Individual estimates of life expectancy and consumption
patterns

Derek R. Lawson, CFPa*, Stuart J. Heckman, Ph.D., CFPa

aDepartment of Personal Financial Planning, Kansas State University, 1324 Lovers Lane, 303 Justin Hall,
Manhattan, KS 66506, USA

Abstract

Previous research reveals a paucity of retirement preparedness in the United States. With the shift
to defined contribution plans, retirement planning now falls on the shoulders of consumers. Using the
2013 Survey of Consumer Finances (SCF), the relationship between subjective life expectancy and
consumption is investigated. Specifically, regression analyses examine the relationship between
subjective life expectancy and three indicators of consumption: financial assets, credit card debt, and
saving behavior. A secondary analysis separated respondents near retirement (i.e., within 10 years)
and far from retirement (i.e., more than 10 years) to determine if retirement saliency affects the
relationship between life expectancy and consumption. The influence of life expectancy on consump-
tion is analyzed by separating life expectancy into two periods: remaining work life and retirement
life. Results indicate that remaining work life expectancy and retirement life expectancy are associated
with financial assets and that retirement life expectancy is associated with savings behavior. © 2017
Academy of Financial Services. All rights reserved.

JEL classification: D14; E21

Keywords: Consumption; Life expectancy; Retirement; Saving

1. Introduction

One result of the shift away from defined benefit plans to defined contribution plans is that
household consumption decisions have a substantial and lasting impact on a household’s
ability to prepare for retirement. Retirement planning is a complex task for the average

* Corresponding author. Tel.: �1-515-707-3729; fax: �1-785-532-5505.
E-mail address: drlawson@ksu.edu (D.R. Lawson)

Financial Services Review 26 (2017) 1–18

1057-0810/17/$ – see front matter © 2017 Academy of Financial Services. All rights reserved.



consumer given all the uncertainty involved, for example, length of time in retirement,
inflation, rates of return, and so forth. Additionally, consumers have difficulty in making
decisions under uncertainty and are prone to many biases and mistakes (Tversky & Kahne-
man, 1974). Now that most consumers bear the investment and longevity risk associated with
retirement funding, consumption decisions throughout the life cycle have an important role
to play in shaping retirement adequacy.

The current study focuses on one particular source of uncertainty, an individual’s life
expectancy, and examines how subjective estimates of life expectancy influence household
consumption. Because of the dynamic relationship between current age, retirement age, and
life expectancy and consumption choice, the current study analyzes the influence of life
expectancy on consumption by dividing the respondents’ remaining life expectancy into two
periods—remaining work life expectancy and retirement life expectancy. From a life cycle
perspective, examining these two periods of time (i.e., remaining working period and
retirement period) highlights the tradeoff between saving and dissaving.

1.1. Literature review

Some studies find subjective life expectancy to be a valuable predictor of actual mortality
(Hamermesh, 1985; Hurd & McGarry, 2002). Elder’s (2007) findings confirm individual
estimates of life expectancy to be accurate but noted an interesting pattern in that individuals
nearing retirement had higher discount rates for their future. His results show that individuals
nearing or just starting retirement were pessimistic about living to age 75, but became
optimistic about living to age 85 once they survived to age 75. This creates a misalignment
in the savings patterns needed to smooth consumption because the initial retirement period
planned for is too short and needs to be corrected for a longer horizon.

There is some evidence to support the notion that subjective life expectancy influences
saving and consumption patterns. For example, Bloom, Canning, Moore, and Song’s (2006)
findings show that increased subjective life expectancy yields increased savings at any age.
They also look at differences between a husband’s and wife’s subjective life expectancy
estimates on household wealth and find that a 10% increase in a husband’s life expectancy
results in an increase of household wealth by $27,600, and a 10% increase in the wife’s life
expectancy results in an increase in household wealth by $32,700. Other findings show that
an increase in life expectancy is associated with an increase in savings, but most of these
findings are based on data from East Asian countries because of their increases in life
expectancies (Bloom, Canning, & Graham, 2003; Tsai, Chu, & Chung, 2000). Yet, other
research finds contradictory results indicating that rises in life expectancy mostly increase
expected retirement age when individuals are not subject to mandatory retirement ages
(Deaton & Paxson, 2000). Bloom et al. (2003) find improved health has an ambiguous effect
on savings rates because it may increase remaining work life expectancy. Because of this,
savings rates do not necessarily need to increase because compounded interest on current
savings can help cover expenditures later on (Bloom, Canning, & Moore, 2004).

Spaenjers and Spira (2015) analyze the empirical relationship between subjective life
expectancy and portfolio choice. Their findings show that life expectancy positively influ-
enced the amount allocated towards equities within portfolios. That is, after controlling for

2 D.R. Lawson, S.J. Heckman / Financial Services Review 26 (2017) 1–18



effects related to gender-specific age, health, and other socioeconomic characteristics, people
with longer life expectancies have greater amounts of equities than their counterparts. They
also compare people with the same subjective life expectancies to see if bequest motives
make a difference on portfolio allocations. They find that those without bequest motives had
a decreased equity allocation as compared to those with bequest motives, indicating that
people with bequest motives may view their life horizons as “indefinite” (Spaenjers & Spira,
2015, p. 104).

An important consideration for retirement planning and savings decisions is an individ-
ual’s planned retirement age. Montalto, Yuh, and Hanna (2000) examine the determinants of
planned retirement age before retirement. They find that as people get older their planned
retirement age also increases. Specifically, they find that as full-time workers age from 35 to
45, their expected retirement age increases by 1.08 years, while workers aging from 45 to 55,
increase their expected retirement age by 3.41 years. Finally, those workers aging from 55
to 65 increase their expected retirement date by 5.72 years. They also find that life
expectancy, non-investment income, and educational attainment are all positively associated
with planned retirement age. Financial assets (excluding IRA/KEOGH and defined contri-
bution plans), having a defined benefit plan, and low-skilled occupations are negatively
associated with expected retirement age (Montalto et al., 2000).

1.2. Summary and gap

While subjective life expectancy has been shown to influence consumer savings (Bloom
et al., 2003), there is little empirical research examining the influence on household con-
sumption behavior, particularly when using the Survey of Consumer Finances (SCF). Other
research utilizing the SCF finds subjective life expectancy influences annuity purchase
decisions (Gupta & Li, 2012) and construction of investment portfolios (Spaenjers & Spira,
2015). Therefore, the current study explores the following research question:

RQ: What effect do subjective estimates of life expectancy have on consumption decisions?

1.3. Theoretical framework

1.3.1. Life cycle hypothesis
First posited by Modigliani and Brumberg (1954), the life cycle hypothesis is driven by

individual utility—a function of an individual’s aggregate consumption. In this model, it is
assumed that individuals aim to maximize their lifetime utility given their total lifetime
wealth, W, which is the summation of one’s current level of resources (A0) and their future
earnings (H) (Yuh & Hanna, 2010).

W � A0 � H

If there is no uncertainty about remaining work life expectancy (RWLE) and retirement life
expectancy (RLE), and a zero real rate of interest, as initially assumed in Modigliani and
Brumberg’s model, then optimal consumption, C, is equal to total wealth divided by total life
expectancy (where LE � RWLE � RLE):

3D.R. Lawson, S.J. Heckman / Financial Services Review 26 (2017) 1–18



C �
W

LE

To smooth consumption, individuals may borrow initially until income is high enough to
support expenses. They then save when income is above expenses and dissave during
retirement. Ando and Modigliani (1963) present the following equation for total consump-
tion, C, in year t, for a person that is age T:

Ct
T � �t

T�t
T

where �t
T represents what Ando and Modigliani (1963) call a proportionality factor, which

depends on the rate of return an individual can earn on assets, the individual’s current age,
and the form of the utility function to be used. The second part, �t

T, represents total resources
available to that individual, which can be expressed as net worth.

Based on this framework, holding all else (especially retirement age) constant, it is
expected that an increase in life expectancy would decrease the amount of consumption in
each year, t. In other words, if total resources remain constant and an individual lives longer,
consumption would need to decrease, assuming that the compounded interested earned on
assets from the previous periods are insufficient to fund extra years of income needs.1

1.3.2. Behavioral life cycle hypothesis
To consider some of the behavioral tendencies inherent in humans, Shefrin and Thaler

(1988) developed a derivative of the life cycle model known as the behavioral life cycle
hypothesis. While they state that the life cycle hypothesis was useful in its own right,
particularly as a prescriptive model, they acknowledge that individuals do not always act in
an optimal manner. Therefore, modifications to the model were made to make the “theory
more behaviorally realistic” (Shefrin & Thaler, 1988, p. 1). They modify the life cycle model
to account for three human behavior factors often missing from economic theory: (a)
self-control, (b) mental accounting, and (c) framing.

According to Shefrin and Thaler (1988), self-control is made up of three components: (a)
internal conflict, (b) temptation, and (c) willpower. Internal conflict is described as an
inherent battle between the planner and the doer within each individual. The planner is
concerned with the long term while the doer aims to maximize utility in the present.
Temptation deals with the choices that are presented to individuals, recognizing that for each
choice there are differing degrees of enticement. Willpower is the mental cost associated with
resisting temptation and the degree of temptation influences how much willpower the
individual must expend to make a choice that aligns with their utility-maximizing preference
for both the long- and short-term. The more an individual is tempted to indulge in something,
the more willpower they must expend to resist the temptation. Although willpower can be
taxing, it is necessary to reduce consumption (Shefrin & Thaler, 1988). Individuals will
engage in willpower if the utility lost from less consumption (i.e., not succumbing to
temptation) is less than the utility cost of engaging in willpower. Hence, willpower is exerted
when the net marginal cost of using it is greater than zero (Shefrin & Thaler, 1988).
Self-control, then, is the relationship among internal conflict, temptation, and willpower.

4 D.R. Lawson, S.J. Heckman / Financial Services Review 26 (2017) 1–18



Consumption in today’s economy is heavily influenced by advertising, which is designed
to convince consumers that they need products they previously had not considered. Approx-
imately $180 billion is spent annually on advertising in the United States (and projected to
grow to nearly $220 billion by 2018) (Statista, 2016), which results in nearly constant
exposure to advertisements. Consumers are constantly at conflict with themselves because
they are tempted to buy things that they may not actually need or want. Shefrin and Thaler
(1988) argue that an economic savings model that does not factor in self-control is a
mis-specified model. Thus, willpower—the mental costs of resisting temptation—must be
considered.

A lack of self-control can lead to financial problems down the road (e.g., inadequate
retirement preparation), whereas too much self-control can deprive someone of their wants
and needs. Although individuals may expect to live a long time in retirement, they may lack
the ability to exert self-control and defer consumption from today to future periods by saving.
From this perspective, it is predicted that an increase in life expectancy would not result in
a change in consumption because of willpower and self-control issues among working
households for which retirement is not salient. That is to say, retirement may be too far in
the future to exert self-control and reduce consumption for young households. However, as
retirement becomes more salient (i.e., as households near retirement) it is expected that
longer life expectancies would have a negative effect on consumption.

1.3.3. Summary
Subjective life expectancy should have a negative effect on consumption according to the

life cycle hypothesis. In general, those who expect to live longer should save more and spend
less. However, consumers may struggle with self-control according to the behavioral life
cycle hypothesis. In the case of self-control problems, life expectancy is expected to have a
negligible or inconsistent effect on consumption for individuals far from retirement but may
have a negative effect on consumption for individuals near retirement. That is to say, for
individuals near retirement, those who expect to live longer should save more and spend less
than individuals far from retirement.

2. Method

2.1. Data

The 2013 SCF, a triennial survey sponsored by the Federal Reserve Board (Fed), is used
and our sample is limited to respondents who work full-time, expect to retire, and have a
retirement life expectancy period greater than zero (N � 2,632). The reason for this
restriction is that full-time workers have the greatest flexibility in adjusting saving and
consumption decisions based on their life expectancy. Retired households or those who work
less than full-time may systematically differ from full-time workers. Additionally, the sample
is limited to those with a RLE period greater than zero because a few respondents had a
negative RLE (i.e., the respondent expected to retire at a date later than his or her life
expectancy), possibly indicating that they did not understand the questions. The switch

5D.R. Lawson, S.J. Heckman / Financial Services Review 26 (2017) 1–18



variable in the SCF is used to obtain respondent characteristics (Lindamood, Hanna, & Bi,
2007). For consistency, weighted statistics and regression models are presented, although
multivariate results are robust to weighting. The SCF relies on a complex sampling design
(Nielsen & Seay, 2014; Shin & Hanna, 2016) and uses multiple imputation to handle missing
data and to help protect participants’ privacy (Montalto & Sung, 1996). Therefore, an
available Stata program (see Nielsen, 2015) is used to implement the Repeated Imputation
Inference (RII) method (Montalto & Sung, 1996) while also using the replicate weight file
provided by the Fed to bootstrap the standard errors (Nielsen & Seay, 2014). This procedure
provides estimates of variance that are closer to the true variance.

2.2. Variables

Three dependent variables were chosen to represent consumption decisions in the SCF: (a)
financial assets, (b) credit card balance, and (c) saving behavior. Both financial assets and
credit card balances are log-transformed to correct for the highly skewed distribution. Our
third dependent variable is saving behavior (whether or not the respondent saved or not).2

This is measured by whether or not the household reported spending less than income, after
considering durable goods purchases (in particular, costs associated with purchasing a car,
home, or investments are considered savings in the SCF). Households who spent all or more
of their income are described as non-savers.

The key independent variables are remaining work life expectancy (RWLE) and retire-
ment life expectancy (RLE). RWLE is computed by subtracting respondent age from
respondent’s estimated retirement age. RLE is computed by subtracting respondent’s retire-
ment age from respondent’s estimated life expectancy. From a life cycle perspective,
examining RWLE and RLE is reasonable as it highlights the tradeoff between saving (during
RWLE) and dissaving (during RLE). Delaying retirement while holding life expectancy
constant has the beneficial effect (assuming a life cycle hypothesis perspective) of increasing
wealth while reducing the number of years without earned income, which has an overall
positive effect on consumption during the remainder of the life. The opposite also holds—
retiring early reduces wealth while increasing the number of years without earned income.
Therefore, coding each period should yield clearer results.

To identify the ceteris paribus influence of life expectancy on consumption decisions, a
number of variables were included as controls, including relationship status and gender
(married, single male, single female, or partner), racial or ethnic status (White, Black,
Hispanic, or Other), education (less than high school, high school, some college, bachelor’s
degree, or graduate degree), whether or not the respondent has children in the household that
are under the age of 18, and health status (poor health vs. good health). Other controls are
if the respondent contributes to a retirement plan at work, if they have access to $3,000 from
a family member or friend in the event of an emergency, their planning horizon (short,
intermediate, or long), income (log), net worth (log; only for the saved model), and the
degree to which the survey administrator believed that the respondent understood the SCF
questions (poor/fair, good, or excellent). To account for negative values of net worth, log
(0.01) is used. Additional controls are whether or not the respondent owns a home, plans to

6 D.R. Lawson, S.J. Heckman / Financial Services Review 26 (2017) 1–18



give a bequest (Spaenjers & Spira, 2015), and if they expect to receive an inheritance in the
future.

2.3. Models

Ordinary least squares is used to model log credit card balance and log financial assets and
binary logistic regression is used to model whether or not the household saved. The
coefficients on RLE and RWLE are the key parameters of interest. After running each of the
models on the full sample of full-time workers who expect to retire and have a RLE greater
than zero (N � 2,632), each model is also run to compare those with a RWLE of less than
or equal to 10 years (n � 617) with those who have a RWLE greater than 10 years (n �
1,212). The secondary analysis helps examine whether there were differences between the
two groups based on the saliency of retirement, as suggested by the behavioral life cycle
model, and is limited to those between the ages of 35 and 62. Other recent studies have used
35 to 60 (Kim & Hanna, 2015a) and 35 to 70 (Kim, Hanna, & Chen, 2014), stating that those
under the age of 35 are likely to have significant changes in jobs and/or marital status, which
reduces the quality of the expected retirement measure. Given that many respondents
expect to retire in their mid-60s (about 30% expect to retire at age 65 and about 56% expect
to retire between the ages of 63 and 70), age 62 is used as the cutoff age for the secondary
analysis.

3. Results

3.1. Descriptive results

Most respondents are in a relationship—53% are married and 11% have a partner. Most
of the respondents are White (70%) and 48% have completed a college degree. Fifty-nine
percent of the respondents have a retirement plan at their workplace. Most respondents (72%)
have access to at least $3,000 in case of an emergency through a family member or friend.
Half of the respondents (50%) plan for the intermediate-term (2–9 years) while many (35%)
plan for the short-term (within a year). Approximately half of the respondents have children
under the age of 18 (51%). Almost all of the respondents are in good health (85%). Nearly
48% of the respondents have an excellent understanding of the SCF questions while 45%
have a good understanding and approximately 6% have a poor or fair understanding of the
SCF questions. A majority of respondents spend less than they make (63%), while nearly
11% say that they spend more than they earned over the last year. Respondents report a
median income of $68,988 with a median net worth of $103,980, median financial assets
totaling approximately $225,000, and a majority (69%) owned a home. Approximately 36%
plan to give a bequest while 17% are planning on receiving an inheritance. Most continue to
carry some level of debt (86%) with a median debt level of $65,600. Nearly 47% hold credit
card debt with an average balance of $2,792. The average age of the respondent is in the
mid-40s (M � 44.45, SD � 5.52) with an average expected retirement age of 63.56 (SD �
2.92), and an average subjective life expectancy of 84.60 (SD � 4.36).

7D.R. Lawson, S.J. Heckman / Financial Services Review 26 (2017) 1–18



Fig. 1 shows the cumulative distribution of life expectancy for the full sample, and for
those near and far from retirement. The range on life expectancy is 39 years to 150 years with
approximately 95% of respondents expecting to live somewhere between 70 and 100 years
old. Given the purpose of this study, even if people have unreasonable expectations about
their life expectancy, outliers should not affect our results because we are interested in how
their life expectancy perceptions influence consumption decisions. Fig. 2 presents the
cumulative distribution of expected retirement age for the full sample, and for those near and
far from retirement. The range for expected retirement age is 24 years to 102 years of age
with nearly 60% expecting to retire between age 63 and 75, and 95% of respondents expect
to retire between ages 49 and 79. Again, it was considered important to keep the outliers on
retirement age because the purpose to see how estimates of life expectancy and retirement
age influence consumption decisions during each period. Finally, mean remaining work life
expectancy (RWLE) is 19.11 (SD � 5.06) and the average retirement life expectancy (RLE)
is 21.05 (SD � 4.96).

In the secondary analysis, respondents who are within 10 years of retirement (i.e., near
retirement) have an average age of 55.44 (SD � 2.20) with an expected retirement age of
61.91 years (SD � 2.17), whereas those who are further from retirement (11 or more years
away) have an average age of 45 years (SD � 2.65) and expect to retire just before age 65
(M � 64.88, SD � 2.16). Those who are closer to retirement expect to live a little longer
(M � 84.10, SD � 4.17) than those who are further from retirement (M � 83.72, SD � 4.08).

Fig. 1. Cumulative distribution of subjective life expectancy for the full sample (N � 2,632), for those near
retirement (n � 617), and for those further from retirement (n � 1,212).

Fig. 2. Cumulative distribution of expected retirement age for the full sample (N � 2,632), for those near
retirement (n � 617), and for those further from retirement (n � 1,212).

8 D.R. Lawson, S.J. Heckman / Financial Services Review 26 (2017) 1–18



Respondents expecting to retire in the next 10 years expect to have a longer retirement period
(M � 22.19, SD � 4.70) than those who are further from retirement (M � 18.85, SD � 4.21).
Respondents who expect to retire within the next 10 years expect to work an average of 6.47
years more (SD � 1.12) while those expecting to retire later expect to work an average of
19.85 more years (SD � 2.70).

Respondents near retirement have slightly more than twice the median net worth
($260,000) than those who are further from retirement ($128,770), and have approximately
42% more financial assets (median � $367,400) than those further from retirement (me-
dian � $257,300). About 84% of those near retirement own their home whereas nearly 72%
of those 11 or more years away from retirement own their home. Of those near retirement,
approximately 66% report spending less than they made in the prior year and 11% report
spending more than they make, whereas only 60% of those further from retirement report
saving and nearly 13.5% spend more than they make. An interesting find was that both
groups have a higher prevalence of having debt (87% for those near retirement, 88% those
further from retirement) than the full sample (86%) and holding credit card debt (49% for
those near retirement, 48% for those further from retirement) than the full sample (47%).
Absolute dollar amounts of total debt ($72,000 for those near retirement, $94,000 for those
further from retirement) and credit card debt ($3,150 for those close to retirement, $3,203 for
those further from retirement) are also higher than the full sample ($65,600, $2,792,
respectively). Approximately 35% of those within 10 years of retirement have minor children
in the home whereas 64% of those 11 or more years away from retirement have minor
children at home. Almost 70% of those nearing retirement have a retirement plan at their
place of employment, while only 60% of those not expecting to retire within the decade have
such a plan. Finally, proportionately more of those who are nearing retirement (56%) plan
for the intermediate-term than did those who were further from retirement (49%), while those
further from retirement are planning for the short-term proportionately more (35%) than
those near retirement (32%). Full descriptive results can be seen in Table 1.

3.2. Multivariate results

The regression models are presented in Tables 2 through 4 with each table containing the
overall model first, followed by the model representing full-time workers ages 35 to 62
expecting to retire in the next 10 years, and ending with the model for full-time workers ages
35 to 62 expecting to retire beyond 10 years.

3.2.1. Financial assets
Considering first the factors that influence the log of financial assets, RLE and RWLE are

significant. Contrary to expectations, RLE is negatively associated, indicating that a one-year
increase in RLE is associated with a 1% decrease in financial assets (B � �0.01, SE � 0.00,
p � 0.01). RWLE is also negatively associated with log financial assets—a one-year increase
in a respondent’s working career is associated with a 4% reduction in financial assets (B �
�0.04, SE � 0.00, p � 0.01). However, this makes conceptual sense given that those with
a longer time until retirement are likely to have less financial assets. Owning a home and
indicating an intent to leave a bequest are both positively associated with financial asset

9D.R. Lawson, S.J. Heckman / Financial Services Review 26 (2017) 1–18



Table 1 Sample descriptives of full-time workers: Full sample, sample of those near retirement, and sample
of those more than 10 years from retirement

Full model 0–10 years 11� years
(N � 2,632) (n � 617) (n � 1,212)

Variable (reference group) M/% SD M/% SD M/% SD

Age 44.45 5.42 55.44 2.20 45.03 2.65
Retirement age 63.56 2.92 61.91 2.17 64.88 2.16
Life expectancy measures

Subjective life
expectancy

84.60 4.36 84.10 4.17 83.72 4.08

Retirement life
expectancy

21.05 4.96 22.19 4.70 18.85 4.21

Remaining work life
expectancy

19.11 5.06 6.47 1.12 19.85 2.70

Income and asset info
Incomea $68,988 $83,192 $78,119
Net wortha $103,980 $260,000 $128,770
Financial assetsa $224,950 $367,400 $257,300
Owns home 69.18% 84.29% 71.96%

Bequest motive 36.02% 35.48% 32.25%
Receive inheritance 17.41% 15.44% 18.18%
Debt burden

Have debt 86.06% 87.05% 88.28%
Have credit card 76.55% 83.35% 76.81%
Have credit card

balance
46.83% 49.35% 48.41%

Total debta $65,600 $72,000 $94,000
Total credit card

balanceb
$2,792 $2,932 $3,150 $3,194 $3,203 $3,049

Spending behavior
Spending � Income 10.98% 11.02% 13.45%
Spending � Income 25.56% 23.03% 26.28%
Spending � Income 63.45% 65.95% 60.27%

Relationship status
Married 53.36% 59.00% 58.26%
Single male 13.42% 10.51% 12.62%
Single female 22.78% 25.13% 19.44%
Partner 10.44% 5.36% 9.69%

Racial/ethnic status
White 69.96% 73.21% 70.07%
Black 13.96% 13.74% 13.60%
Hispanic 11.07% 9.60% 10.68%
Other 5.01% 3.46% 5.65%

Education
No high school diploma 5.96% 8.08% 5.51%
High school 20.05% 20.74% 19.20%
Some college 25.78% 26.04% 25.76%
Bachelor’s 28.58% 24.79% 29.97%
Graduate 19.62% 20.36% 19.57%

Have children � 18 Years 50.58% 35.65% 64.27%
Perceived poor health 14.74% 16.69% 14.22%

10 D.R. Lawson, S.J. Heckman / Financial Services Review 26 (2017) 1–18



ownership. Unmarried individuals have less financial assets than their married counterparts.
When compared to White workers, non-White workers also have less financial assets.
Education is positively associated with financial assets, as is income. Those with minor
children and/or in poor health are found to have less financial assets than those without
children and in good health. Full-time workers that are perceived to have an excellent
understanding of the SCF questions have greater financial assets than those with a good
understanding of the questions, while those with a poor-to-fair understanding of the SCF
questions have less financial assets than do respondents who have a good understanding of
the questions. Workers with a retirement plan and access to $3,000 are found to have a
positive association with financial assets, too. Finally, those that have planning horizons of
less than 10 years are less likely to accumulate financial assets compared with those who plan
for the long term.

There are some slight differences in the secondary analysis where the sample is split into
those who are near retirement (within 10 years) and those who are far from retirement (more
than 10 years). RLE and RWLE are not significantly associated with financial assets for those
who expect to retire within 10 years. However, RLE and RWLE are significant in determining
financial assets for respondents that expect to retire beyond 10 years. Again, results seem odd for
RLE, as an increase of RLE by one year is associated with a decrease in financial assets by 2%
(B � �0.02, SE � 0.01, p � 0.01). An increase in RWLE by one more working year results in
a 3% decrease in financial assets (B � �0.03, SE � 0.01, p � 0.01).

3.2.2. Credit card debt
Among credit card holders, RLE and RWLE have no significant effect on outstanding

credit card balance. Home ownership is positively associated with the amount of credit card
debt held. Single males hold less amounts of credit card debt when compared to their married
counterparts. Compared to White households, households of other races and ethnicities

Table 1 (Continued)

Full model 0–10 years 11� years
(N � 2,632) (n � 617) (n � 1,212)

Variable (reference group) M/% SD M/% SD M/% SD

Understanding of SCF’s
Poor/fair 6.36% 6.45% 5.02%
Good 45.37% 45.61% 44.32%
Excellent 48.27% 47.94% 50.66%

Has retirement plan 58.98% 69.21% 60.54%
Has access to $3,000 72.71% 74.80% 71.31%
Planning horizon

Short 35.49% 31.54% 34.59%
Intermediate 49.89% 55.50% 49.25%
Long 14.63% 12.95% 16.17%

Source: Weighted analysis of the 2013 Survey of Consumer Finances.
Note: Analysis includes only those that are full-time, expect to retire, and have a retirement life expectancy

greater than zero.
a Results are reported as the Median.
b Reported as Mean because of the Median equaling zero.

11D.R. Lawson, S.J. Heckman / Financial Services Review 26 (2017) 1–18



(e.g., Asian, Native American, etc.) have lower balances and Hispanic households have
higher balances. More educated workers hold more credit card debt than did their peers
(with the exception of those having a graduate degree). Those who are in worse health
have less credit card debt, and those who have a retirement plan have more credit card
debt than those who did not have a retirement plan at work. Lastly, workers with shorter
planning horizons hold more credit card debt than do their counterparts with longer planning
horizons.

Similar results are found in the secondary analysis—RLE and RWLE are not significantly
associated with credit card balances for respondents near or far from retirement.

Table 2 OLS regressions on the log of financial assets, full sample of full-time workers, sample of those
near retirement, and sample of those more than 10 years from retirement

Full model 0–10 years 11� years
(N � 2,632) (n � 617) (n � 1,212)

Variables (reference group) B SE. B p B SE. B p B SE. B p

Retirement life expectancy �0.01 (0.00) �.001 0.01 (0.01) 0.408 �0.02 (0.01) 0.007
Remaining work life expectancy �0.04 (0.00) �.001 �0.02 (0.03) 0.499 �0.03 (0.01) 0.004
Own’s home 0.79 (0.10) �.001 0.53 (0.23) 0.020 0.95 (0.14) �.001
Bequest motive 0.52 (0.07) �.001 0.32 (0.15) 0.040 0.72 (0.09) �.001
Receive inheritance 0.15 (0.09) 0.074 �0.01 (0.15) 0.962 0.09 (0.12) 0.449
Relationship status (married)

Single male �0.31 (0.11) 0.004 �0.58 (0.29) 0.048 �0.46 (0.16) 0.005
Single female �0.50 (0.13) �.001 �0.81 (0.31) 0.009 �0.60 (0.19) 0.002
Partner �0.26 (0.12) 0.031 �0.27 (0.34) 0.425 �0.01 (0.18) 0.955

Racial/ethnic status (White)
Black �0.64 (0.13) �.001 �0.94 (0.30) 0.002 �0.37 (0.18) 0.042
Hispanic �0.48 (0.13) �.001 �0.95 (0.36) 0.009 �0.50 (0.23) 0.029
Other 0.04 (0.11) 0.717 0.16 (0.39) 0.673 0.08 (0.14) 0.546

Education (no high school
diploma)

High school 0.87 (0.19) �.001 0.95 (0.39) 0.016 0.56 (0.30) 0.063
Some college 1.55 (0.21) �.001 1.52 (0.37) �.001 1.34 (0.31) �.001
Bachelor’s 1.97 (0.22) �.001 1.62 (0.42) �.001 1.95 (0.31) �.001
Graduate 2.29 (0.22) �.001 1.81 (0.44) �.001 2.28 (0.34) �.001
Log of income 0.62 (0.13) �.001 0.73 (0.27) 0.007 0.61 (0.16) �.001
Have kids �0.16 (0.04) �.001 �0.26 (0.09) 0.004 �0.17 (0.05) 0.001
Perceived poor health �0.21 (0.10) 0.040 0.13 (0.21) 0.540 �0.26 (0.18) 0.148

Understand SCF questions?
(good)

Poor/fair �0.45 (0.18) 0.015 �0.49 (0.38) 0.202 �0.87 (0.40) 0.029
Excellent 0.15 (0.07) 0.024 �0.11 (0.16) 0.495 0.18 (0.10) 0.088
Having retirement plan 1.26 (0.07) �.001 1.27 (0.15) �.001 1.38 (0.10) �.001
Access to $3,000 0.91 (0.10) �.001 0.75 (0.22) 0.001 0.87 (0.15) �.001

Planning horizon (long)
Short �0.64 (0.11) �.001 �0.84 (0.25) 0.001 �0.48 (0.16) 0.003
Intermediate �0.29 (0.09) 0.001 �0.31 (0.21) 0.132 �0.24 (0.11) 0.031
Constant 1.28 (1.43) 0.369 0.56 (3.09) 0.857 1.03 (1.81) 0.569

R2 0.57 0.58 0.56
F 142.68 34.58 63.86

Source: Weighted analysis of the 2013 Survey of Consumer Finances; RII technique and bootstrapping used.
Note: Analysis includes only those that are full-time, expect to retire, and have a retirement life expectancy

greater than zero.

12 D.R. Lawson, S.J. Heckman / Financial Services Review 26 (2017) 1–18



3.2.3. Saving behavior
Looking at factors that influence whether or not a worker saved, RLE is again significant

and positively associated with the likelihood of saving (B � 0.01, SE � 0.00, p � 0.05, odds
ratio [OR] � 1.01). Other positive factors are having a bequest motive, being a single male
or in a partnership, income, and net worth. Hispanics, having kids, and being in poor health
are negatively associated with saving. Having access to $3,000 is positively associated with
saving, while those with short-term planning horizons (i.e., less than one year), intermediate-
term planning horizons (i.e., two to nine years), and credit card debt are less likely to save.

A few differences exist in the secondary analysis. For those near retirement, RLE is not
significant but RWLE is significantly and negatively associated with saving. As RWLE
increases by one year for individuals near retirement, the odds of saving are about 12% lower

Table 3 OLS regressions on the log of credit card balance, full sample of full-time workers, sample of those
near retirement, and sample of those more than 10 years from retirement

Full model 0–10 years 11� years
(N � 2,632) (n � 617) (n � 1,212)

Variables (reference group) B SE. B p B SE. B p B SE. B p

Retirement life expectancy �0.01 (0.01) 0.338 �0.01 (0.01) 0.399 �0.01 (0.01) 0.469
Remaining work life expectancy 0.00 (0.01) 0.627 �0.04 (0.06) 0.529 0.01 (0.02) 0.410
Own’s home 0.73 (0.14) �.001 0.42 (0.43) 0.329 0.52 (0.23) 0.024
Bequest motive �0.11 (0.14) 0.455 �0.48 (0.29) 0.103 0.10 (0.22) 0.653
Receive inheritance �0.24 (0.19) 0.196 0.56 (0.43) 0.197 �0.30 (0.25) 0.228
Relationship status (married)

Single male �0.93 (0.19) �.001 �0.22 (0.41) 0.596 �0.99 (0.39) 0.010
Single female �0.41 (0.22) 0.056 �0.94 (0.40) 0.021 �0.01 (0.28) 0.969
Partner 0.34 (0.21) 0.116 0.93 (0.72) 0.193 0.50 (0.33) 0.134

Racial/ethnic status (White)
Black 0.22 (0.24) 0.370 1.56 (0.47) 0.001 �0.46 (0.35) 0.199
Hispanic 0.73 (0.24) 0.003 0.32 (0.83) 0.699 0.25 (0.37) 0.506
Other �0.92 (0.33) 0.005 0.87 (0.73) 0.232 �1.24 (0.45) 0.006

Education (no high school diploma)
High school 0.80 (0.29) 0.006 �0.16 (0.66) 0.806 1.02 (0.40) 0.010
Some college 1.44 (0.31) �.001 0.83 (0.59) 0.158 1.46 (0.41) �.001
Bachelor’s 1.24 (0.34) �.001 �0.14 (0.66) 0.833 1.20 (0.45) 0.007
Graduate 0.70 (0.36) 0.056 �0.30 (0.72) 0.680 0.99 (0.50) 0.048

Log of income 0.07 (0.07) 0.316 �0.53 (0.18) 0.004 0.03 (0.09) 0.783
Have kids �0.06 (0.06) 0.324 �0.04 (0.19) 0.840 �0.06 (0.09) 0.483
Perceived poor health �0.58 (0.17) 0.001 �1.02 (0.46) 0.027 �0.33 (0.28) 0.234
Understand SCF questions? (good)

Poor/fair �0.36 (0.25) 0.157 �1.41 (0.79) 0.075 0.13 (0.44) 0.761
Excellent �0.06 (0.14) 0.691 0.07 (0.33) 0.839 �0.12 (0.24) 0.605

Having retirement plan 0.46 (0.14) 0.002 0.92 (0.34) 0.006 0.57 (0.21) 0.007
Access to $3,000 0.27 (0.15) 0.074 0.82 (0.33) 0.013 0.51 (0.25) 0.039
Planning horizon (long)

Short 0.99 (0.21) �.001 1.25 (0.51) 0.015 0.65 (0.38) 0.087
Intermediate 0.59 (0.18) 0.001 0.52 (0.45) 0.253 0.69 (0.36) 0.054

Constant 0.78 (0.86) 0.364 8.24 (2.11) �.001 0.95 (1.26) 0.447
R2 0.05 0.09 0.04
F 5.39 2.30 2.06

Source: Weighted analysis of the 2013 Survey of Consumer Finances; RII technique and bootstrapping used.
Note: Analysis includes only those that are full-time, expect to retire, and have a retirement life expectancy

greater than zero.

13D.R. Lawson, S.J. Heckman / Financial Services Review 26 (2017) 1–18



(B � �0.13, SE � 0.03, p � 0.001, OR � 0.88). For individuals further from retirement,
RLE is significant and positively associated with saving while RWLE did not have a
significant effect.

4. Discussion

When breaking down subjective estimates of life expectancy into a retirement life
expectancy period and a remaining work life expectancy period, there is confusing, contra-
dictory, and ultimately very little influence on household consumption decisions. Retirement

Table 4 Logistic regressions on the likelihood of saving, full sample of full-time workers, sample of those
near retirement, and sample of those more than 10 years from retirement

Full model 0–10 years 11� years
(N � 2,632) (n � 617) (n � 1,212)

Variables (reference group) B SE. B p OR B SE. B p OR B SE. B p OR

Retirement life expectancy 0.01 (0.00) 0.014 1.01 �0.02 (0.01) 0.054 0.98 0.01 (0.01) 0.045 1.01
Remaining work life expectancy 0.01 (0.00) 0.159 1.01 �0.13 (0.03) �.001 0.88 �0.02 (0.01) 0.093 0.98
Own’s home 0.10 (0.10) 0.344 1.10 0.46 (0.29) 0.120 1.58 �0.11 (0.16) 0.480 0.89
Bequest motive 0.19 (0.08) 0.015 1.21 �0.01 (0.24) 0.959 0.99 0.02 (0.13) 0.894 1.02
Receive inheritance �0.14 (0.12) 0.255 0.87 0.29 (0.33) 0.392 1.33 �0.03 (0.16) 0.840 0.97
Relationship status (married)

Single male 0.37 (0.16) 0.017 1.45 1.33 (0.48) 0.006 3.77 0.10 (0.23) 0.676 1.10
Single female �0.13 (0.13) 0.328 0.88 0.07 (0.35) 0.848 1.07 �0.34 (0.17) 0.043 0.71
Partner 0.28 (0.13) 0.030 1.32 0.92 (0.66) 0.161 2.50 0.16 (0.18) 0.379 1.17

Racial/ethnic status (White)
Black �0.13 (0.12) 0.302 0.88 �0.27 (0.34) 0.430 0.77 0.01 (0.18) 0.947 1.01
Hispanic �0.34 (0.11) 0.002 0.71 �0.26 (0.45) 0.558 0.77 �0.45 (0.19) 0.018 0.64
Other 0.04 (0.18) 0.818 1.04 �0.89 (0.40) 0.027 0.41 0.35 (0.26) 0.177 1.42

Education (no high school diploma)
High school �0.09 (0.18) 0.622 0.92 0.47 (0.47) 0.318 1.59 �0.28 (0.33) 0.397 0.76
Some college �0.11 (0.21) 0.607 0.90 0.50 (0.47) 0.287 1.64 �0.41 (0.37) 0.272 0.67
Bachelor’s 0.34 (0.21) 0.110 1.40 1.34 (0.52) 0.010 3.84 0.08 (0.38) 0.827 1.09
Graduate 0.27 (0.22) 0.220 1.31 1.00 (0.55) 0.067 2.72 0.37 (0.40) 0.360 1.45

Log of income 0.48 (0.09) �.001 1.61 0.65 (0.19) 0.001 1.91 0.55 (0.13) �.001 1.73
Log of net worth 0.04 (0.01) �.001 1.04 0.02 (0.03) 0.513 1.02 0.05 (0.02) 0.004 1.05
Have kids �0.20 (0.03) �.001 0.82 �0.27 (0.12) 0.026 0.77 �0.21 (0.05) �.001 0.81
Perceived poor health �0.30 (0.12) 0.010 0.74 �0.56 (0.31) 0.067 0.57 �0.56 (0.20) 0.005 0.57
Understand SCF questions? (good)

Poor/fair �0.22 (0.16) 0.171 0.81 �0.02 (0.52) 0.962 0.98 �0.02 (0.29) 0.952 0.98
Excellent 0.07 (0.08) 0.421 1.07 �0.22 (0.22) 0.319 0.80 �0.03 (0.13) 0.794 0.97

Having retirement plan 0.14 (0.09) 0.111 1.15 �0.23 (0.21) 0.268 0.79 0.37 (0.12) 0.002 1.45
Access to $3,000 0.27 (0.10) 0.008 1.31 0.13 (0.26) 0.621 1.14 0.16 (0.17) 0.351 1.17
Planning horizon (long)

Short �0.67 (0.15) �.001 0.51 �1.17 (0.45) 0.010 0.31 �0.53 (0.22) 0.016 0.59
Intermediate �0.46 (0.14) 0.001 0.63 �0.99 (0.37) 0.008 0.37 �0.30 (0.21) 0.152 0.74

Have credit card balance �0.62 (0.07) �.001 0.54 �1.13 (0.18) �.001 0.32 �0.64 (0.11) �.001 0.53
Constant �4.80 (1.01) �.001 0.01 �4.77 (2.25) 0.034 0.01 �5.06 (1.33) �.001 0.01
Pseudo R2 0.14 0.21 0.17
Log likelihood 473.51 166.81 277.85

Source: Weighted analysis of the 2013 Survey of Consumer Finances; RII technique and bootstrapping used.
Note: OR � odds ratio; analysis includes only those that are full-time, expect to retire, and have a retirement

life expectancy greater than zero.

14 D.R. Lawson, S.J. Heckman / Financial Services Review 26 (2017) 1–18



life expectancy is not a significant predictor in our primary analysis of credit card debt.
Additionally, RWLE is not a significant predictor in our primary analysis of credit card debt
and whether or not a respondent saved. Even when separating individuals into those who are
near and far from retirement, RLE is not significantly associated with financial assets, credit
card debt, and savings behavior for those near retirement, and it is not significantly associated
with credit card debt for those further from retirement. RWLE, on the other hand, is not
significantly associated with financial assets and credit card debt for those near retirement,
and is not associated with credit card debt and savings behavior for those further from
retirement. RLE exhibits effect directions contrary to expectations in financial assets for both
the primary and secondary analysis. Subjective life expectancy was only positively associ-
ated with saving behavior for those far from retirement.

Although the life cycle hypothesis suggests that life expectancy should influence house-
hold consumption choices, the behavioral life cycle model suggests that there may be other
factors that prevent households from smoothing consumption according to the life cycle
hypothesis. For example, even if a household realizes they may live longer, holding
retirement age constant, they may lack the self-control or willpower to change consumption
today to smooth consumption over the remainder of their lifetime.

The negative direction of the RLE factor on financial assets for the overall model may
indicate what previous literature has found—U.S. consumers are not doing a good job at
preparing for retirement (Aegon, 2016; Kim & Hanna, 2015b; Yuh, Montalto, & Hanna,
1998). Additionally, the fact that RLE and RWLE are not significant for those near
retirement may show that people are not able to account for their life expectancy when
making consumption choices. Alternatively, it could be that these households believe that,
although retirement is salient, there is nothing they can do, so they do not factor in their RLE
and RWLE when making consumption decisions related to their financial assets. Finally,
because RLE is significant, but negatively associated with financial assets for those who plan
to retire more than a decade from now, these individuals may be making consumption
choices that are not in-line with the life cycle hypothesis. That is, there may be a self-control
or willpower problem associated with their ability to properly plan for retirement.

4.1. Limitations

The most serious limitations center on measurement issues. Although the authors feel that
the best available consumption measures in the SCF are chosen, the analysis would be
improved with better measures of consumption. The saving measure is fairly limited as it is
simply whether or not the respondent spent less than they made last year, and it assumes that
the respondent actually knows whether or not they really did spend less than they made.
Another limitation is that the secondary analysis age group restriction may be limited
because of the fact that those in their mid-to-upper 30s, and even those in their mid-to-low
40s, are likely considerably different than those in their 50s and early 60s, regardless of
RWLE. For example, someone that is 40 years of age with an RWLE of 10 may be quite
different than a 55-year old with an RWLE of 10.

15D.R. Lawson, S.J. Heckman / Financial Services Review 26 (2017) 1–18



4.2. Implications

The results of this study have implications for professionals who help individuals and
families plan for retirement, employers who offer retirement plans, and policymakers
implementing legislation on retirement plans. Financial planners, financial counselors, and
financial therapists have an important role to play in helping clients plan for retirement and/or
make decisions around consumption choices that influence retirement preparedness. Given
that it is likely that clients are not taking into account their life expectancy, particularly the
amount of time they expect to spend in retirement, these professionals need to carefully help
clients consider how their consumption choices impact their future financial resources
available to them and how these choices may negatively influence their retirement lifestyle.
Having conversations early and often, and holding clients responsible for their actions is of
utmost importance. Professionals can help clients stick to the plan and illustrate the long-
term impact of short-term consumption. This may help the client to think twice before
making a choice that is detrimental to their financial assets available for retirement.

Second, these results may have meaningful impact for Human Resources directors and
those who manage retirement plans. Given that we have shifted from a defined benefit (i.e.,
pensions) system to a defined contribution (i.e., 401(k)s) system, and that willpower and
self-control may be an issue, incorporating programs that automatically opt-in employees to
retirement plans, as well as automatically increase the employees’ initial level of savings to
some maximum reasonable percentage, may be a welcome adjustment. Previous research has
found that this is a significant way to help increase employees’ savings (Beshears, Choi,
Laibson, & Madrian, 2009; Thaler & Benartzi, 2004). Although our results are not conclu-
sive, the general lack of impact of life expectancy on consumption choices suggest there is
a need for policies aimed at improving retirement preparation in the United States. If our
result is explained by willpower and self-control problems, legislation requiring employers
to, at minimum, automatically enroll their employees into defined contribution plans seem
beneficial. According to Aegon (2016), three-quarters (76%) of respondents agreed on some
level that the government should mandate employers to auto-enroll employees into the
defined contribution plans.

This article contributes to the literature by showing that even if consumers are considering
subjective life expectancy and their work life and retirement periods, they are doing a poor
job. This may be due, in part, to self-control and willpower issues in delaying current
consumption to provide consumption during retirement. The most likely explanation is that
consumers do not consider their life expectancy and associated work and retirement periods
when making consumption choices, resulting in a failure to adjust their consumption levels
over the course of their life. Subsequently, this results in a lack of retirement preparedness
for these individuals.

Notes

1 For example, using an 80% retirement income replacement rate for our sample’s
median income ($68,988) results in $55,190 in annual income needs during retirement.

16 D.R. Lawson, S.J. Heckman / Financial Services Review 26 (2017) 1–18



Based on the median financial assets of our sample ($224,950), a 24.5% annual return
would be needed to fund this retirement lifestyle in perpetuity ($55,190/$224,950).
Alternatively, a portfolio of approximately $1 million would be needed to earn the
amount necessary for retirement spending, given the assumptions above (based on a
more realistic 5.5% annual return; $1M*.055 � $55,000). Hence, it is plausible that
many households do not have the resources available to fund such a retirement income
requirement, and that current financial asset income would not be sufficient to fund
retirement income needs.

2 Spending relative to income (less, equal to, or more) in a multinomial logit was
modeled. Results did not differ so the simpler (binary) model is presented.

References

Aegon. (2016). A Retirement Wake-Up Call: The Aegon Retirement Readiness Survey 2016: United States
Country Report. Aegon: Center for Longevity and Retirement (available at http://www.aegon.com/
Documents/aegon-com/Research/2016-Retirement-Survey/2016-Aegon-Retirement-Readiness-Report-USA.pdf).

Ando, A., & Modigliani, F. (1963). The “life cycle” hypothesis of saving: Aggregate implications and tests. The
American Economic Review, 53, 55–84.

Beshears, J., Choi, J. J., Laibson, D., & Madrian, B. C. (2009). The importance of default options for retirement
savings outcomes: Evidence from the United States. In J. R. Brown, J. B. Liebman, & D. A. Wise (Eds.),
Social Security Policy in a Changing Environment (pp. 167–195). Chicago, IL: University of Chicago Press.

Bloom, D. E., Canning, D., & Graham, B. (2003). Longevity and life-cycle savings. Scandinavian Journal of
Economics, 105, 319–338.

Bloom, D. E., Canning, D., & Moore, M. (2004). Health, Longevity, and Optimal Retirement (available at SSRN:
http://dx.doi.org/10.2139/ssrn.594801).

Bloom, D. E., Canning D., Moore, M., & Song, Y. (2006). The Effect of Subjective Survival Probabilities on
Retirement and Wealth in the United States (No. 13688). Cambridge, MA: National Bureau of Economic
Research.

Deaton, A., & Paxson, C. (2000). Growth, demographic structure, and national savings in Taiwan. Population and
Development Review, 26, 141–173.

Elder, T. E. (2007). Subjective survival probabilities in the health and retirement study: Systematic biases and
predictive validity. Working Paper 2007–159, University of Michigan, Michigan Retirement Research Center.

Gupta, A., & Li, Z. (2013). Optimal annuity purchase decisions under uncertain lifetime. Journal of Family
Economic Issues, 34, 447–459. doi:10.1007/s10834–012-9340–1

Hamermesh, D. (1985). Expectations, life expectancy, and economic behavior. Quarterly Journal of Economics,
100, 389–408.

Hurd, M. D., & McGarry, K. (2002). The predictive validity of subjective probabilities of survival. Economic
Journal, 112, 966–985.

Kim, K. T., & Hanna, S. D. (2015). Does financial sophistication matter in retirement preparedness? Journal of
Personal Finance, 14, 9–20.

Kim, K. T., & Hanna, S. D. (2015). Do U.S. households perceive their retirement preparedness realistically?
Financial Services Review, 24, 139–155.

Kim, K. T., Hanna, S. D., & Chen, S, C-C. (2014). Consideration of retirement income stages in planning for
retirement. Journal of Personal Finance, 13, 52–64.

Lindamood, S., Hanna, S. D., & Bi, L. (2007). Using the Survey of Consumer Finances: Some methodological
considerations and issues. Journal of Consumer Affairs, 41, 195–222.

17D.R. Lawson, S.J. Heckman / Financial Services Review 26 (2017) 1–18



Modigliani, F., & Brumberg, R. (1954). Utility analysis and the consumption function: An interpretation of
cross-section data. The collected papers of Franco Modigliani (Vol. 6, pp. 3–45). Cambridge, MA: The MIT
Press.

Montalto, C. P., Yuh, Y., & Hanna, S. (2000). Determinants of planned retirement age. Financial Services
Review, 9, 1–15.

Montalto, C. P., & Sung, J. (1996). Multiple imputation in the 1992 survey of consumer finances. Financial
Counseling and Planning, 7, 133–141.

Nielsen, R. B. (2015). SCF complex sample specification for Stata. Technical note. Department of Financial
Planning Housing and Consumer Economics, University of Georgia, Athens, GA.

Nielsen, R. B., & Seay, M. C. (2014). Complex samples and regression-based inference: Considerations for
consumer researchers. The Journal of Consumer Affairs, 48, 603–619. doi:10.1111/joca.12038.

Shefrin, H. M., & Thaler, R. H. (1988). The behavioral life-cycle hypothesis. Economic Inquiry, 26, 609–643.
Shin, S. H., & Hanna, S. D. (2016). Accounting for complex sample designs in analysis of the survey of consumer

finances. The Journal of Consumer Affairs, 1–15. doi:10.1111/joca.12106.
Spaenjers, C., & Spira, S. M. (2015). Subjective life horizon and portfolio choice. Journal of Economic Behavior

& Organization, 116, 94–106.
Statista (2016). Media advertising spending in the United States from 2011 to 2018 (in billion U.S. dollars;

available at https://www.statista.com/statistics/272314/advertising-spending-in-the-us/).
Thaler, R. H. & Benartzi, S. (2004). Save more tomorrow: Using behavioral economics to increase employee

saving. Journal of Political Economy, 112, S164–S187.
Tsai, I-J., Chu, C. Y. C., & Chung, C-F. (2000). Demographic transition and household saving in Taiwan.

Population and Development Review, 26, 174–193.
Tversky, A., & Kahneman, D. (1974). Judgement under uncertainty: Heuristics and biases. Science, 185,

1124–1131.
Yuh, Y., & Hanna, S. D. (2010). Which households think they save? The Journal of Consumer Affairs, 44, 70–97.
Yuh, Y., Montalto, C. P., & Hanna, S. D. (1998). Are Americans prepared for retirement? Financial Counseling

and Planning, 9, 1–13.

18 D.R. Lawson, S.J. Heckman / Financial Services Review 26 (2017) 1–18