the
median retirement age in Canada between 2001 and 2009 has ranged from 60.6
to 61.9 years (Schwartz, 2010)).
The rst column of Table 1 presents the broad characteristics — gender, mar-
ital status, work sector and education — of our sample (1951—1958 birth cohort
retiring at age 61 with a 65%–75% conventional earnings replacement rate as
given in Equation (1)). Among the simulated 1951-1958 birth cohort who are
“working” at age 60 (i.e. employed more than 75% of the weeks within the scal
year), just over 4% retire at age 61 (that is, did not work for more than 25% of
any subsequent scal year), and 5.4% of this group have a conventional earn-
ings replacement rate between 65% and 75%. Reducing the broader population
to the sample that we want to address resulted in a much smaller sample size
(3,122 workers). Our conclusions cannot speak, therefore, meaningfully to the
available at http:/www.cambridge.org/core/terms. http://dx.doi.org/10.1017/asb.2016.20
Downloaded from http:/www.cambridge.org/core. IP address: 142.167.14.111, on 01 Oct 2016 at 18:18:06, subject to the Cambridge Core terms of use,
642 B.-J. MACDONALD, L. OSBERG AND K.D. MOORE
TABLE 1
CHARACTERISTICS OF (1) 1951–1958 BIRTH COHORT OF SIMULATED INDIVIDUALS WHO RETIRE AT AGE 61
WITH 65%–75% CONVENTIONAL EARNINGS REPLACEMENT RATE; (2) 1951–1958 BIRTH COHORT OF
SIMULATED INDIVIDUALS WHO RETIRE AT AGE 61 AND (3) ENTIRE 1951–1958 BIRTH COHORT OF SIMULATED
INDIVIDUALS.
1951–1958 Birth Cohort
Those Who Retire
at Age 61 with 65%–75% Those Who Entire
Conventional Retire at Birth
Replacement Rate Age 61 Cohort
Household
Single Male 11% 12% 11%
Single Female 19% 15% 14%
Member of Couple 69% 73% 74%
Sector
Public 24% 18% 17%
Private 76% 82% 82%
Education
Less than High School 9% 17% 23%
High School Graduate 33% 29% 25%
Certicate (Non-University) 31% 33% 29%
University Degree or Certicate 27% 21% 23%
Note: 1% of the entire birth cohort never worked (therefore, do not receive the private- nor public-sector
classication).
retirement preparedness of Canadians in general. The sample does, however,
serve its purpose to test the continuity of living standards for a sample of work-
ers who achieve approximately a 70% conventional earnings replacement rate
(a target that assumes the complete cessation of employment at a single age of
retirement - an assumption that our reduced sample size suggests is unrealistic).
Table 1 compares the characteristics of the 1951–1958 birth cohort of sim-
ulated individuals who retire at age 61 with a 65%–75% conventional earnings
replacement rate, to those who retire at age 61 (with any conventional earnings
replacement rate level) as well as to those of the entire birth cohort. The sample
under examination is largely consistent with the broader birth cohort, except its
members are somewhat more highly educated and more likely to work in the
public sector (the sector status is assigned at age 60).
4. RESULTS
4.1. Result #1 — a wide range of living standards continuity
Figure 2 plots the distribution of the LSRR for the 1951–1958 birth cohort of
simulated individuals retiring at age 61 with a 65%–75% conventional earnings
available at http:/www.cambridge.org/core/terms. http://dx.doi.org/10.1017/asb.2016.20
Downloaded from http:/www.cambridge.org/core. IP address: 142.167.14.111, on 01 Oct 2016 at 18:18:06, subject to the Cambridge Core terms of use,
THE LIVING STANDARDS REPLACEMENT RATE (LSRR) 643
Living Standards Replacement Rate (1 = 100%)
Proportion of Sample Population (%)
012345
0.0 0.2 0.4 0.6 0.8
FIGURE 2: Distribution of working-life living standards replacement rates after retirement for the 1951–1958
birth cohort of simulated individuals retiring at age 61 with a 65%–75% conventional earnings replacement
rate. (Distribution smoothed using a Gaussian kernel with a bandwidth based on normal distribution
approximation.12)
Notes: Living Standards Replacement Rate is this paper’s measure of living standards continuity between
working-life and retirement (see Equation (2)). The horizontal bars enclose the range of approximate living
standards continuity (80%–120%). Note that the smoothing causes the left tail of the distribution to
incorrectly fall below zero.
replacement rate. The two dotted lines mark the LSRR at 80% and 120% (the
assumed range of living standards continuity — see Section 3.3). As Figure 2
shows, most of those people satisfying the narrow 65%–75% earnings replace-
ment rate criterion can actually expect to improve their living standards after
retirement (but to various degrees). Specically, some 80% of the sample will
improve their living standards by over 20% after retirement (that is, LSRR >
120%).
Figure 2 primarily shows that retirees satisfying the narrow 65%–75% earn-
ings replacement rate criterion can actually expect a wide range of changes in
living standards after retirement. This nding was consistent across supplemen-
tary analysis that examined other samples of workers (such as those retiring at
ages 65 and 70, and testing targets of 50%, 60% and 80%). Moreover, the ob-
served spread would have been even greater had the results included the risks
associated with post-retirement (such as ination and returns on investments).
(Recall that our simplifying annuitization assumption effectively removed post-
retirement risks, although in reality voluntary-annuitization is extremely low —
see Section 3.3.)
available at http:/www.cambridge.org/core/terms. http://dx.doi.org/10.1017/asb.2016.20
Downloaded from http:/www.cambridge.org/core. IP address: 142.167.14.111, on 01 Oct 2016 at 18:18:06, subject to the Cambridge Core terms of use,
644 B.-J. MACDONALD, L. OSBERG AND K.D. MOORE
(a) Four replacement rate (RR) measures
Rate (1 = 100%)
Proportion of Sample Population (%)
RR#1
Conventional RR from Equation (1)
RR#2
Conventional RR measured over
lifetime
RR#3
Conventional RR measured over
lifetime at household level
and adjusted for family size
Living Standards Replacement
Rate (from Equation (2))
012345
024681012
(b) Measure the numerator and denominator
over broader measurement period
Rate (1 = 100%)
Proportion of Sample Population (%)
RR#1
Conventional RR from Equation (1)
RR#2
Conventional RR measured over
lifetime
012345
024681012
(c) Each year, measure at the household level
and then adjust for family size
Rate (1 = 100%)
Proportion of Sample Population (%)
RR#2
Conventional RR measured
over lifetime
RR#3
Conventional RR measured
over lifetime at household level
and adjusted for family size
012345
0.0 0.2 0.4 0.6 0.8 1.0
(d) Use a fuller measure of income
available for consumption expenditure
Rate (1 = 100%)
Proportion of Sample Population (%)
RR#3
Conventional RR measured
over lifetime at household level
and adjusted for family size
Living Standards Replacement
Rate (from Equation (2))
012345
0.0 0.2 0.4 0.6 0.8 1.0
FIGURE 3: (a) Distribution of four replacement rate (RR) measures for the 1951–1958 birth cohort retiring at
age 61 with a 65%–75% conventional earnings replacement rate. Figures (b), (c), (d) open up Figure (a),
plotting each new measure one by one (and comparing it with the previous measure). (Other than RR#1,
distributions are smoothed using a Gaussian kernel with a bandwidth based on normal distribution
approximation). (a) Four replacement rate (RR) measures. (b) Measure the numerator and denominator over
broader measurement period. (c) Each year, measure at the household level and then adjust for family size. (d)
Use a fuller measure of income available for consumption expenditure.
Notes: Living Standards Replacement Rate is this paper’s measure of living standards continuity between
working-life and retirement (see Equation (2)). Conventional earnings replacement rate is the ratio of gross
retirement income in the rst year of retirement, divided by the subject’s gross earnings in the last year of
employment (see Equation (1)).
4.2. Result #2 — the omissions in the conventional earnings replacement rate
are signicant and interacting
The people in Figure 2 started retirement with a conventional earnings replace-
ment rate in the narrow range of 65%–75%. The dispersion of simulated living
standard continuity is striking. What is causing it? Figure 3 shows how each
available at http:/www.cambridge.org/core/terms. http://dx.doi.org/10.1017/asb.2016.20
Downloaded from http:/www.cambridge.org/core. IP address: 142.167.14.111, on 01 Oct 2016 at 18:18:06, subject to the Cambridge Core terms of use,
THE LIVING STANDARDS REPLACEMENT RATE (LSRR) 645
renement to the conventional earnings replacement rate calculation from
Equation (1) affects the distribution of results. Recall that Equation (1) used
only a single year before and after retirement, looked exclusively at individual
income and had a narrow denition of income. Section 3.3 broke down the evo-
lution of the conventional earnings replacement rate formula into the LSRR
formula using three steps — similarly, Figures 3(a)–3(d) examine the inuence
of measurement period, family size and income:
Measure the numerator and denominator over broader measurement period (as in the LSRR).
Each year, measure at the family level and then adjust for family size (as in the LSRR)
Use a fuller measure of income available for consumption expenditure (as in the LSRR).
RR#1: Conventional earnings replacement rate from Equation (1)
M
RR#2: Conventional earnings replacement rate measured over lifetime
E
RR#3: Conventional earnings replacement rate measured over lifetime at household level
U
LSRR from Equation (2)
Figure 3(a) graphically illustrates this progression between the conventional
earnings replacement rate (RR#1) to the LSRR by showing the distribution of
these four replacement rate measures for our sample of Canadians with a 65%–
75% conventional earnings replacement rate. Figures 3(b), (c) and (d) opens
up Figure 3(a) to more clearly show the relative change of moving from one
replacement rate concept to the next. We next discuss each in turn.
In the following discussion, recall that it is not so much the shift in the dis-
tribution that is relevant to our investigation, but the widening of the range.
Under the conventional earnings replacement rate theory, retirement income
adequacy occurs when approximately a 70% target is achieved, and under our
LSRR framework, it is achieved at approximately 100%. Accordingly, we ex-
pect a rightward shift in the distribution when moving from one replacement
rate concept to the next.
In Figure 3(b), the distribution of the conventional earnings replacement
rate (RR#1) is, by denition, limited to 65%–75%. A single year’s income is the
sum of permanent and transitory income; hence, broadening the measurement
period by moving from RR#1 (single year) to RR#2 (longer period) spreads
the distribution of actual replacement due to the inuence of transitory income
in both the numerator and the denominator of RR#1.
To illustrate this, let yxrepresent permanent income at age x and σxrep-
resent the transitory component at that age, where σxis drawn from a mean
zero random distribution f(x). Observed annual income at age xequals to yx+
available at http:/www.cambridge.org/core/terms. http://dx.doi.org/10.1017/asb.2016.20
Downloaded from http:/www.cambridge.org/core. IP address: 142.167.14.111, on 01 Oct 2016 at 18:18:06, subject to the Cambridge Core terms of use,
646 B.-J. MACDONALD, L. OSBERG AND K.D. MOORE
σx. When observed annual income is used to dene the 70% target at re-
tirement age 61, Equation (1) states that retirement income adequacy oc-
curs when (y62+σ62
y60+σ60 =70%). However, if we examine the permanent income re-
placement rate (y62
y60 )of those individuals who satisfy this 70% criterion using
observed annual income, manipulation of the previous equation shows that
(y62
y60 =70% +70% σ60
y60 σ62
y60 ).Sinceσ60 and σ62 are random variables that are
uncorrelated by construction, effectively, these transitory components in annual
income have a spreading out effect on the observed “permanent” earnings re-
placement rate at age 61. The choice of measurement period plays a huge role
in replacement rate outcomes, as has been observed in previous literature and
discussed in Section 2.
Using the family as the unit of analysis in Figure 3(c) (RR#2 to RR#3)
has the anticipated effect of marginally shifting the distribution of relative re-
tirement nancial well-being upward. The overall shift of the distribution is
largely due to the incorporation of dependent children over the working-life.
Accounting for dependent children reduces effective working-life consumption,
implying that any given amount of retirement income translates into a greater
improvement in living standards after retirement. Incorporating spouses also
affects replacement rates — for example, low-income spouses who relied on
their spouse’s income during working-years (e.g., while child-rearing) will re-
ceive senior benets from the government that will improve the household’s
standard of living after retirement. On the ip side, some individuals’ measured
replacement rate will decrease once his/her higher income spouse is accounted
for. The overall impact is, however, an upward shift since living standards conti-
nuity into retirement will be higher once the calculation recognizes that workers
need to share their income with dependent children. (We discuss this further in
Figure 4(c).)
Finally, moving from RR#3 to LSRR in Figure 3(d) generally improves the
retiree’s calculated nancial well-being relative to his/her working years because
(1) it reects the preferential tax treatment for Canadian seniors (much of which
comes into effect after age 6513); and (2) it incorporates the accumulation and
drawdown of savings, which simultaneously decreases the denominator and in-
creases the numerator of the replacement rate. Specically, consumption expen-
diture is diminished by the accumulation of savings (which includes mortgage
payments) during working-life, while it is elevated by the drawdown of savings
after retirement. Reducing the denominator and increasing the numerator both
contribute to improving the overall rate, creating the rightward shift between
RR#3 and the LSRR.
The shift of the distribution in Figure 3(d) would have been even more sig-
nicant had we included any drawdown of housing equity. Owner occupied
housing is the single largest net asset of most Canadians, and seniors are more
likely to own their home without a mortgage than any other age group (Statistics
Canada, 2006). Assuming that this asset is drawn upon in retirement would have
resultedinevenhigherLSRRs.
available at http:/www.cambridge.org/core/terms. http://dx.doi.org/10.1017/asb.2016.20
Downloaded from http:/www.cambridge.org/core. IP address: 142.167.14.111, on 01 Oct 2016 at 18:18:06, subject to the Cambridge Core terms of use,
THE LIVING STANDARDS REPLACEMENT RATE (LSRR) 647
(a) Four replacement rate (RR) measures
Rate (1 = 100%)
Proportion of Sample Population (%)
RR#1
Conventional RR from Equation (1)
RR#4
Conventional RR with fuller income
measure
RR#5
Conventional RR with fuller income
measure at household level
and adjusted for family size
Living Standards Replacement
Rate (from Equation (2))
012345
024681012
(b) Use a fuller measure of income available
for consumption expenditure
Rate (1 = 100%)
Proportion of Sample Population (%)
RR#1
Conventional RR from Equation (1)
RR#4
Conventional RR with fuller income
measure
012345
024681012
(c) Each year, measure at the household level
and then adjusted for family size
Rate (1 = 100%)
Proportion of Sample Population (%)
RR#4
Conventional RR with fuller
income measure
RR#5
Conventional RR with fuller
income measure at household
level and adjusted for
family size
012345
0.0 0.2 0.4 0.6 0.8 1.0
(d) Measure the numerator and denominator
over broader measurement period
Rate (1 = 100%)
Proportion of Sample Population (%)
RR#5
Conventional RR with fuller
income measure at household
level and adjusted for
family size
Living Standards Replacement
Rate (from Equation (2))
012345
0.0 0.2 0.4 0.6 0.8 1.0
FIGURE 4: (a) Distribution of four replacement rate (RR) measures for the 1951–1958 birth cohort retiring at
age 61 with a 65%–75% conventional earnings replacement rate. Figures (b), (c), (d) open up Figure (a),
plotting each new measure one by one (and comparing it with the previous measure). (Other than RR#1,
distributions smoothed using a Gaussian kernel with a bandwidth based on normal distribution
approximation). (a) Four replacement rate (RR) measures. (b) Use a fuller measure of income available for
consumption expenditure. (c) Each year, measure at the household level and then adjusted for family size. (d)
Measure the numerator and denominator over broader measurement period.
Notes: Living Standards Replacement Rate is this paper’s measure of living standards continuity between
working-life and retirement. (see Equation (2)). Conventional earnings replacement rate is the ratio of gross
retirement income in the rst year of retirement, divided by the subject’s gross earnings in the last year of
employment (see Equation (1)).
More relevant than the shift in Figure 3(d) is the width of the distribution
for the LSRR compared with RR#3, which is created by the varying impact
of including these various other sources of income in each replacement rate’s
numerator and denominator. One example is taxation, which is conventionally
assumed xed in replacement rate literature but has, in fact, differential impacts
between people and across the life-course owing to differences in
available at http:/www.cambridge.org/core/terms. http://dx.doi.org/10.1017/asb.2016.20
Downloaded from http:/www.cambridge.org/core. IP address: 142.167.14.111, on 01 Oct 2016 at 18:18:06, subject to the Cambridge Core terms of use,
648 B.-J. MACDONALD, L. OSBERG AND K.D. MOORE
the amount of total income;
tax deductions and tax credits according to personal circumstances (such as
medical expense credits for seniors with failing health);
the composition of income by source;
the distribution of income across spouses;
location (since taxes vary by province);
the forms that savings take.
It is reasonable to ask if the order of rening the conventional earnings re-
placement rate affects the impacts. We therefore reverse the steps by rst looking
at the inuence of income denition on the conventional earnings replacement
rate, then family size and nally measurement period. Beginning again with the
conventional earnings replacement rate and ending with the LSRR, we calcu-
late two new replacement rate measures, RR#4 and RR#5, using the following
procedure:
Figure 4 plots our results. Many of the same observations made for Figure 3
can be seen here.
Including all sources of income by moving from RR#1 to RR#4 in Figure
4(b) shifts and expands the distribution owing to the varying, and generally
favorable, impacts of these other sources of income on the replacement rate.
There is minimal impact of using the family as the unit of analysis in Figure
4(c). This occurs because the measurement period is only the year before retire-
ment (age 60), and therefore the most important family change, namely the ar-
rival and departure of dependent children, has occurred long ago for most peo-
ple. It is in Figure 4(d) that the improving impact of dependent children is real-
ized once the measurement period is expanded (which also widens the distribu-
tion owing to the varying impact of income in these other years). This compari-
son illustrates the importance of including dependent children when evaluating
living standards for the purpose of measuring retirement income adequacy.
The central message in Figures 3 and 4 is that the conventional earnings
replacement rate is not a robust indicator of living standards continuity —for
available at http:/www.cambridge.org/core/terms. http://dx.doi.org/10.1017/asb.2016.20
Downloaded from http:/www.cambridge.org/core. IP address: 142.167.14.111, on 01 Oct 2016 at 18:18:06, subject to the Cambridge Core terms of use,
THE LIVING STANDARDS REPLACEMENT RATE (LSRR) 649
example, it is immensely affected by broadening the measurement period (com-
paring RR#1 to RR#2), and the income concept (comparing RR#1 to RR#4).
It is commonly observed that there is a wide range of ndings across the re-
placement rate literature — whether in determining the “correct” target earn-
ings replacement rate or the proportion of the population prepared for retire-
ment (where, “within the economics profession, there is a lot of disagreement”
(Munnell, 2005: 3)). This is not surprising given that the improvements that an-
alysts have made to the conventional earnings replacement rate measure have
varied widely. (Appendix C reviews the various approaches taken across the lit-
erature.) Each improvement can have signicant and interacting impact on the
results (as Figures 3 and 4 illustrate), thus the variety of renements made across
studies have led to varying and often conicting conclusions.
This section shows that the choices made when building the replacement
rate regarding (1) unit of analysis (family versus the individual), (2) sources of
consumption and (3) measurement period all have large impacts. These choices
interact, moreover, and the effect of improving one may not emerge without
the other. The clearest example of the interaction and signicance of the omis-
sions is in the incorporation of family size in the conventional earnings re-
placement rate measure and the impact of the measurement period choice.
Some analysts incorporate family size into the conventional earnings replace-
ment rate, but without simultaneously opening up the measurement period (see
Appendix C.1). We nd that accounting for family composition (by moving
from RR#4 to RR#5 in Figure 4(c)) has little effect on the replacement rate
distribution if using the conventional pre-retirement measurement period of
one year. It is only when income is measured over years when children are de-
pendent that the replacement rate distribution is appreciably affected by fam-
ily composition (moving from RR#5 to LSRR in Figure 4(d)). As empha-
sized by Scholz and Seshadri (2009), the role of children in determining pre-
retirement living standards is crucial, although nearly universally ignored in
this line of research. In future years, the increasing number of young adults
living with their parents suggests that dependent children may, moreover, be
a consideration both during the pre-retirement period and also after retire-
ment. Between 1981 and 2011, the ratio of young adults (aged 20 to 29) liv-
ing with their parent(s) increased from 26.9% to 42.3% (Statistics Canada,
2011).
4.3. Result #3 — the 70% target does not appear to t any subgroup
Some argue that the 70% target works for specic groups, such as the middle-
class or public-sector workers, but we have not been able to locate a segment of
the population for whom the 70% conventional earnings replacement rate target
accurately predicts living standards continuity. Figure 5 segments the popula-
tion by work sector, family type, education and individual gross employment
earnings at age 60 to examine whether there is a specic characteristic(s) within
the sampled population driving the wider distribution. Controlling for these
available at http:/www.cambridge.org/core/terms. http://dx.doi.org/10.1017/asb.2016.20
Downloaded from http:/www.cambridge.org/core. IP address: 142.167.14.111, on 01 Oct 2016 at 18:18:06, subject to the Cambridge Core terms of use,
650 B.-J. MACDONALD, L. OSBERG AND K.D. MOORE
(a) Sector at Age 60
Living Standards Replacement Rate (1 = 100%)
Proportion of Sample Population (%)
All
Private Sector
Public Sector
012345
0.0 0.2 0.4 0.6 0.8 1.0
(b) Gross Employment Earnings at Age 60 (2012$)
Living Standards Replacement Rate (1 = 100%)
Proportion of Sample Population (%)
All
< $30000
$30,000−60,000
$60,000−90,000
$90,000−120,000
> $120,000
012345
0.0 0.2 0.4 0.6 0.8 1.0
(c) Education at Age 60
Living Standards Replacement Rate (1 = 100%)
Proportion of Sample Population (%)
All
Less than Secondary
Graduate
Secondary Graduate
Only
Non−university post−
secondary certificate
University post−
secondary degree
or certificate
012345
0.0 0.2 0.4 0.6 0.8 1.0
(d) Household Type at Age 60
Living Standards Replacement Rate (1 = 100%)
Proportion of Sample Population (%)
Couple
Single Male
Single Female
012345
0.0 0.2 0.4 0.6 0.8 1.0
FIGURE 5: Distribution of working-life living standards replacement rates after retirement for the 1951–1958
birth cohort of simulated individuals retiring at age 61 with a 65%–75% conventional earnings replacement
rate, by (a) work sector at age 60; (b) gross employment earnings at age 60; (c) education attainment at age 60
and (d) by household type at age 60 (distributions smoothed using a Gaussian kernel with a bandwidth based
on normal distribution approximation).
Notes: Living Standards Replacement Rate is this paper’s measure of living standards continuity between
working-life and retirement (see Equation (2)). Conventional earnings replacement rate is the ratio of gross
retirement income in the rst year of retirement, divided by the subject’s gross earnings in the last year of
employment (see Equation (1)).
characteristics does not appreciably change the distribution of LSRRs — that
is, the wide distributions remain nearly the same across the subgroups. The
only noticeable impact is between singles and couples, where the distribution is
shifted to the right, which is again largely a product of the more likely support
of children among couples (as discussed, the general impact of incorporating
children shifts the rate upward).
available at http:/www.cambridge.org/core/terms. http://dx.doi.org/10.1017/asb.2016.20
Downloaded from http:/www.cambridge.org/core. IP address: 142.167.14.111, on 01 Oct 2016 at 18:18:06, subject to the Cambridge Core terms of use,
THE LIVING STANDARDS REPLACEMENT RATE (LSRR) 651
4.4. Result #4 — the conventional earnings replacement rate and living
standards continuity are poorly correlated
Does the traditionally measured 70% earnings replacement rate offer much
guidance as a retirement income adequacy target for living standards continu-
ity? We conclude not. Among the simulated individuals who obtained a 65%–
75% conventional earnings replacement rate, only 22.5% actually achieved liv-
ing standards continuity between working-life and retirement (dened as 80% <
LSRR <120%). Once we included all the simulated individuals from that birth
cohort who retired at age 61 (that is, without ltering on the 65%–75% con-
ventional earnings replacement rate), we found that even a higher proportion
(28.5%) achieved living standards continuity. Testing people with lower (60%)
and higher (80%) conventional earnings replacement rate targets produced sim-
ilar results — between 20% and 30% of them achieved living standards conti-
nuity. This suggests that the conventional earnings replacement rate target has
little information content.
Is there any relationship between the conventional earnings replacement rate
(from Equation (1)) and living standards continuity (estimated using the LSRR
in Equation (2))? If conventional earnings replacement rates had some valid-
ity, we would expect that people with higher conventional earnings replacement
rates would, on average, also have improved living standards in retirement, and
vice versa. This is not the case. Among the simulated individuals who retire at
age 61 (not just those with a 65%–75% conventional earnings replacement rate),
we nd that the correlation (Pearson product–moment correlation) between
conventional earnings replacement rates and the LSRR is only 0.11, suggest-
ing that the conventional earnings replacement rate formula is a very unreliable
measure of living standards continuity. Indeed, Figure 6 shows that the distribu-
tion of living standards continuity given in Figure 2 remains nearly unchanged
regardless of whether we had captured workers with a 50%, 70%, or 100% con-
ventional earnings replacement rate. The lack of movement in the distribution
signies that the earnings replacement rate simply does not predict living stan-
dards continuity between working-life and retirement.
5. ADOPTING THE LSRR — WHY,WHO AND HOW?
5.1. Why?
“One-size-ts-all” nal employment earnings replacement rate targets have his-
torically been used widely to evaluate retirement income adequacy because com-
prehensive individualized measures are difcult to compute (often on account of
data constraints). This paper nds, however, that any single year’s employment
earnings does not provide a good estimate of working-life living standards, and
therefore is not a reliable benchmark for determining retirement income ade-
quacy. If retirement income adequacy is to be meaningfully tested, it is necessary
available at http:/www.cambridge.org/core/terms. http://dx.doi.org/10.1017/asb.2016.20
Downloaded from http:/www.cambridge.org/core. IP address: 142.167.14.111, on 01 Oct 2016 at 18:18:06, subject to the Cambridge Core terms of use,
652 B.-J. MACDONALD, L. OSBERG AND K.D. MOORE
Living Standards Replacement Rate (1 = 100%)
Proportion of Sample Population (%)
50% Conventional RR
70% Conventional RR
100% Conventional RR
012345
0.0 0.2 0.4 0.6 0.8
FIGURE 6: Distribution of working-life living standards replacement rates after retirement for the 1951–1958
birth cohort of simulated individuals retiring at age 61 with a conventional earnings replacement rate
45%–55% (dashed line), 65%–75% (solid line), and 95%–105% (dotted line). (Distribution smoothed using a
Gaussian kernel with a bandwidth based on normal distribution approximation.)
Notes: Living Standards Replacement Rate is this paper’s measure of living standards continuity between
working-life and retirement (see Equation (2)). The horizontal bars enclose the range of approximate living
standards continuity (80%–120%). Note that the smoothing causes the left tail of the distribution to
incorrectly fall below zero.
to develop individualized measures using more comprehensive and evidence-
based assessments of pre- and post-retirement living standards.
Nearly three decades ago, Palmer (1988) developed target gross earnings
replacement rates by assigning average levels of social security benets, taxes
and savings rates (based on survey data) to workers in various subgroups
(differentiated by broad characteristics earnings level, region and marital sta-
tus) (see Appendix B for a description of the methodology). This work led to
a long line of follow-up reports, whose results have been widely disseminated.
The 2008 Palmer report wrote:
For twenty years, Aon Consulting and Georgia State University have pub-
lished data on retirement income needs. The Replacement Ratio StudyTM
has become a premier source of retirement planning information for
employers, employees, and their advisors. The 2008 Replacement Ratio
Study is the seventh update of this report and builds on a 1980 edi-
tionissued bythePresident’s Commissionon PensionPolicy...This2008
available at http:/www.cambridge.org/core/terms. http://dx.doi.org/10.1017/asb.2016.20
Downloaded from http:/www.cambridge.org/core. IP address: 142.167.14.111, on 01 Oct 2016 at 18:18:06, subject to the Cambridge Core terms of use,
THE LIVING STANDARDS REPLACEMENT RATE (LSRR) 653
update continues to answer the original question, ‘How much income will
I need at retirement to maintain my standard of living? Palmer (2008: 2)
A noteworthy user of these rates is the popular U.S. National Retirement
Risk Index (NRRI) (Munnell et al., 2006), which has garnered substantial at-
tention in both academia and in the media. In 2013, the NRRI study pre-
dicted that almost half of Americans are nancially unprepared for retirement
(Munnell et al., 2013), using Palmer’s targets as its measure of retirement income
adequacy.
The rst issue is that these target rates omit or poorly deal with the many
components of living standards discussed throughout this paper (and listed
in Section 2). Second, the Palmer target rates were developed based on styl-
ized individuals — the problems associated with stylized individuals have been
discussed by, among others, Steuerle et al. (2000). Third, the authors have been
unable to locate any empirical demonstration that the targets actually accom-
plish their goal — that is, for a sufcient sample of workers who hit these pre-
scribed targets, living standards are, in fact, approximately maintained after re-
tirement.
In the three decades since its inception, moreover, the underlying method-
ology has not changed very much. On the other hand, the world of employer
pensions, family dynamics and nancial market expectations have changed dras-
tically. At one time, it may have seemed justiable that earnings in the year
leading up to retirement should be the benchmark for retirement income ad-
equacy, as it was more common to spend entire careers with the same employer,
when salaries were much more stable and DB pension plans were the norm.
In addition, married couples were very unlikely to divorce after retirement and
seniors were most likely cared for by family and friends in the case of chronic
illness (see Appendix A for supporting evidence). This is no longer the case.
Retirement income adequacy is a growing concern in Canada and industrial-
ized countries around the world due to the low interest rate environment, high
investment fees, longer life expectancies, rising divorce rates among seniors, ag-
ing populations, an increasing reliance on paid services to care for chronic health
conditions, the shift from DB to DC pension plans and reduced condence in
nancial markets — see Appendix A. In addition to signaling a greater need for
accurate measures of retirement income adequacy, these changes in culture and
the economic environment strongly suggest that the simplistic assumptions that
may have underlain the conventional earnings replacement rate are insupport-
able.
By solely relying on employment earnings, and doing so in a single year,
the conventional earnings replacement rate does not capture living standards.
Housing wealth, nancial debt, savings decisions, number of children and the
earnings of one’s spouse are among the many determinants of a person’s living
standards, but the conventional earnings replacement rate ignores them. Con-
sider two male 40 year olds, who both earn $50,000. The conventional earn-
ings replacement rate would advise them to save enough so that their retirement
available at http:/www.cambridge.org/core/terms. http://dx.doi.org/10.1017/asb.2016.20
Downloaded from http:/www.cambridge.org/core. IP address: 142.167.14.111, on 01 Oct 2016 at 18:18:06, subject to the Cambridge Core terms of use,
654 B.-J. MACDONALD, L. OSBERG AND K.D. MOORE
income is $35,000 per year (assuming that wage growth keeps pace with ina-
tion). The rst man has, however, a wife who is a homemaker, three children and
rents his home (at $1,500/month). The second man has a wife who earns twice
his income, no children and paid off his home two years ago. Even if we make
the simplifying assumption that both men saved 10% of their gross earnings, and
their homes are comparable, the estimated value of the second man’s personal
living standard is over ve times that of the rst man ($85,000 versus $16,000,
assuming 2015 average tax rates in Ontario). If both men targeted $35,000 in re-
tirement, the second man would experience a 70% drop in his living standards
after retirement, while the rst man would increase his living standards by 70%.
In other words, the 70% target would have led the rst man to over-save signif-
icantly (likely at the expense of his family’s needs) and led the second man to
under-save to a similar degree.
A second major problem when relying exclusively on a single year’s employ-
ment earnings to estimate pre-retirement living standards is that the employ-
ment earnings of actual people can swing dramatically from year to year (see
Section 2). So, not only is any single year’s employment earnings unlikely to
approximate working-life living standards, it may not even represent “typical”
earnings for that individual. Consequently, in addition to the necessity to cap-
ture the full nancial picture of a person’s living standards (presented in Figure
1), it is also necessary to calculate average working-life living standards over a
sufciently representative number of years.
5.2. Who and how?
One of the primary goals typically identied for retirement income policy is
that individuals should not experience substantial declines in their standard of
living upon retirement — the LSRR is a measure that directly evaluates this
policy goal. Another common objective of policy analysts when evaluating re-
tirement income adequacy is to identify vulnerable groups in the population,
as well as understand the diverse implications of policy features, and proposed
changes, across the population. The LSRR could be employed to identify the
“winners” and “losers” of policy designs by providing a meaningful measure
of retirement income adequacy, where the changes in an individual’s retirement
living standards are compared directly to his/her own personal pre-retirement
living standards (and not to a universal and misleading target). Changing LSRR
values signal “who” is affected, and the degree of change measures “how much”
that person is affected in relation to his/her personal living standard.
Employer pension plan sponsors may well be interested in designing a pen-
sion plan that, in conjunction with government benets, will adequately serve a
specic “target” type of employee — for example, career employees who retire
at age 65 and follow a plausible, or “typical”, career path. In this respect, the
LSRR concept is a useful metric for understanding the true effectiveness of de-
sign formulas in serving the type of employee that the plan is targeting. When
available at http:/www.cambridge.org/core/terms. http://dx.doi.org/10.1017/asb.2016.20
Downloaded from http:/www.cambridge.org/core. IP address: 142.167.14.111, on 01 Oct 2016 at 18:18:06, subject to the Cambridge Core terms of use,
THE LIVING STANDARDS REPLACEMENT RATE (LSRR) 655
building the LSRR, therefore, the pension plan sponsor’s analyst may choose
to use “representative” levels of housing wealth, number of children, spousal
earnings and other components of the LSRR suitable to the type of employee
that the plan is targeting. A useful follow-up analysis would be to then use the
actual data of a few sample employees to examine the validity of the stylized
results. This approach would not only likely be more in line with the objectives
of the plan sponsor, but would also reduce the data and modeling requirements
tremendously. In fact, such calculations could be carried out deterministically
with relative simplicity.
Another line of analysis would be if a plan sponsor wanted to use the LSRR
to test the employee populations’ ability to retire. With the greater shift to DC
pension plans, where the employer has much less control within the pension
plan design to encourage a particular age of retirement, understanding the true
retirement preparedness of employees can help employers take proactive mea-
sures to help employees who need additional retirement planning assistance.
This helps to ensure an orderly transition of the workforce and reduce other
costs associated with having older employees continuing to work past an opti-
mal age. In this type of investigation, the analyst would use individual-specic
LSRR inputs for each employee when possible, and make assumptions to ll in
gaps.14
Employers could also use the LSRR to help employees test their personal
level of retirement nancial preparedness and make retirement savings deci-
sions. Such a service would be very benecial to an employee who does not
t the “targeted” employee prole, such as somebody who joins the plan later
in life or wishes to retire early. Similarly, for workers saving for retirement and
the nancial analysts helping them, the LSRR provides guidance on how to
evaluate current living standards and determine how much retirement income
should be aimed for if living standards continuity is the goal. It provides a more
understandable measure so that workers can better appreciate the true impact
of alternative nancial planning decisions on living standards continuity after
retirement. Some workers may wish for better living standards after retirement
(i.e., more money to spend to support personal consumption) and thus they
would aim for more than 100%, and the opposite may also be true.
In Canada, Eckler Ltd.’s Guided Outcomes (GO) is an industry example of
a model that employs the LSRR for its DC plan sponsor clients to evaluate pen-
sion design and support participants. This model uses both deterministic and
stochastic elements in its projections. In terms of personal nancial planning,
Avery and Morrisons (2009) Ruthen package employs the LSRR framework
to compare alternative nancial strategies for individuals; it uses sophisticated
dynamic micro-simulation modeling that captures both consumption outcomes
and the dispersion that emerges from a variety of risks.
Section 3.3 presented the LSRR framework, and discussed some additional
considerations when building the LSRR for both industry and policy analysts.
Readers wishing to adopt the LSRR to industry should, however, keep in mind
that our tool of analysis was a large-scale micro-simulation model, and some
available at http:/www.cambridge.org/core/terms. http://dx.doi.org/10.1017/asb.2016.20
Downloaded from http:/www.cambridge.org/core. IP address: 142.167.14.111, on 01 Oct 2016 at 18:18:06, subject to the Cambridge Core terms of use,
656 B.-J. MACDONALD, L. OSBERG AND K.D. MOORE
components of our analysis design would be overly burdensome for industry
analysis. For example, when calculating “working-life” living standards (that is,
the denominator of LSRR), the general goal of the industry analyst should be
to capture a representative number of years. This paper had the opportunity to
look at the whole working-career (30 years) and average across the most “rep-
resentative” 20 years, which is likely impractical for industry analysis. The mes-
sage of this paper is that it is crucial that the calculation covers a representative
number of years. Overall, “the more, the better” when capturing pre-retirement
living standards, but the entire career is not necessary. When calculating “retire-
ment” living standards in the LSRR numerator, the industry analyst will simi-
larly aim to capture a “representative” number of years. Relative to the working
life, retirement income shows much less volatility (see Appendix C-3 for further
discussion). Nevertheless, employer pension plan benets that are not indexed
to ination will increasingly lose value over time. As a result, using a snapshot
of retirement income in the early years of retirement can overstate the real value
of the retirement income levels. In our analysis, we averaged living standards
from retirement until death, but this can be impractical for industry analysts
adopting the LSRR. One shortcut is to employ a single year in retirement, but
chose a later age, such as age 75. Using a later age incorporates some of the
eroding effect of ination over retirement, as well as avoids having to deal with
the complicated transition years between employment and retirement (such as
part-time employment and partial retirement benets).
A further question when applying the LSRR is the treatment of high-/low-
income workers when assessing the retirement preparedness of employer pen-
sion plan participants. The LSRR is a relative measure, and therefore for peo-
ple who were poor during their working years, “continuity in living standards”
means being poor in old age too (in other words, continuity does not imply
true adequacy). Similarly, the relative LSRR measure may fall below 100% for
afuent workers who would otherwise be viewed as adequately prepared for
retirement. One approach is to combine the relative LSRR measure with an
absolute “low” and “high” measure of retirement living standards. For exam-
ple, the analyst could ag individuals whose LSRR numerator falls below an
absolute value intended to signal poverty (such as the “HRSDC Market Basket
Measure”, which represents the basic cost of necessities for individuals living in
Canada) or above an absolute measure whose value is deemed adequately suf-
cient to support an afuent standard of living. The treatment of low- and high-
income employees depends largely on the goals of the pension plan sponsor.
A common concern when applying the LSRR to industry has been “im-
puted rent”. While academic analysis tends to include theoretical consumption,
practitioners and individuals who are planning for retirement tend to think in
terms of cash ow needs. Although not commonly used in industry, imputed
rent is an important consideration in the analysis of retirement income ade-
quacy since it enables analysts to incorporate the differing nancial situations
of a homeowner versus a non-homeowner. In terms of actual cash ow needs,
it accounts for the freed-up “money to support personal consumption” of a
available at http:/www.cambridge.org/core/terms. http://dx.doi.org/10.1017/asb.2016.20
Downloaded from http:/www.cambridge.org/core. IP address: 142.167.14.111, on 01 Oct 2016 at 18:18:06, subject to the Cambridge Core terms of use,
THE LIVING STANDARDS REPLACEMENT RATE (LSRR) 657
homeowner who does not need to pay rent for his/her shelter, and therefore a
relevant and recommended component of the living standards calculation (see
Appendix C.2 for further discussion on imputed rent).
A nal comment is that the future is uncertain, so the LSRR values (like any
metric) should be routinely reassessed and inputs adjusted to recognize actual
experience.
6. CONCLUSION
The conventional nal employment earnings replacement rate is a widely used
benchmark to evaluate retirement income adequacy, commonly employed by
public policy analysts, employer pension plan advisors and academics when
evaluating the retirement preparedness of a population or comparing pension
systems, as well as by nancial advisors and individuals making retirement -
nancial planning decisions. It underlies pension systems, drives the research that
determines whether populations are prepared or not prepared for retirement,
and is the backbone of retirement planning software. Our analysis nds that it
predicts living standards continuity in retirement very poorly.
Specically, we nd that workers who approximately attain a nal employ-
ment earnings replacement of 70%, — i.e., actually hit the “right” target — will
experience a wide range of living standards continuity after retirement. This is
problematic at both extremes — both for those who end up experiencing rela-
tive deprivation in retirement and for those whose relatively high consumption
during their retirement years was purchased by the over-sacrice of working-life
welfare. We have been unable to locate a segment of the population for whom
the 70% target accurately predicts standard of living continuity.
This paper nds that the conventional earnings replacement rate is a poor
metric for understanding how well an individual’s living standards are main-
tained after retirement. The primary reason is that a single year’s employment
earnings are not a reliable representation of working-life living standards. The
conventional earnings replacement rate formula relies on an inadequate mea-
surement period, does not incorporate important components of consumption,
and ignores family size. These omissions are crucial to accurate estimation of
pre-retirement living standards. Further, we nd that these omissions interact,
and rening the conventional earnings replacement rate measure requires si-
multaneous improvement in all three of these spheres since the full effect of
improving one may not emerge without the others. These signicant and inter-
acting omissions have led to a wide range of often conicting reports on the
retirement preparedness of populations.
Analysts have suggested that the use of single replacement rate “tar-
gets” are not reasonable owing to the existence of risk, and our results go
beyond this and suggest that the conventional earnings replacement rate has
little predictive value owing to its omission of critical components of living stan-
dards in its formula. Our paper demonstrates that before one can decide whether
available at http:/www.cambridge.org/core/terms. http://dx.doi.org/10.1017/asb.2016.20
Downloaded from http:/www.cambridge.org/core. IP address: 142.167.14.111, on 01 Oct 2016 at 18:18:06, subject to the Cambridge Core terms of use,
658 B.-J. MACDONALD, L. OSBERG AND K.D. MOORE
living standards in retirement are adequate or inadequate, one needs a measure
that accurately measures living standards both before and after retirement.
Replacement rate statistics play an important role in providing a single sta-
tistical criterion for retirement income adequacy, and the authors recognize the
importance of “Not throwing the baby out with the bathwater”. Accordingly,
this paper suggests the LSRR as an alternative metric for assessing how well a
worker’s living standards are maintained after retirement. The LSRR is a more
accurate, understandable, and consistent measure of living standards continu-
ity for the purpose of measuring retirement income adequacy. The goal of the
LSRR is to determine how well a worker’s living standards will be maintained
after retirement, by comparing how much money a worker has available to sup-
port his/her personal consumption of goods and services before and after retire-
ment. As outlined in this paper, it is crucial that the LSRR calculation considers
the entire family, includes consumption components comprehensively and cov-
ers a representative number of years.
The framework presented in this paper is intended to provide a guide for aca-
demics, nancial planners, employer pension plan advisors and policy makers to
follow when analyzing individuals or populations. While the conventional earn-
ings replacement rate has been calculated in a wide variety of ways, the LSRR
framework will enable a more consistent measure of retirement income ade-
quacy that analysts can reference, which will facilitate the interpretation, com-
parison and integration of ndings across different analysis (between authors,
over time and across nations), greatly helping the study of retirement income
adequacy to move forward.
Employer pension plan sponsors could use the LSRR to investigate the
retirement income adequacy provided by their pension plan and savings pro-
gram for the “typical” type of employee that the plan is targeting, test the
true effectiveness of adopting different plan features, assess the level of retire-
ment preparedness among its plan participants, and engage plan participants
to help them better appreciate the true impact of alternative nancial planning
decisions on their living standards after retirement. Policy analysts could use
the LSRR to evaluate the retirement income adequacy of their current system
across the population, identify vulnerable groups, and understand the impli-
cation of policy changes. Financial advisors could use the LSRR as a retire-
ment income target that clients can understand when making nancial planning
decisions. The LSRR can be computed with sophisticated dynamic stochastic
micro-simulation modeling, simple deterministic projections or something in
between. In all of these roles, the analysts will inevitably run into gaps in the
data, which they will need to ll. Nevertheless, data constraints should not cre-
ate a barrier to adopting the LSRR. It is far better to acknowledge and grapple
with weaknesses and gaps, and make informed assumptions when necessary,
than to blindly rely on simplistic and invalid benchmark targets. Analysts invest
signicant time and effort in the study of retirement income adequacy, but em-
ploying an unreliable benchmark for “adequacy” not only effectively invalidates
that effort, but it can also lead to misleading, and possibly harmful, conclusions.
available at http:/www.cambridge.org/core/terms. http://dx.doi.org/10.1017/asb.2016.20
Downloaded from http:/www.cambridge.org/core. IP address: 142.167.14.111, on 01 Oct 2016 at 18:18:06, subject to the Cambridge Core terms of use,
THE LIVING STANDARDS REPLACEMENT RATE (LSRR) 659
With population aging in many industrialized countries, now more than ever is
the time to adopt better measures of retirement income adequacy.
ACKNOWLEDGEMENTS
We would like to thank Malcolm Hamilton, James Davis, Deborah Ng, Fred
Vettese, Keith Ambachtsheer, Rob Brown, Richard J. Morrison, Janice Hol-
man, Eckler Ltd, and the Economics Department of Dalhousie University. We
also wish to thank an anonymous reviewer for his/her helpful comments. This
project began in response to a call for proposals by the Society of Actuaries’
Pension Section over ve years ago, and we gratefully acknowledge their nan-
cial support and the feedback given by the project oversight group on earlier
drafts. This paper was helped funded by the Rotman International Center for
Pension Management (ICPM) at the Rotman School of Management, Univer-
sity of Toronto, whom we thank. We thank the ICPM review committee for their
valuable feedback. Finally, progressive versions of this work was presented at
the following gatherings, and we thank and acknowledge the organizers of these
events and the valuable feedback from its audiences: NEST Insight 2016 Confer-
ence, Benets Canada 17th Annual Dened Contribution Plan Summit (2016),
Association of Canadian Pension Management Panel Discussion (2016), Rot-
man International Centre for Pension Management 2014 Discussion Forum,
30th International Congress of Actuaries (2013), Towson University Actuarial
Science 2013 Annual Colloquium, North America Society of Actuaries Webcast
Speaker on Retirement Income Adequacy (2012) and 2011 Canadian Institute
of Actuaries Pension Seminar, Toronto. An earlier draft of this paper won the
ICA 2014 Pension, Benets and Social Security Scientic Committee Award
Prize for Best Paper at the 30th International Congress of Actuaries in April
2014.
NOTES
1. Appendix A discusses each of these seven factors.
2. Between 2001 and 2009, the median retirement age for Canadians has ranged from 60.6 to
61.9 years (Schwartz, 2010).
3. The opinions expressed and conclusions reached by the authors are their own and do not
represent any ofcial position or opinion of Statistics Canada. We take full responsibility for the
assumptions underlying the projection scenario used.
4. Examples of studies that prescribe or depend on replacement rates as a criterion for retire-
ment income adequacy (in terms of living standards continuity) include Dexter (1984), Boskin and
Shoven (1987), Palmer (1988, 2008), Gustman and Steinmeier (1998), Mitchell and Moore (1998),
Moore and Mitchell (2000), Alford et al. (2004), Schieber (2004), Steinberg and Lucas (2004),
Vanderhei (2004), Munnell and Soto (2005), Haveman et al. (2006), Munnell et al. (2006), Munnell
et al. (2007a, 2007b), OECD (2009), Brady (2010), Dodge et al. (2010), McGill et al. (2010), TD
Economics (2010), Munnell et al. (2011), U.K. Government DWP (2013) and Shillington (2016).
5. Sustaining living standards after retirement is the most prevalent denition of retirement
income adequacy in the replacement rate literature (see footnote 4 for samples of published re-
search), although not the universal one. Engen et al. (1999) and Scholz and Seshadri (2009) are
examples of authors who employ replacement rates but apply a utility maximization framework
in an augmented life-cycle model. They argue that retirement income adequacy occurs when the
discounted marginal utility of consumption is smoothed over time.
available at http:/www.cambridge.org/core/terms. http://dx.doi.org/10.1017/asb.2016.20
Downloaded from http:/www.cambridge.org/core. IP address: 142.167.14.111, on 01 Oct 2016 at 18:18:06, subject to the Cambridge Core terms of use,
660 B.-J. MACDONALD, L. OSBERG AND K.D. MOORE
6. While replacement rates are most commonly used as a prescriptive statistic that signals re-
tirement income adequacy, they have also been used as a descriptive statistic to examine trends
over time and between groups of people (e.g., Boskin and Shoven, 1987; GAO, 2001; Smith, 2003;
Butrica et al., 2003; Fidelity, 2007; LaRochelle-Cote et al., 2008; Ostrovsky and Schellenberg,
2009).
7. In Canada, “registered” plans (such as Registered Retirement Savings Plans (RRSPs) and
Registered Retirement Income Funds (RRIFs)) are tax-sheltered retirement savings, where con-
tributions and investment returns are tax-exempt while taxation occurs when the funds are with-
drawn.
8. To capture the income pooling and economies of scale that individuals experience within
a household, a commonly used equivalence scale is the square root of household size (Buhmann
et al., 1988), i.e., if two households had the same income ($X), but one was a single-person house-
hold and the other had four members (e.g., two adults and two children), then the equivalent in-
come of the person in the one-person household would be $X, while each person in the four-person
household would be assigned an equivalent income of $X/4=$X/2. Note that we measure at
the census-family level as LifePaths does not model economic households.
9. A drop in necessary expenditure after retirement to sustain pre-retirement living standards
is also suggested by research explaining the “retirement consumption puzzle”, which nds that re-
tirement provides additional leisure time for home-production, which lowers the expense of main-
taining working life living standards (Aguiar and Hurst, 2005; Brzozowski and Lu, 2010).
10. For annuities purchased with non-registered funds, we assume the Canadian 2012 taxation
treatment of “prescribed annuities” (calculating the taxed amount as a proportion of the annuity
payment). Owing to the ination indexation assumption, however, the appropriate tax rate would
depend on the more complex treatment of “non-prescribed annuities” (see Milevsky, 2010 for fur-
ther information). However, the empirical impact of this simplifying assumption is trivial.
11. For the purpose of federal employment insurance benets, Canadians are considered “at-
tached” to the labor force if they have worked 490 hours in the previous year (which is 24% of the
year assuming a standard 40 hour work-week), thus leading to the choice of 25% in our retirement
concept. Source: www.servicecanada.gc.ca/eng/ei/types/regular.shtml#Number
12. Wikipedia provides a good explanation of kernel smoothing, which includes a descrip-
tion of Gaussian kernel smoothing (https://en.wikipedia.org/wiki/Kernel smoother#Gaussian
kernel smoother)
13. In 2013, these advantages included: the splitting of certain types of income from pensions
and registered savings between Canadian spouses; up to 6,854 of personal federal income tax ex-
emption for people 65 years plus; a federal tax-exemption for the rst 2,000 of pension income;
the exemption of Guaranteed Income Supplement (low-income senior benet) and corresponding
provincial/territorial exemptions.
14. For example, Canadians have approximately two children on average, so this would be a
reasonable assumption in the absence of data. The analyst may wish to test the impact of assum-
ing more or less children, and the implications for adequacy. Similarly, most Canadians own a
home at retirement, so the analyst could use regional housing prices to assign housing values by
earnings. The LSRR calculation is simplied further if the employer pension plan sponsor is inter-
ested in the adequacy of the pension plan (combined with government benets) without depending
on supplementary individual savings from employees. In this case, the analyst simply assumes no
savings/debt in the LSRR calculation, which would eliminate the necessity to collect that data (or
otherwise input assumptions).
15. “Earnings mobility refers to changes in the relative earnings of individual workers through
time” (Beach and Finnie, 2004: 5).
16. See, for example, the line of research by the Society of Actuaries Committee on Post-
Retirement Needs and Risks at http://www.soa.org/research/research-projects/pension/research-
post-retirement-needs-and-risks.aspx
REFERENCES
AGUIAR,M.andHURST, E. (2005) Consumption versus expenditure. Journal of Political Economy,
113(5), 919–948.
available at http:/www.cambridge.org/core/terms. http://dx.doi.org/10.1017/asb.2016.20
Downloaded from http:/www.cambridge.org/core. IP address: 142.167.14.111, on 01 Oct 2016 at 18:18:06, subject to the Cambridge Core terms of use,
THE LIVING STANDARDS REPLACEMENT RATE (LSRR) 661
ALFORD, S., FARNEN, D.B., and SCHACHET, M. (2004) Affordable Retirement: Light at the End
of the Tunnel. Benets Quarterly, 20(4), 7–14.
AVERY,M.andMORRISON, R.J. (2009) Spending down one’s retirement assets in an uncertain
world. Prepared for the June 2009 Meetings of the International Microsimulation Association,
Ottawa, Ontario, Canada.
BALDWIN, B. (2009) Research study on the Canadian retirement income system: Final report.
Prepared for the Ministry of Finance, Government of Ontario.
BEACH,C.andFINNIE, R. (2004) A longitudinal analysis of earnings change in Canada. Canadian
Journal of Economics,37(1), 219–240.
BERNHEIM, B.D., FORNI, L., GOKHALE,J.,andKOLTIKOFF, L.J. (May 2000) How Much
Should Americans Be Saving for Retirement? American Economic Review,90(2), 288–292.
doi:10.1257/aer.90.2.288.
BIGGS, A.G., and SPRINGSTEAD, G.R. (2008) Alternate Measures of Replacement Rates for Social
Security Benets and Retirement Income. Social Security Bulletin,68(2), 1–19.
BINSWANGER,J.andSCHUNK, D. (2012) What is an adequate standard of living during Retire-
ment?. Journal of Pension Economics and Finance,11(2), 203–222
BOSKIN,M.andSHOVEN, J.B. (1987) Concepts and measures of earnings replacement during re-
tirement. In Issues in Pension Economics. (eds. Z. Bodie, J.B. Shoven and D.A. Wise), Chicago:
University of Chicago Press.
BRADY, P. (2010) Measuring retirement resource adequacy. Journal of Pension Economics and Fi-
nance,9(2), 235–262.
BRANCATI, C.U., BEACH, B., FRANKLIN,B.,andJONES, M. (2014) Understanding Retirement Jour-
neys: Expectations vs. Reality. London: International Longevity Centre UK. November.
BROWN, J.R. (2009) Understanding the role of annuities in retirement planning. In Overcoming the
Saving Slump (A. Lusardi ed.), Chicago: University of Chicago Press.
BROWN, S.L. and LIN, I.-F. (2012) The gray divorce revolution: Rising divorce among middle-
aged and older adults, 1990–2010. Journals of Gerontology Series B: Psychological Sciences
and Social Sciences,67(6), 731–741.
BROWN, W., HOU,F.andLAFRANCE, A. (2010) Incomes of retirement-age and working-age Cana-
dians: Accounting for home wwnership, Statistics Canada Economic Analysis Research Paper
Series, Ottawa, July 2010.
BRZOZOWSKI,M.andLU, Y. (2010) Home cooking, food consumption, and food production
among retired Canadian households. Canadian Public Policy,36(1), 107–128.
BUHMANN, B., RAINTER, L., SCHMAUS,G.andSMEEDING, T. (1988) Equivalence scales, well-
being, inequality and poverty: Estimates across ten countries using the luxembourg income
study database. Review of Income and Wealth,34(2), 115–142.
BUTRICA, B.A., IAMS, H.M. and SMITH, K.E. (2006) The changing impact of social security
on retirement income in the United States. In The Distributional Effects of Government
Spending and Taxation,edited by Dimitri B. Papadimitriou, 112–132. Palgrave Macmillan UK,
http://link.springer.com/chapter/10.1057/9780230378605 4.
CHIURI, M.C. and JAPPELLI, T. (2010) Do the elderly reduce housing equity? An international
comparison. Journal of Population Economics 23(2), 643–63. http://dx.doi.org/10.1007/s00148-
008-0217-4.
CLARK, W. (2005) What do seniors spend on housing?. Canadian Social Trends, 78, 2–7 (Statistics
Canada Catalogue No. 11–008. Ottawa).
DEKKERS,G.andBELLONI, M. (2009) Microsimulation, pension adequacy and the dynamic
model MIDAS: An introduction, projection AIM – deliverable 4.10.
DENTON, F.T., FINNIE,R.andSPENCER, B.G. (2011) Income replacement in retirement: Longitu-
dinal evidence from income tax records. Canadian Journal on Aging,30(3), 1–17.
DENTON, F.T. and SPENCER, B.G. (2009) What is retirement? A review and assessment of alterna-
tive concepts and measures. Canadian Journal on Aging,28(1), 63–76.
DEXTER, M. (1984) Replacement ratios: A major issue in employee pension systems. In National
Committee on Public Employee Pension Systems, pp. 42–43. Washington, D.C.: National Com-
mittee on Public Employee Pension Systems, 1984.
DODGE, D.A., LAURIN,A.andBUSBY, C. (2010) The piggy bank index: Matching Canadians’
saving rates to their retirement dreams, C.D. Howe Institute. E-brief no. 95, Toronto, March
2010. Available at: http://www.cdhowe.org/pdf/ebrief 95.pdf
available at http:/www.cambridge.org/core/terms. http://dx.doi.org/10.1017/asb.2016.20
Downloaded from http:/www.cambridge.org/core. IP address: 142.167.14.111, on 01 Oct 2016 at 18:18:06, subject to the Cambridge Core terms of use,
662 B.-J. MACDONALD, L. OSBERG AND K.D. MOORE
ENGEN, E.M., GALE, W.G., and UCCELLO, C.E. (1999) The Adequacy of Household Saving.
Brookings Papers on Economic Activity,2, 65–187.
FERNANDEZ-VILLAVERDE,J.andKRUEGER, D. (2007) Consumption over the life cycle: Facts
from consumer expenditure survey data. The Review of Economics and Statistics,89(3), 552–
565.
Fidelity Research Institute (2007) The Fidelity Research Institute Retirement Index. Research In-
sights Brief; March.
FINNIE, R. (1999) Earnings mobility of Canadians, 1982–1992, Perspectives on Labour and Income.
Statistics Canada Catalogue No. 75-001.XPE: 9-15. Statistics Canada.
FINNIE,R.andGRAY, D. (2011) Labour-force participation of older displaced workers in Canada:
Should i stay or should i go?, IRPP Study No. 15. Institute for Research and Public Policy,
Montreal, Quebec, February.
FLOOD, L., JANSSON, F., PETTERSSON, T., PETTERSSON, T., SUNDBERG,O.andWESTERBERG,A.
(2012) SESIM III - A Swedish dynamic micro simulation model.
GIBSON, M.J., and HOUSER, A. (2007) Valuing the Invaluable: A New Look at the Economic Value
of Family Caregiving. Issue Brief (Public Policy Institute (American Association of Retired
Persons)), no. IB82, 1–12.
GOURINCHAS,P.-O.andPARKER, J. (2002) Consumption over the life cycle. Econometrica,70(1),
47–89.
Government Accountability Ofce (GAO) (2001) Social security: Program’s role in helping ensure
income adequacy, Report No. GAO-02-62, Washington, D.C., November 2001.
GUAY,R.andJEAN, L.A. (2013) Long-term returns: A reality check for pension funds and retirement
savings, C.D. Howe Institute Commentary No. 395 Pension Policy. December, Toronto.
GUSTMAN,A.andSTEINMEIER, T. (1998) Effects of pensions on savings: Analysis with data from
the health and retirement study, Working Paper No. 6681, Cambridge, MA, August 1998.
HARDING, A. (2007) The Australian dynamic population and policy microsimulation model. In
IMA Conference, Vienna, Austria.
HAVE MA N, R., HOLDEN, K., WOLFE,B.,andSHERLUND, S. (2006) Do newly retired workers in
the United States have sufcient resources to maintain well-being? Economic Inquiry,44(2),
249–64. doi:10.1093/ei/cbj023.
HOLMER, M. (2009) Projecting Future U.S. Pension Benets. Washington, D.C.: Policy Simulation
Group.
Human Mortality Database (2013) University of California, Berkeley (USA), and Max
Planck Institute for Demographic Research (Germany). Available at www.mortality.org or
www.humanmortality.de (data downloaded on [06/01/2013]).
JAMES, E., VITTAS,D.,andSONG, X. (2001) Annuity markets around the world: Money’s worth and
risk intermediation, CERP Working Paper 16/01, Center for Research on Pensions and Welfare
Policies, Moncalieri, Italy.
KEEFE, J., L ´
EGAR ´
E, J., CHARBONNEAU,P.andD
´
ECARIE, Y. (2012) Intergenerational support to
older Canadians by their adult children: Implications for the future. In The Family, the Market
or the State? Intergenerational Support Under Pressure in Ageing Societies. International Stud-
ies in Population, Vol. 10 (ed. G. De Santis), pp. 141–158. Berlin: Springer Publisher. ISBN
97-94-007-4338.
KING,M.andLOW, D. (2014) Measuring the ‘World’ real interest rate, NBER Working Paper
19887. Cambridge, MA.
KLEVMARKEN, N.A. (1997) Behavioral Modeling in Micro Simulation Models. A Survey. Work-
ing Paper. Uppsala - Working Paper Series. http://econpapers.repec.org/paper/fthuppaal/1997-
31.htm.
KOTLIKOFF, L. (2006) Is conventional nancial planning good for your nancial health, Boston Uni-
versity working paper 2006.
LAROCHELLE-COTE, S., MYLES,J.andPICOT, G. (2008) Income security and stability during re-
tirement in Canada, Analytical Studies Branch Research Paper Series. Cat. No 11F0019M,
Ottawa.
LI, J. and O’DONOGHUE, C. (2013) A survey of dynamic microsimulation models: Uses, model
structure and methodology. International Journal of Microsimulation,6(2), 3–55.
LIU, H., OSTROVSKY,Y.,andZHOU, J. (2013) Statistics Canada, and Economic Anal-
ysis Division. Saving and Wealth: The Adequacy of Household Saving in Canada,
http://www.deslibris.ca/ID/238491.
available at http:/www.cambridge.org/core/terms. http://dx.doi.org/10.1017/asb.2016.20
Downloaded from http:/www.cambridge.org/core. IP address: 142.167.14.111, on 01 Oct 2016 at 18:18:06, subject to the Cambridge Core terms of use,
THE LIVING STANDARDS REPLACEMENT RATE (LSRR) 663
MACDONALD,B.-J.,ANDREWS,D.andBROW N, R. (2010) The Canadian elder standard - pricing
the cost of basic needs for the Canadian elderly. Canadian Journal on Aging/La Revue canadi-
enne du vieillissement 29(1): 39–56.
MACDONALD,B.-J.,MOORE, K., CHEN,H.andBROW N, R. (2011) Canadian national retirement
risk index: employing statistics Canada’s LifePaths to measure the nancial security of future
Canadian seniors. Journal of Canadian Public Policy,37(1), 73–94.
MACDONALD, B.J. and OSBERG, L. (2014) Canadian retirement incomes: How much do nancial
market returns matter?. Canadian Public Policy,40(4), 315–335.
MACDONALD, D. (2015) The feeling’s not mutual. the high costs of Canada’s mutual fund based
retirement system. Canadian Centre for Policy Alternatives March 2015.
MACKENZIE, G.A. (2010) The Decline of the Traditional Pension: A Comparative Study of Threats
to Retirement Security. Cambridge, New York: Cambridge University Press.
MCGILL, D.M. and WHARTON SCHOOL,EDS. (2010) Fundamentals of Private Pensions.NewYork:
Oxford University Press.
MILEVSKY, M.A. (2012) The 7 Most Important Equations for Your Retirement. Mississauga: John
Wiley & Sons,.
MILEVSKY, M.A. and SHAO, L. (2010) Annuities and their derivatives: The recent Canadian ex-
perience. Pension Research Council Working Paper. Pension Research Council, University of
Pennsylvania. Philadelphia. Available at: http://ssrn.com/abstract=1706893.
MILEVSKY, M.A. and YOUNG, V.R. (2007) Annuitization and asset allocation. Journal of Economic
Dynamics and Control,31(9), 3138–3177.
MILLS, M.S. and YOUNG, M.L. (2004) Facing the new facts of retirement income adequacy. Ben-
ets Quarterly,20(4), 40–47.
MITCHELL,O.andMOORE, J. (1998) Can americans afford to retire? New evidence on retirement
saving adequacy. Journal of Risk and Insurance,65(3), 371–400.
MODIGLIANI,F.andBRUMBERG, R. (1954) Utility analysis and the consumption function: An
interpretation of cross-section data. In Post-Keynesian Economics(ed. K. Kurihara), New
Brunswick: Rutgers University Press. Pp 388–436.
MOORE, J.F. and MITCHELL, O.S. (2000) Projected retirement wealth and savings adequacy in
the health and retirement study, In Forecasting Retirement Needs and Retirement Wealth (eds.
S.M. Olivia, P. Brett Hammond and A.M. Rappaport), pp. 68–94. Philadelphia, PA: Pension
Research Council and the University of Pennsylvania Press.
MOORE, K., ROBSON,W.andLAURIN, A. (2010) Canada’s looming retirement challenge: Will fu-
ture retirees be able to maintain their living standards upon retirement?, C.D. Howe Institute
Commentary No. 317.
MORISSETTE, R., ZHANG,X.andFRENETTE, M. (2007) Earnings losses of displaced workers: Cana-
dian evidence from a large administrative database on rm closures and mass layoffs, Analytical
Studies Branch Research Paper No. 291. Catalogue No. 11F0019MIE, Ottawa.
MORRISON, R.J. (2000) Assessing the quality of DYNACAN’s synthetically-generated earnings
histories. In Paper presented at the 6th Nordic Conference on Microsimulation, June, 2000,
Copenhagen, Denmark.
MUNNELL, A. (2005.May) Do people save enough for retirement? TIAA-CREF Trends and Issues.
MUNNELL, A., GOLUB-SASS,F.andWEBB, A. (2011) How much to save for a secure retirement,
Center for Retirement Research at Boston College Issue Brief Number 11-13. August.
MUNNELL, A., ORLOVA,N.andWEBB, A. (2013) How important is asset allocation to nancial
security in retirement? In The Market for Retirement Financial Advice (eds. O.S. Mitchell and
K. Smetters), pp. 89–106. Oxford: Oxford University Press.
MUNNELL,A.andSOTO, M. (2005) What Replacement Rates Do Households Actually Experience
in Retirement? Boston: Center for Retirement Research.
MUNNELL, A., WEBB,A.andDELORME, L. (2006) A New National Retirement Risk Index.Chest-
nut Hill, MA: Center for Retirement Research at Boston College.
MUNNELL, A., WEBB,A.andGOLUB-SASS, F. (2007a)What moves the national retirement risk
index? A look back and an update, Center for Retirement Research at Boston College Issue
Brief Number 7–1, January.
MUNNELL, A., WEBB,A.andGOLUB-SASS, F. (2007b) Is there really a retirement savings crisis? An
NRRI analysis, Center for Retirement Research at Boston College Issue Brief Number 7-11,
August.
available at http:/www.cambridge.org/core/terms. http://dx.doi.org/10.1017/asb.2016.20
Downloaded from http:/www.cambridge.org/core. IP address: 142.167.14.111, on 01 Oct 2016 at 18:18:06, subject to the Cambridge Core terms of use,
664 B.-J. MACDONALD, L. OSBERG AND K.D. MOORE
O’HARRA, J., SABELHAUS,J.E.andSIMPSON, M. (2004) Overview of the congressional budget
ofce long-term (CBOLT) policy simulation model. Congressional Budget Ofce.
ORGANIZATION FOR ECONOMIC CO-OPERATION AND DEVELOPMENT, OECD. (2009) Private
Pensions and the Financial Crisis: How to Ensure Adequate Retirement Income from Dened
Contribution Pension Plans? Working Party on Private Pensions.
OSTROVSKY,Y.andSCHELLENBERG, G. (2009) Pension coverage, retirement status, and earnings
replacement rates among a cohort of Canadian seniors, Statistics Canada Analytical Studies
Branch Research Paper Series, Ottawa, December 2009.
PALMER, B. (1988) The Impact of Tax Reform on Wage Replacement Ratios. Atlanta, GA: Center
for Risk Management and Insurance Research at Georgia State University.
PALMER, B. (2008) 2008 GSU/Aon RETIRE Project Report. Atlanta, GA: Center for Risk Man-
agement and Insurance Research at Georgia State University.
PICKARD, L. (2008) Informal care for older people provided by their adult children: Projections of
supply and demand to 2041 in England, Report to the Strategy Unit (Cabinet Ofce) and the
Department of Health, Discussion Paper 2515, London.
PICKARD, L. (2011) The supply of informal care in Europe, ENEPRI Research Report No. 94,
CEPS, Brussels.
QUINN, J.P. (1999) Retirement patterns and bridge jobs in the 1990’s, Employee Benet Research
Institute EBRI Issue Brief 206, Washington, DC.
REINHART, C.M. and ROGOFF, K. (2009) This Time is Different: Eight Centuries of Financial Folly.
Princeton, NJ: Princeton University Press.
RUCKMAN, K. (2003) Expense ratios of North American mutual funds. Canadian Journal of Eco-
nomics,36(1), 192–223.
RUSSELL, L.H., BRUCE, E.A., and CONAHAN, J. (2006) A methodology to determine economic
security for elders. Washington, DC: Gerontology Institute, University of Massachusetts
Boston, and Wider Opportunities for Women.
SCHIEBER, S.J. (2004) Retirement income adequacy: Good news or bad? Benets Quarterly,20(4),
27–39.
SCHOLZ,J.andSESHADRI, A. (2009) What replacement rates should households use?, Michigan
Retirement Research Center Working Paper WP 2009-2014, Wisconsin.
SCHWARTZ, D. (2010) Will a retirement boom start in 2011?. CBC News. December 31, 2010.
Available at: http://www.cbc.ca/news/canada/story/2010/12/30/f-boomers-retire.html
SKINNER, J. (2007) Are you sure you’re saving enough for retirement? Journal of Economic Per-
spectives,21(3), 59–80.
SMITH, J. (2003) Trends and projections in income replacement during retirement. Journal of Labor
Economics,21(4), 755–781.
Social Security Administration (2008) The future of social security. Washington, DC: SSA (SSA
Publication No. 05-10055, ICN 462560).
Social Security Administration (2012a) Social security programs throughout the world: Asia.Wash-
ington, DC: SSA, Ofce of Policy (SSA Publication No. 13-11802).
Social Security Administration (2012b) Social security programs throughout the world: Europe.
Washington, DC: SSA, Ofce of Policy (SSA Publication No. 13-11801).
Social Security Administration (2013a) Social security programs throughout the world: Africa.
Washington, DC: SSA, Ofce of Policy (SSA Publication No. 13-11803).
Social Security Administration (2013b) Social security programs throughout the world: Americas.
Washington, DC: SSA, Ofce of Policy (SSA Publication No. 13-11804).
SPIELAUER,MARTIN with contributions from Chantal Hicks, Steve Gribble, Geoff Rowe, Xiaofen
Lin, Kevin Moore, Laurie Plager, and Huan Nguyen (2013) The LifePaths Microsimulation
Model: An Overview.
Statistics Canada Demography Division (2010) Population projections for Canada, provinces and
territories 2009–2036, Cat. No. 91-520-X, Ottawa.
Statistics Canada (2006) The wealth of Canadians: An overview of the results of the survey of nancial
security 2005, Pension and Wealth Research Paper Series (Cat. No. 13F0026MIE-No. 001),
Ottawa, 2006.
Statistics Canada (2011) Census. Living Arrangements of Young Adults Aged 20 to 29.2011 Census
in Brief. Ottawa: Statistics Canada, 2012 (Cat. No. 98-312-X2011003).
available at http:/www.cambridge.org/core/terms. http://dx.doi.org/10.1017/asb.2016.20
Downloaded from http:/www.cambridge.org/core. IP address: 142.167.14.111, on 01 Oct 2016 at 18:18:06, subject to the Cambridge Core terms of use,
THE LIVING STANDARDS REPLACEMENT RATE (LSRR) 665
STEINBERG,A.andLUCAS, L. (2004) Shifting responsibility: The future of retirement adequacy
in America. Benets Quarterly: 15–26. Adapted from: Hewitt Associates. (2004) Total Retire-
ment Income at Large Companies: The Real Deal.
STEUERLE, C., SPIRO,C.andCARASSO, A. (2000) Do analysts use atypical workers to evaluate
social security? In Straight Talk on Social Security and Retirement Policy, (19). Washington
DC: The Urban Institute.
TD Economics (2010) Retirement income security reform: Rush prudently, don’t run blindly. Spe-
cial Report. Available at: http://www.td.com/economics/index.jsp.
THE CANBERRA GROUP (2001) Expert group on household income statistics: Final report and
recommendations. Ottawa, Canada.
U.K. Government Department for Work and Pensions (DWP) (2013) Framework for the analysis
of future pension incomes. Ref: ISBN 978-1-78153-799-2
VANDERHEI, J. (2004) Measuring retirement income adequacy, traditional replacement ratios, and
results for workers at large companies. EBRI Notes, no. 9 (Employee Benet Research Insti-
tute), September 2004. Washington DC.
VANDERHEI, J. (2006) Measuring Retirement Income Adequacy: Calculating Realistic Income Re-
placement Rates. Washington DC: Employee Benet Research Institute.
VENTI,S.F.andWISE, D.A. (2001) Aging and housing equity. In Innovations for Financing Retire-
ment. Pension Research Council Publications (eds. O.S. Mitchell, Z. Bodie, P. Brett Hammond
and S. Zeldes) pp. 254–81. Philadelphia: University of Pennsylvania Press.
VENTI, S.F. and WISE, D.A. (2004) Aging and housing equity: Another look. In Perspectives on the
Economics of Aging (ed. David A. Wise), pp. 127–75. Chicago: University of Chicago Press for
NBER. http://dx.doi.org/10.7208/chicago/9780226903286.003.0004.
VINCENT, G.K. and VELKOFF, V.A. (2012) US Census Bureau. “The next four decades the
older population in the United States: 2010 to 2050. Population estimates and projections
P25-1138.” Washington, DC: US Dept. of Commerce, Economics and Statistics Administra-
tion, US Census Bureau; 2010.
WOLFF, E.N. (2011) Pensions in the 2000s: The lost decade?, NBER Working Paper No. 16991.
Cambridge, MA.
WOLFSON, M. (2011) Projecting the adequacy of Canadians’ retirement incomes: Current
prospects and possible reform options, Institute for Research on Public Policy IRPP Study
17, Montreal.
BONNIE-JEANNE MACDONALD (Corresponding author)
School of Mathematical and Computational Sciences
University of Prince Edward Island
Halifax, Nova Scotia, Canada
E-Mail: Bonniejeanne macdonald@hotmail.com
LARS OSBERG
McCulloch Professor
Department of Economics
Dalhousie University, Halifax,
Nova Scotia, Canada
E-Mail: lars.Osberg@Dal.Ca
KEVIN D. MOORE
Principal Researcher
Statistics Canada
Ottawa, Ontario, Canada
E-Mail: kevin.Moore@statcan.gc.ca
available at http:/www.cambridge.org/core/terms. http://dx.doi.org/10.1017/asb.2016.20
Downloaded from http:/www.cambridge.org/core. IP address: 142.167.14.111, on 01 Oct 2016 at 18:18:06, subject to the Cambridge Core terms of use,
666 B.-J. MACDONALD, L. OSBERG AND K.D. MOORE
APPENDIX A
Accurate measures of retirement income adequacy are probably more important today than
ever before (in Canada and elsewhere), owing to
1. the low interest rate environment: Reinhart and Rogoff (2009), King and Low (2014),
Guay and Jean (2013) all suggested a long-term low interest rate environment for
Canada, and similar situations exist across industrialized countries. A continuing low
interest rate environment reduces the accumulation of personal savings, heightens the
cost of annuitization, and drives sponsors of underfunded DB pension plan sponsors to
freeze and/or close pension plans (see MacDonald and Osberg , 2014 for an examination
of the impact of a long-term low interest rate environment on the nancial well-being of
Canadian seniors),
2. high investment fees: In Canada, for example, the average Canadian mutual fund
charged 2.1% in fees in 2014, while comparable average pension plan fees were 0.38%
(MacDonald, 2015). Ruckman (2003) also found similarly high expense ratios for mu-
tual funds, particularly in Canada,
3. longer life expectancies: From 1950 to 2009, the average life expectancy of a 65 year-
old grew from 15.0 to 21.7 years for female Canadians and 13.3 to 18.6 years for male
Canadians (Human Mortality Database, 2013),
4. rising divorce rates among seniors (with likely negative nancial implications): Divorce
rates more than doubled between 1990 and 2010 from 4.87% to 10.05% for Americans
aged 50 and above (while U.S. population-wide divorce rates have declined) (Brown and
Lin, 2012),
5. an aging population: A rapidly rising number of seniors accompanied by a slowly
increasing labor force have raised concern regarding the capacity of governments to
deliver promised benets. For example, the ratio of Canadians aged 65 and over to
those ages 20–64 in 2010 was 22% — growing to 41% by 2030 (Statistics Canada,
2010),
6. an increasing reliance on paid services for the potentially costly expenses associated with
chronic health conditions: People aged 85 and above are most affected by chronic health
conditions and their share of the population is rapidly growing (in 2010, there were
three American seniors aged 85 and above for every hundred working-aged Americans
— this is projected to grow to eight by 2050 (Vincent and Velkoff, 2012)). While the
bulk of caregivers have historically been family and friends in most countries, this is
expected to decline owing to fewer children, greater mobility of family members, greater
workforce participation of women and changing expectations of care within families
(among many, see Gibson and Houser, 2007; Pickhard, 2008, 2011; Keefe et al., 2012)
and
7. less secure sources of retirement income: This uncertainty has arisen from
a. the shift among the design of employer pension plans from dened benet (DB)
(where the risk of the benet payments is the responsibility of the employer) to de-
ned contribution (DC) (where the risk lies with the individual) in many countries
around the world (MacKenzie, 2010),
b. similar reforms of state pension systems in many countries from DB designs to
DC individual accounts (countries with mandatory individual retirement schemes
include Chile, Uruguay, Peru, Nicaragua, Mexico, Ecuador, Dominican Republic,
Chile, Columbia, Costa Rica, Bolivia, Bulgaria, Croatia, Estonia, Latvia, Poland,
available at http:/www.cambridge.org/core/terms. http://dx.doi.org/10.1017/asb.2016.20
Downloaded from http:/www.cambridge.org/core. IP address: 142.167.14.111, on 01 Oct 2016 at 18:18:06, subject to the Cambridge Core terms of use,
THE LIVING STANDARDS REPLACEMENT RATE (LSRR) 667
Romania, Russia, San Marino, Slovak Republic, Sweden, Nigeria, Brunei, China,
Kazakhstan, Kyrgyzstan and Uzbekistan (USA Social Security Administration,
2012a, 2012b, 2013a, 2013b)), as well as reductions in state pension benets (such as
increases in normal retirement ages, which removes guaranteed benets during the
delayed period — for example, the prominent proposal to delay U.S. Social Security
benets to ensure the solvency of the program),
c. the 2008 nancial crisis and its immediate impact of reducing personal retirement
savings (Wolff, 2011).
APPENDIX B
Rather than use a single target rate, some studies estimate a range of target gross replacement
rates using survey data. This included Palmer (1988, 2008), which produced target gross re-
placement rates for workers classied by earnings level, region and marital status, using data
from the U.S. Consumer Expenditure Survey. By matching workers and retirees with similar
incomes in the survey year, its aim was to calculate, within each subgroup, the level of gross
retirement income necessary to maintain a worker’s available income (disposable income less
net savings). Specically, for each subgroup of workers, the goal was to use the survey data to
solve for annual retirement gross income and taxes (retirement gross income and taxes were
solved simultaneously since one affects the other) in
working gross income taxes workers net savings
=retirement gross income retirement taxes retirement net savings,
where
workers net savings were calculated averages for the working members of the subgroup,
retirement net savings were also calculated averages for retired members of the subgroup
with similar income levels as the workers.
With the “retirement gross income” calculated, he estimated the target gross replacement rate
for the subgroup as:
=retirement gross income/working gross income.
Palmer further produced a second set of replacement rate targets, which incorporated the
impact of period-specic expenses that do not exist in both pre- and post-retirement, such as
those associated with employment and senior aging.
The methodology underlying Palmer’s empirical measure of target gross replacement
rates was rst developed in Dexter (1984). For an illustration and additional explanation
of the conceptual model of building target gross replacement rates, see McGill et al. (2010,
Chap. 7).
available at http:/www.cambridge.org/core/terms. http://dx.doi.org/10.1017/asb.2016.20
Downloaded from http:/www.cambridge.org/core. IP address: 142.167.14.111, on 01 Oct 2016 at 18:18:06, subject to the Cambridge Core terms of use,
668 B.-J. MACDONALD, L. OSBERG AND K.D. MOORE
Mitchell and Moore (2000) and Munnell et al. (2006) took conceptually similar ap-
proaches to estimate target gross replacement rates, but they further evaluated whether cur-
rent workers (future retirees) appear to be on track to achieving them.
APPENDIX C: METHODOLOGICAL
ISSUES OF THE CONVENTIONAL
EARNINGS REPLACEMENT RATE
This section discusses the methodological issues associated with the conventional earnings
replacement rate listed in Section 2, and gives some examples of the various approaches em-
ployed by analysts.
C.1. Household-level differences in consumption due to family size
The conventionally measured earnings replacement rate is done at the individual level.
Economists have long recognized, however, that most individuals live in households and
share consumption with others, implying that household or family income should be used
when determining living standards. In the context of retirement income adequacy, this also
implies that parents need less income in retirement to maintain their pre-retirement standard
of living, relative to childless individuals with similar pre-retirement income, since a large
fraction of a parent’s pre-retirement budget is devoted to supporting children.
A widely accepted method of accounting for family size and economies of scale is to
assign each person in the household an “equivalent income”. The Luxembourg Income Study
(LIS) equivalence scale sets this amount of income equal to total household income divided
by the square root of household size, and is the methodology employed in this study.
The replacement rate literature has been diverse in terms of unit of analysis. Some replace-
ment rate studies have used individuals (VanDerhei, 2006; Ostrovsky and Schellenberg, 2009),
while others include spouses (Mitchell and Moore, 1998; Munnell et al., 2006). Some studies
have used individuals as the unit of analysis, but calculate per-capita income at the level of
couples (Butrica et al., 2006) (for example, if the consumption of one spouse is $100,000 and
the other zero, each spouse would be assumed to have consumed $50,000). Studies that use
longitudinal data have to contend with the fact that marital status changes through time.
For example, it is not obvious whether a man who marries in the year leading up to his
retirement should be considered married or single for the purposes of computing his replace-
ment rate. Smith (2003) discussed these issues (including transitions to widowhood) between
pre-retirement and post-retirement, and adjusted replacement rates accordingly by using an
equivalency scale.
While analysts often incorporate marital status, children are routinely not considered
(for discussion, see Scholz and Seshadri, 2009: 3). Examples of replacement rate studies that
construct replacement rate measures where pre-retirement consumption is explicitly reduced
to reect the costs of supporting dependent children include Larochelle-Cote et al. (2008),
Scholz and Seshardri (2009), Moore et al. (2010) and MacDonald et al. (2011).
available at http:/www.cambridge.org/core/terms. http://dx.doi.org/10.1017/asb.2016.20
Downloaded from http:/www.cambridge.org/core. IP address: 142.167.14.111, on 01 Oct 2016 at 18:18:06, subject to the Cambridge Core terms of use,
THE LIVING STANDARDS REPLACEMENT RATE (LSRR) 669
C.2. Components of income
Early literature tended to restrict the included sources to those found in income data (Boskin
and Shoven, 1987; Palmer, 1988), and this approach continues in studies that rely exclusively
on this data source (Smith, 2003; Larochelle-Cote et al., 2008; Ostrovsky and Schellenberg,
2009). Although some sources of income are adequately captured in traditional income data
or administrative data, others are not. For example, the drawdown (or dissavings) of non-
registered assets, which includes non-tax-sheltered savings held in bank accounts, bonds,
mutual funds, stock markets and certain annuities, is not captured in traditional income data
sources; only the annual yield on such capital is “income”. For savings in the form of home
ownership, neither the imputed rent, mortgage payments, nor the consumption represented
byitsdrawdownarevisibleinsuchdata.
Instead of using earnings to solely determine the replacement rate denominator, Munnell
and Soto (2005) and Munnell et al. (2006) included pre-retirement investment income in the
replacement rate denominator. Smith (2003), LaRochelle-Cote et al. (2008) and MacDonald
et al. (2011) incorporated investment income as well as government transfers (such as social
assistance, EI and child benets).
“Imputed rent” has been considered by Munnell and Soto, 2005, Munnell et al., 2006,
Moore et al., 2010, MacDonald et al., 2011 and Wolfson, 2011. Imputed rent can be regarded
as a form of investment income — the return from savings that are held in the form of real
property, rather than in nancial assets. Including imputed rent in the replacement rate rec-
ognizes that homeowners who have some equity in their homes are enjoying a ow of services
that is an important source of household consumption (Brown et al., 2010). The importance
of imputed rent continues to apply in retirement, particularly given that the cost of shelter
is the largest expenditure from among a healthy senior’s basic costs of living (MacDonald
et al., 2010) and that seniors are more likely to own their home than any other age group (in
Canada, see Statistics Canada, 2006).
With regards to pre-retirement savings, Palmer (1988) developed target gross replacement
rates using a measure that reasonably captured most forms of savings, but appeared to omit
the savings represented by paying down one’s mortgage. The target gross replacement rates
developed by Mitchell and Moore (1998) and Brady (2010), on the other hand, included the
reduction of mortgage principal in pre-retirement savings. Using micro-simulation model-
ing, Moore et al. (2010) and Wolfson (2011) incorporated pre-retirement savings in the form
of employer-sponsored pension plans, registered retirement savings plans and paying down
mortgages. MacDonald et al. (2011) took a somewhat more comprehensive approach by also
measuring non-registered wealth and debt accumulation (including nancial assets/debts,
real-estate investments and equity in businesses owned).
Personal wealth includes equity in an owner-occupied home, which therefore could be
seen as a component of retirement savings. Many argue that homes are a special case, however,
and that an assessment of the replacement adequacy of the retirement income system should
not assume that retirees sell their homes or otherwise deplete their home equity. Excluding
this asset class, which for most Canadians is the single largest net asset (Statistics Canada,
2006), clearly distorts important differences in nancial security between homeowners and
renters, and between homeowners with very different amounts of home equity. Gustman
and Steinmeier (1998), Moore and Mitchell (2000), Munnell et al. (2006) and Munnell et al.
(2007a, 2007b) explicitly included the drawdown of all housing wealth in addition to nancial
assets. Moore et al. (2010) included both imputed rent and the drawdown of registered assets,
and also explored the impact of drawing down different proportions of home equity. Engen
et al. (1999) included half of home equity (this half-way point was supported by Munnell,
available at http:/www.cambridge.org/core/terms. http://dx.doi.org/10.1017/asb.2016.20
Downloaded from http:/www.cambridge.org/core. IP address: 142.167.14.111, on 01 Oct 2016 at 18:18:06, subject to the Cambridge Core terms of use,
670 B.-J. MACDONALD, L. OSBERG AND K.D. MOORE
2005). MacDonald et al. (2011) included imputed rent and the drawdown of registered and
non-registered wealth, but did not include home equity, which was assumed to not be con-
sumedinretirement.
Replacement rate studies are divided on the issue of employment earnings as a component
of retirement income. In most of the literature, earnings are expressly excluded on conceptual
grounds that the purpose of a “replacement rate” is to evaluate the extent that employment
earnings have been replaced after retirement (where retirement is the cessation of employ-
ment) (Mitchell and Moore, 1998; Munnell et al., 2006; Biggs and Springstead, 2008; Moore
et al., 2010; MacDonald et al., 2011). In other studies, earnings are either purposely included
in the numerator, or are included in a broader “total income” variable (Boskin and Shoven,
1987; Smith, 2003; Larochelle-Cote et al., 2008).
Income taxes should be subtracted from gross pre- and post-retirement income. This is
done directly for each individual if individualized targets are being employed (such as in Mac-
Donald et al., 2011). Otherwise, when averages are used to determine target gross replacement
rates (such as explained in Appendix B), differentials in income taxes pre- and post-retirement
are incorporated into the target gross replacement rates.
C.3. The volatility of earnings and retirement income
One of the most central methodological issues in designing a replacement rate measure is
the choice of measurement period for the denominator. The measurement period used to
calculate pre-retirement earnings can have an enormous impact on the resulting replacement
rate, leading to quite different conclusions about replacement adequacy.
The “replacement rate” goal is the preservation of pre-retirement living standards, but
this leads to the question: pre-retirement living standards measured over what period? Much
of the consumption literature nds that the average shape of consumption over an individual’s
life has a distinct “hump” shape (Gourinchas and Parker, 2002; Fernandez-Villaverde and
Krueger, 2007). On average, annual consumption early in an individual’s career closely tracks
earnings and typically rises at a relatively rapid pace. After an individual reaches their “prime”
working years, however, annual consumption increases much more gradually or even levels
off completely. Consumption tends to peak in late middle-age, and starts declining gradually
thereafter, even as annual earnings remain at or continue to increase.
The most important complications arise, however, because replacement rate denomina-
tors are typically calculated using gross earnings and not income for consumption expendi-
ture (see Equation (1)). Annual earnings have considerable year-over-year variation (Mor-
rison, 2000), which creates complex and diverse earnings histories across individuals. Many
individuals exhibit tremendous “earnings mobility”,15 and consequently occupy quite differ-
ent rankings in the distribution of earnings over the course of their careers (Finnie, 1999;
Beach and Finnie, 2004). Overall, it cannot be assumed that a narrow measurement period
is “representative” of earnings since it will reect different things for different individuals.
Further, much of the theoretical literature relating to life-cycle income and consumption
suggests that the relationship between an individual’s standard of living and his/her earnings
measured over any particular short-term period may be quite weak (a theory that began with
Modigliani and Brumberg, 1954). Diminishing marginal utility implies that individuals prefer
a relatively consistent standard of living, so they smooth their consumption over long time
horizons, such as their entire lifetimes, rather than making current consumption decisions
based primarily on current income (ibid).
In the applied replacement rate literature, there is tremendous diversity in the measure-
ment periods used to calculate the pre-retirement denominator. As we discuss below, this
available at http:/www.cambridge.org/core/terms. http://dx.doi.org/10.1017/asb.2016.20
Downloaded from http:/www.cambridge.org/core. IP address: 142.167.14.111, on 01 Oct 2016 at 18:18:06, subject to the Cambridge Core terms of use,
THE LIVING STANDARDS REPLACEMENT RATE (LSRR) 671
diversity applies both to the part of the life-course chosen, and the number of years averaged
in the calculation.
Some studies use only one year of data. For example, Palmer’s work (1988, 2008) relied on
a single cross-sectional year of data with individuals between ages 50 and 64, whom he aver-
aged across to produce the denominator of his replacement rate measure. Mitchell and Moore
(1998) essentially used nal earnings, but the denominator was calculated deterministically
by taking a single year of data (an individual’s earnings in 1992) and assuming constant real
wage growth until retirement. Munnell et al. (2006) similarly began with data on a house-
hold’s pre-retirement income in a single year, 2003, and projected it forward to retirement
age according to an average earnings prole. Using only one year of data, and projecting any
remaining years from this one observation in a highly stylized manner, ignores the substantial
variation in many individuals’ earnings.
Brady (2010) calculated average career earnings from ages 30 to 67, but relied on a handful
of illustrative individuals with stylized earnings histories.
Using longitudinal earnings data, Smith (2003), LaRochelle-Cote et al. (2008) and Os-
trovsky and Schellenberg (2009) all averaged the earnings of each sampled individual from
ages 54 to 56, which were characterized as “peak” and “permanent” earnings. Given the sub-
stantial variability in employment and earnings across many individuals’ life-courses, average
earnings for this age range will represent neither peak nor permanent earnings for a signicant
number of individuals (Finnie, 1999; Morissette et al., 2007; Finnie and Gray, 2011).
Munnell and Soto (2005) used approximately forty years of longitudinal micro-data on
individual earnings to calculate both career-average and nal average (best 5 of nal 10) mea-
sures for the denominator. Boskin and Shoven (1987) similarly used 23 years of longitudinal
earnings micro-data to calculate career average and nal average (best 3 of nal 10) measures.
The impact of the chosen measurement period had a substantial impact in both studies.
As noted in Section 3, studies that use large-scale, dynamic micro-simulation models have
nearly complete exibility in choosing the pre-retirement measurement period in the denom-
inator — such as in Butrica et al. (2006), Moore et al. (2010), MacDonald et al. (2011) and
Biggs and Springstead (2008).
Relative to earnings, retirement income sources tend to be much smoother, although there
are some exceptions, including marital transitions during the retirement years, especially to
widowhood, which can affect retirement income stability. In addition, pension income that is
not indexed to ination, such as some employer pensions or most private annuities, can fall
substantially in real value over the course of an individual’s retirement. This suggests that
a broad “retirement average” measurement period should be preferred for post-retirement
income or consumption, rather than a narrow measurement period (such as the rst year
after retirement).
Most of the literature has used a single, cross-sectional snapshot of post-retirement in-
come (Palmer, 1988; Smith, 2003; Ostrovsky and Schellenberg, 2009). Conventionally, re-
placement rate studies have tended to look specically at income in the rst year of retirement
(GAO, 2001).
Some researchers have applied adjustments to contend with the short post-retirement
measurement period of one year in conventional replacement rates. For example, assuming a
one-year measurement period in retirement, Steinberg and Lucas (2004) suggested increasing
conventional targeted replacement rates by 10 to 15 percentage points to compensate for the
lack of ination indexing in many retirement benets (the lower range would be for low-
income seniors, who rely largely on ination-indexed Social Security benets). Alford et al.
(2004) made similar adjustments.
available at http:/www.cambridge.org/core/terms. http://dx.doi.org/10.1017/asb.2016.20
Downloaded from http:/www.cambridge.org/core. IP address: 142.167.14.111, on 01 Oct 2016 at 18:18:06, subject to the Cambridge Core terms of use,
672 B.-J. MACDONALD, L. OSBERG AND K.D. MOORE
Some studies have taken snapshots of replacement rates at several different ages post-
retirement, including Biggs and Springstead (2008). Larochelle-Cote et al. (2008) stands out
in this regard, as it used longitudinal data to follow individuals and couples through their
retirement years, from ages 55 to 80, with a particular focus on the evolution of their replace-
ment rates over time and the stability of their retirement income.
C.4. Pre-and post-retirement nancial risks
Workers and seniors face different risks. The major income source risk for workers are labor
market risks such as unemployment. For seniors, it is generally investment rate risk, ination
risk, generating an expensive medical condition and living “too long” (that is, outliving their
nancial resources). Unlike workers, seniors have fewer opportunities to react to these risks
— for example, it is usually less feasible to return to the labor market if investments perform
poorly. Consequently, post-retirement risk is an important area of study.16
Typically, replacement rate measures have not explicitly accounted for important risks
of retirement that can affect a senior’s nancial well-being — namely, the risk of accelerating
ination, the death of a spouse, divorce, insurer default, low investment returns, annuitization
rates, longevity, developing a health condition that generates signicant out-of-pocket expen-
ditures, unanticipated expenses (such as replacing a roof and other home repairs), increase in
public taxes, DB pension plan default and changes in retirement benets by government and
private plan sponsors (such as reductions in pension income, retiree medical benets, pension
cost-of-living adjustments and other plan design features — see Mills and Young, 2004 for a
discussion of such changes over the past decade in U.S. employer pension plans).
Schieber (2004) wrote that “singular rules of thumb for replacement rates are naive
and that estimates should take into account the unforeseen risks that individuals face” (ab-
stract). How to incorporate these risks is unclear, however, particularly for conventional re-
placement rate measures where the typical post-retirement measurement period is the rst
year of retirement. Choosing average or median costs is problematic — for instance, an
individual either has a major health condition or does not, and therefore incorporating the
median cost of any particular illness in a replacement adequacy measure will be insufcient
for half of the population and cause the other half to unduly reduce their pre-retirement
standard of living in order to save for an event that does not occur. The VanDerhei (2006)
study offered one possible approach — he explicitly modeled the risk of catastrophic med-
ical expenses, low investment returns and longevity, and used micro-simulation to simu-
late the distribution of possible future outcomes in order to demonstrate the uncertain -
nancial impact of each contingency on retirement income adequacy targets. VanDerhei ex-
plained that the individual could choose the target that most appropriately t his or her
level of risk aversion (for example, a highly risk-averse person might choose the target re-
placement rate providing a 90% likelihood of maintaining a specied standard of living in
retirement).
Holmer (2009) provided another approach. Based on ve hundred different macroeco-
nomic projection scenarios, he calculated replacement rates from expected risk-adjusted re-
tirement income ows (rather than expected average retirement income ows). For the 1990
American birth cohort, as an example of the results, he found that “the risk-adjusted pension
benet is substantially smaller than the risk-adjusted social security benet, even though the
average (non-risk-adjusted) pension benet is roughly the same as the average social security
benet” (abstract).
available at http:/www.cambridge.org/core/terms. http://dx.doi.org/10.1017/asb.2016.20
Downloaded from http:/www.cambridge.org/core. IP address: 142.167.14.111, on 01 Oct 2016 at 18:18:06, subject to the Cambridge Core terms of use,
THE LIVING STANDARDS REPLACEMENT RATE (LSRR) 673
C.5. Purchasing power differences of income before and after retirement
How do replacement rates account for the fact that the denominator will be measured at a
different time than the numerator (the former being before retirement and the latter being
after retirement)? Most studies express the numerator and denominator in constant dollars
using the consumer price index. An alternative method is to use wage indexation, which is
the approach traditionally taken by the U.S. Social Security Administration in its calculation
of social security replacement rates, and was also used by Butrica et al. (2006) and Wolfson
(2011). Using a consumer price index compares an individual’s retirement income to the ab-
solute level of pre-retirement consumption he/she actually experienced, while using a wage
index also incorporates a comparison to the consumption of currently working generations.
For most purposes, replacement rates are used to evaluate whether retirees’ own consumption
falls after retirement, rather than to make intergenerational comparisons and, for this reason,
price indexation is most commonly used.
C.6. Individual preferences for risk aversion, leisure and bequest
As Section 2 noted, an alternative conception of retirement income adequacy employs a util-
itarian framework (such as in Engen et al., 1999; Scholz and Seshadri, 2009; Liu et al.; 2013).
Although a utilitarian framework suffers from complexity in modeling true individual pref-
erences, it is considered an ideal framework in the study of living standards as it has the
capacity to “(1) reect differences in leisure as well as all forms of potential consumption,
including home production and publicly provided goods; ...(2)account (for) differences in
constraints faced both by people living in the same country, and differences in constraints
faced by people in different countries; ...(and) (3) account for differences in the ability to
smooth income across periods” (Canberra Group, 2001: 2). Despite these advantages of the
utilitarian framework, it is commonly agreed to be too complex to implement for most users
of replacement rates.
C.7. Changes in expenses/spending over the life course
Replacement rates are concerned only with the replacement of pre-retirement consumption
expenditure in post-retirement. There can be, however, period-specic consumption that does
not need to be replaced in post-retirement and/or new post-retirement consumption that does
not occur in pre-retirement. Consequently, an individual with an unchanging standard of
living before and after retirement could in fact have different consumption levels owing to
the effects of aging and the cessation of employment.
For example, expenses associated with employment generally cease after retirement, such
as professional development fees, commuting to work costs and the expense of special cloth-
ing (Figure 1 incorporates this commonly acknowledged changing expense, although we do
not model it in our analysis). Further, retirees are better able to “stretch their dollar” owing
to senior discounts and greater time for home production, consequently a dollar of retire-
ment income could be more valuable than a dollar while working — for instance, it has been
observed in both Canada (Brzozowski and Lu, 2010) and the U.S. (Aguiar and Hurst, 2005)
that retirees spend less on food but still maintain the same quality of diet owing to more
efcient shopping and cooking more at home. Similarly, Brancati et al. (2015) and Milevsky
(2012) found that real consumption declines with age after retirement. On the other hand,
medical expenses are likely to rise with age, particularly those associated with a chronic health
condition.
available at http:/www.cambridge.org/core/terms. http://dx.doi.org/10.1017/asb.2016.20
Downloaded from http:/www.cambridge.org/core. IP address: 142.167.14.111, on 01 Oct 2016 at 18:18:06, subject to the Cambridge Core terms of use,
674 B.-J. MACDONALD, L. OSBERG AND K.D. MOORE
Should period-specic consumption be incorporated in the study of retirement income
adequacy? The cost of medical care is likely the most material since it is both a necessary cost
and a potentially signicant one depending on the public/private nature of the state health
care system and the coverage afforded by the retiree’s employer. This is particularly true at
advanced ages when the likelihood of suffering from acute and chronic health conditions is
much higher. For instance, for those who need it, the expense of home care is a great concern
since the onset of a chronic illness can be sudden and beyond the individual’s control, and
the out-of-pocket cost can become quite substantial in both the U.S. (Russells et al., 2006)
and Canada (MacDonald et al., 2010).
Analysts have the option to either integrate period-specic consumption (all components
or those that are deemed most important) or to treat them as a topic outside of a replace-
ment rate framework. For instance, new and signicant retirement expenses (notably medical)
could be investigated separately such as through a precautionary saving, or insurance type
analysis. Alternatively, period-specic consumption could be built into a replacement rate
analysis by being subtracted from the income for consumption expenditure from Figure 1.
In past replacement rate literature, the topic of period-specic consumption has been
either not addressed, or done so very loosely in a wide variety of manners. For instance,
Dexter (1984) included all components as a one-off net change in consumption requirements
at retirement when developing target replacement rates. Building on Dexter (1984), Palmer
(1988) dened this variable as:
Work-related expenses +Net Change in Age-Sensitive Expenditures.
Palmer then developed two sets of target replacement rates — those with and without age-
and work-related changes in consumption. McGill et al. (2010) took a similar approach, but
modeled these two categories of expenses using regression analysis. As explained in Appendix
C.4, however, using average or median change in necessary expenditures can be problematic,
particularly when the expense is associated with an unknown post-retirement risk.
Steinberg and Lucas (2004), Alford et al. (2004) and Schieber (2004) used a conceptu-
ally similar approach as Palmer (1988) and subtracted net retiree medical costs from post-
retirement income in the conventional gross employment earnings replacement rate calcu-
lation. VanDerhei (2006) integrated the cost of medical expenses after retirement into the
replacement rate analysis using a different approach. He explicitly modeled medical expenses
over the lifetime of the retiree to assess the adequacy of conventional gross employment earn-
ings replacement rates. As noted above, however, most replacement rate studies do not ad-
dress period-specic consumption.
APPENDIX D: LIFEPATHS –TOOL OF
ANALYSIS
Statistics Canada’s LifePaths is one of the world’s largest dynamic micro-simulation models
of society. By integrating many data sets within Statistics Canada, LifePaths builds entire syn-
thetic populations by simulating the detailed life-courses of virtual Canadians case by case.
These virtual individuals attend school, make educational choices, leave home, form fami-
lies, migrate, become parents, divorce and remarry, lose and nd jobs, earn money, acquire
homes, save, pay taxes, contribute to pension schemes, receive benets and pensions, become
available at http:/www.cambridge.org/core/terms. http://dx.doi.org/10.1017/asb.2016.20
Downloaded from http:/www.cambridge.org/core. IP address: 142.167.14.111, on 01 Oct 2016 at 18:18:06, subject to the Cambridge Core terms of use,
THE LIVING STANDARDS REPLACEMENT RATE (LSRR) 675
(
(((((
(A)
(
(B)
Age
Date of Birth
Gender
Education
Marital Status
Number of Children
Employment Status
Earnings
Province of Residence
Duration in Current State
Date
Events
Marriage, Separation, Divorce, Widowed,
and Common-Law Union Formation
Birth of a Child
High School Dropout, High School
Graduation, Post Secondary School Graduate
(30 levels and 100 fields of study)
Employed, No Employed, Self- Employed
Marital Status
Immigration, Emigration, Return to Canada,
Inter-Provincial Migration
Province of Residence
Duration in Current State
Date
FIGURE D.1: Illustration of LifePaths’ simulation of a Canadian life-course.
Source: MacDonald et al. (2010, Fig. 1) .
disabled, and eventually die. LifePaths simulations aggregate to historical data over the past
half-century and allow for detailed projections into the future (projections that incorporate
the realistic complexity and diversity both across individuals and within life-courses).
Figure (4) represents the evolution of a simulated life in LifePaths. This is a simplied
ow chart for illustration purposes, and is not intended to convey the true complexity
of LifePaths. We list only some of the components of LifePaths—marital status, fertil-
ity, education, employment, and migration. For each simulated life, LifePaths tracks
the individual’s relevant characteristics, such as those listed in the rst box. These char-
acteristics enter as explanatory variables to determine the times until the occurrence of
each possible event (arrow A). The event with the shortest wait time “wins” and, once
it occurs, the individual’s characteristics are updated (arrow B). These characteristics
then enter again as explanatory variables to determine the next event (arrow A). This
continues until death, thus creating a complete life course with all of the necessary
details for millions of simulated Canadians. (MacDonald et al., 2010: 76).
LifePaths integrates a tremendous amount of micro-data and aggregate data in its be-
havioral equations and other modeling elements to faithfully reproduce the historical socioe-
conomic experience of Canadians. The model simulates detailed and diverse individual life
courses using a variety of statistical methods, with particular emphasis on statistical event-
history equations estimated from a broad array of data sources. Micro-data with a longitu-
dinal component are taken advantage of wherever possible. Key data sources that are used
to develop the model are historical demographic estimates of population, immigration, emi-
gration, fertility, mortality, census micro-data from 1971 to 2006, longitudinal Labour Force
Survey (LFS) data from 1976 to 2005, Family History surveys from 1984 to 2001, admin-
istrative data on post-secondary education, the 1999 and 2005 Survey of Financial Security
(SFS), and longitudinal and/or cross-sectional taxation data from 1980 to 2011. Many other
data sources have also been used to a lesser degree. Other than the exceptions noted below,
the scenario underlying the main results of this report is the “default” LifePaths scenario,
which generally assumes the continuation of recently observed socio-economic outcomes.
LifePaths is publicly available and has been under development for over two decades.
An overview of LifePaths can be found at Statistics Canada Modelling Division (Spielauer,
available at http:/www.cambridge.org/core/terms. http://dx.doi.org/10.1017/asb.2016.20
Downloaded from http:/www.cambridge.org/core. IP address: 142.167.14.111, on 01 Oct 2016 at 18:18:06, subject to the Cambridge Core terms of use,
676 B.-J. MACDONALD, L. OSBERG AND K.D. MOORE
2013), which is publicly available for the interested reader and can be found on the Statistics
Canada website: http://www.statcan.gc.ca/microsimulation/lifepaths/lifepaths-eng.htm.
Li and O’Donoghue (2013) provides an international context for LifePaths through com-
parison with other dynamic micro-simulation models. LifePaths is particularly notable in its
use of continuous rather than discrete time, and its use of an open, rather than closed, pop-
ulation methodology.
We carried out our analysis by building on LifePaths Model version 5.1.4.4. The assump-
tions and calculations underlying the simulation results were prepared by the authors and the
responsibility for the use and interpretation of these data is entirely that of the authors. In
addition to working code that collects and calculates our measures of interest, we developed
and integrated into LifePaths an improved annuity price calculator that realistically incor-
porates the relevant personal and nancial market inputs akin to actual Canadian annuity
providers. We also updated the nancial market model up to the end of 2012.
available at http:/www.cambridge.org/core/terms. http://dx.doi.org/10.1017/asb.2016.20
Downloaded from http:/www.cambridge.org/core. IP address: 142.167.14.111, on 01 Oct 2016 at 18:18:06, subject to the Cambridge Core terms of use,
... As argued above, old age insecurity is a risk that can only be realized in the long-run, so the projected replacement rate might vary significantly from the real replacement rate 30 years from now, and it might still not be enough for maintaining a decent old-age life. Studies have continued to argue over what parameters to use and what would be the optimal target replacement rate (e.g., Vanderhei 2006;Scholz and Seshadri 2009;Macdonald et al. 2016). Second, and more importantly, even if the actual risk environment remains constant across time, individuals may still experience and interpret the investment risks differently. ...
... Following the 70% ideal replacement rate suggested by the literature, I divide the respondents into the secure/insecure groups. Because the survey item does not allow me to identify the exact 70% and because the scholarship has started to question the adequacy of the 70% replacement rate (e.g., Macdonald et al. 2016), I picked 75%, a more conservative replacement rate, as the cutoff point. Individuals whose perceived replacement rate is above 75% are the secure group; those who perceived replacement rate below 75% are considered as the insecure group. ...
... There is a considerable portion of citizens who think their income replacement rate will be above 75% but still perceive not having enough money during retirement. This contradicting picture echoes some skeptics who argue that 70% is not an adequate replacement rate for retirement (e.g., Vanderhei 2006;Scholz and Seshadri 2009;Macdonald et al. 2016). In sum, even though it is far from perfect to claim rising retirement insecurity with only two waves of data, the trend goes upward from the available evidence. ...
... As argued above, old age insecurity is a risk that can only be realized in the long-run, so the projected replacement rate might vary significantly from the real replacement rate 30 years from now, and it might still not be enough for maintaining a decent old-age life. Studies have continued to argue over what parameters to use and what would be the optimal target replacement rate (e.g., Vanderhei 2006;Scholz and Seshadri, 2009;Macdonald 2016). Second, and more importantly, even if the actual risk environment remains constant across time, individuals may still experience and interpret the investment risks differently. ...
... This contradicting picture echoes some skeptics who argue that 70 percent is not an adequate replacement rate for retirement (e.g., Vanderhei 2006;Scholz and Seshadri, 2009;Macdonald, Osberg, and Moore 2016). ...
... Moreover, the paper supplements the subjective income replacement measure with the feeling of income sufficiency, another proxy of retirement insecurity. As the feeling of income insufficiency is only minimally correlated with the subjective replacement rates, this paper shows that the assumed 70 percent replacement rate might not lead people to feel secure about retirement, resonating with recent studies (Macdonald et al., 2016). Under a more fluctuated financial market, being in DC programs is associated with higher retirement insecurity, either proxied by the subjective income replacement rate or the feeling of income insufficiency. ...
Article
Full-text available
Why do we observe higher and higher retirement insecurity over the years? This paper argues that an individual's retirement insecurity hinges on the interplay between her registered pension scheme and the level of market volatility. A pension scheme reflects the extent to which pension investment risk is individualized, providing the seed for retirement insecurity. The fear for retirement was triggered as capital markets became more volatile over the past decade. Market volatility creates an immediate information shortcut for individuals when forming expectations about post-retirement income that will only be realized in the long run. Due to the "available information bias," citizens have developed overall downward expectations of retirement income under growing market volatility, forming pessimistic perceptions about retirement. This pessimism is more striking among those with a Defined Contribution pension scheme than those with a Defined Benefit plan because of the risk individualization feature. The argument is tested with data from the British Household Panel Survey (BHPS). The findings suggest that individuals' perceptions of insecurity in post-working life are driven by the type of pension plan in which they are enrolled and the degree of market fluctuation facing them.
... The actual replacement rate is defined as the ratio of a person's income after retirement to that before retirement [24]. The living standards replacement rate aims to capture the continuity of a person's living standards after retirement by comparing the amount of money she/ he needs to support personal consumption of goods and services before and after retirement [25]. Considering almost all the pension benefits are used to meet residents' basic living needs in China's poor regions [25], this study utilizes a modified living standards replacement rate (hereafter referred to as MLSRR) that removes the excess demand for commodities. ...
... The living standards replacement rate aims to capture the continuity of a person's living standards after retirement by comparing the amount of money she/ he needs to support personal consumption of goods and services before and after retirement [25]. Considering almost all the pension benefits are used to meet residents' basic living needs in China's poor regions [25], this study utilizes a modified living standards replacement rate (hereafter referred to as MLSRR) that removes the excess demand for commodities. By comparing the ARR and MLSRR, this study assesses the pension adequacy of China's NRPS in the poor areas. ...