Focusing on both sides of the balance sheet: the potential

benefit of liability management

Zhikun Liua,*, David M. Blanchettb

aEmpower Retirement, 8515 E. Orchard Road, 4T2, Greenwood Village, CO 80111, United States
bMorningstar Investment Management LLC, 22 W. Washington Street, Chicago, IL 60602, United States

Abstract

Debt has become a significant issue among U.S. households with average household interest pay-

ments on liabilities exceeding expected returns on investment assets by more than 50%. In this study,

we explore the role of U.S. household debt and analyze the impact of different economic, demo-

graphic, and behavioral factors on household borrowing decisions, with a particular focus on “good”

and “bad” debts, which depend on type and interest rate. We estimate significant potential benefits

with improved liability management and find that households with lower asset, income, and educa-

tion levels are likely to benefit most from assistance with debt optimization. © 2021 Academy of

Financial Services. All rights reserved.

JEL classification: D12; G4; D15

Keywords: Debt management; Retirement financial planning; Behavioral finance; Financial decision making

1. Introduction

Debt is an increasingly significant part of the U.S. household balance sheet. After the

2007–2009 economic recession, debt levels of American households have increased signifi-

cantly (Bricker et al., 2017). The total U.S. household indebtedness was approximately

$14.27 trillion as of June 30, 2020, according to the Federal Reserve Bank of New York.

This is higher than the previous peak of $12.68 trillion in the third quarter of 2008 (adjusted

to 2019 dollars) and has increased by 27.9% since the second quarter of 2013 (Federal

Reserve Bank of New York, 2020). Additional information on this effect is shown in Fig. 1.

*Corresponding author: Tel.: +1-303-737-6207; fax: +1-303-737-6544.

Email address: zhikun.liu@empower-retirement.com (Z. Liu)

1057-0810/21/$ – see front matter © 2021 Academy of Financial Services. All rights reserved.

Financial Services Review 29 (2021) 121–145



Financial firms and advisors tend to spend significantly more time focusing on the assets

side of the household balance sheet compared with the liability side. This focus is consistent

with the traditional skill set of financial advisors—building portfolios—and reflects how

they are typically compensated (e.g., as a percentage of assets under management).

However, in this study, we demonstrate that this predominant attention paid to the assets

does not necessarily reflect the economic importance within the context of the household’s

entire balance sheet (i.e., when liabilities are taken into consideration). For example, data

from the 2016 Survey of Consumer Finances (SCF) suggest that among “low-to-affluent”

U.S. households, the total interest payments on debts exceed the expected gains from their fi-

nancial assets.1 Therefore, spending time on “debt optimization” is likely to result in better

outcomes than focusing on assets alone.

In this article, we explore the composition of household balance sheets in the United

States to understand the potential benefits associated with making more intelligent debt deci-

sions. Consistent with past research, we find that certain types of “bad” debts, such as credit

cards, are relatively common on household balance sheets today despite their high interest

rates (averaging approximately 15%).2 It is not clear to what extent interest rates could be

lower had the household done more due diligence on its debt decisions, or the extent to

which these debts can be refinanced, but it is likely that some, and possibly many, house-

holds’ situations can be improved (i.e., the household could reduce the interest rate on out-

standing debt). This analysis suggests more work should be done to understand the potential

benefits of improving household credit decisions.

The objective of this study is to demonstrate the urgency, importance, and potential

impact of household liability management by answering the following questions: What is

the current financial situation and retirement outlook of low-to-affluent U.S. households?

What factors are associated with household debts and leverage ratios? What is the difference

between “good” and “bad” debts?3 Will the attributes related to households carrying differ-

ent types of debts be similar? What kinds of families are more likely to have higher average

debt interest rates and how much could they save by accessing liability optimization?

Fig. 1. Growth trends in U.S. consumer credit owned. Source: Federal Reserve Board NY 2020 Consumer Credit
Panel/Equifax.

122 Z. Liu and D. M. Blanchett / Financial Services Review 29 (2021) 121–145



2. Literature review

Using data from multiple waves of the SCF, Emmons and Noeth (2013) report that the

household leverage ratio, defined by the sum of total debts divided by total assets, is higher

among younger families. Also using SCF data, Barba and Pivetti (2009) demonstrate that

the rising household indebtedness is associated with a decrease in the household savings

rate. This phenomenon is partially explained by lagging real wage growth and the tendency

for U.S. households to sustain their relative consumption level. Based on data from

Consumer Finance Monthly, Jiang and Dunn (2013) show that younger consumers have

higher levels of credit card debt and are repaying that debt at lower rates than previous

generations.4

Using the Health and Retirement Study (HRS), Gustman, Steinmeier, and Tabatabai

(2011) find that relative debt levels have been increasing for households that are near retire-

ment since the 2007–2009 recession and that much of the growth in debt appears to be

related to mortgage and housing expenses. The Quarterly Report on Household Debt and

Credit for the second quarter of 2020 supports this finding, reporting $9.78 trillion of mort-

gage balances for U.S. households (Federal Reserve Bank of New York, 2020). Using the

HRS, Lee, Lown, and Sharpe (2007) study the dissaving behavior of older Americans and

point out that financial debt carried into later life may result in reduced access to essential

health care, restrictions on activities, and delayed retirement.

Among the different categories of household liabilities, high-interest debts such as con-

sumer revolving credit debts can have significant negative impacts on household balance

sheets and cash flows. Based on information from the Federal Reserve Bank of New York

(2019), credit card balances stood at $870 billion as of the last quarter of 2018, with a sea-

sonally adjusted annual growth rate of 3%. Auto loan originations reached the highest

amount in the 19-year recorded history of the New York Fed in 2018, amounting to $584 bil-

lion. Unlike certain good debts, which tend to have lower relative interest rates and are typi-

cally used to purchase assets that are expected to generate long-term income or grow in

value (e.g., mortgages), bad debts such as credit cards, payday loans, and some auto loans

typically have higher interest rates and are generally associated with purchases (and assets)

that do not generate positive long-term returns (Hanson, 2006). In other words, the cost of

the good debts can often be outweighed by their potential long-term benefits, while the bad

debts’ high interest costs typically have little-to-zero long-term returns. Bad debts are not

only expensive, but they may also negatively influence the borrowers’ credit scores, hinder

their financial and retirement goals, and even cause stress and health issues. Davies,

Montgomerie, and Wallin (2015) report a positive relationship between individuals who are

deeply in debt and those who report mental health problems such as depression and physical

illness. Behavioral studies also indicate that consumers may be more likely to accumulate a

larger revolving credit card balance if they frequently pay behind schedule or miss payments

(Kim & DeVaney, 2001; Wärneryd, 1999). Therefore, helping consumers stay away from

“bad” debt and coaching them to develop good borrowing and accumulation habits are

essential approaches for advisors and financial planning firms to support their clients’ liabil-

ity management.

Z. Liu and D. M. Blanchett / Financial Services Review 29 (2021) 121–145 123



This effect, in which households spend more on their debt than they earn on their sav-

ings, is likely to continue in the future given the growth in debt among American house-

holds, as noted in Fig. 2 Therefore, it is essential for financial planning firms and

advisors to start putting a greater emphasis on their clients’ debt structures and help

them better manage their liabilities in order to help ensure that they can achieve a suc-

cessful retirement.

Zinman (2015) notes that research on the household debt has significantly lagged its sister

literature on the asset side of the balance sheet. While one may assume that households

make rational decisions regarding debt, Stango and Zinman (2016) find that cross-consumer

dispersion in credit card borrowing costs remains substantial even after controlling for debt

levels, credit risk, and product characteristics.

While the share of U.S. households with debt has been relatively constant, ranging from

72.3% in 1989 to 77.1% in 2016 (Bricker et al., 2017), the mean value of debt for American

families has increased significantly, from $66,900 in 1989 (in 2016 dollars) to $123,400 in

2016. This magnitude of debt increase has been observed across age levels. Within the 2016

SCF survey wave, the percentage of households carrying debt peaked around middle age

(approximately 45 years old), with the most common debt categories being mortgages, credit

card debts, auto loans, and student loans, as noted in Fig. 3.

Not surprisingly, interest rates differ significantly across different types of loans. In Fig. 4

we provide context regarding the distribution of interest rates for households by loan type,

again using 2016 SCF data.

Fig. 4 shows that unsecured personal loans (such as credit card loans and other consumer

loans) typically have the highest interest rates. These loans are also typically categorized as

bad debts because they are not used to purchase assets that improve the long-term financial

condition of the household and rather are used to purchase items that are more consumption-

based in nature.

Fig. 5 jointly illustrates the prevalence of different loan types and the median interest

rates among the households in which the head-of-household is 45 years old. While me-

dian interest rates are relatively static across ages, age 45 is selected as the representa-

tive age because it is the approximate peak age for indebtedness, as previously noted in

Fig. 3.

Our study explores the urgency and importance of liability management for American

households. The article consists of the following sections: First, this study utilizes SCF data

to develop a general picture of U.S. households’ financial situations in terms of their balance

sheet characteristics. Second, we review the liability side of households’ balance sheets to

investigate the prevalence of different types of consumer debts and the interest rates associ-

ated with them. Third, we analyze a number of economic and demographic factors that are

associated with household debts. After exploring the attributes that potentially relate to the

households carrying bad debts, we then identify the characteristics of households that have

higher average debt interest rates. Finally, we demonstrate the impact of liability manage-

ment in terms of investment alpha-equivalent (“excess investment return”-equivalent) analy-

sis and the potential dollar amount that can be saved through interest rate reduction relative

to financial asset considerations.

124 Z. Liu and D. M. Blanchett / Financial Services Review 29 (2021) 121–145



3. Theory

The household consumption decision involves a trade-off between consuming more today

(borrowing) and consuming more in the future (saving). The borrowing and saving behavior

of households is largely driven by their intertemporal consumption choices, affected by their

time-discounting preference, investment interest rates, and other factors.

Fig. 2. Mean value of debt for U.S. families with debt holdings. Source: Federal Reserve Board Survey of Consumer
Finances (SCF) Bulletin 2017. Notes: All respondents in the 2016 SCF data are included in this graph. The age of the house-
hold is represented by the age of the household head.

Fig. 3. Probability of a household having debt. Source: Federal Reserve Board Survey of Consumer Finances 2016 sur-
vey wave. Notes: Weights applied. Other consumer loans include loans for household appliances, furniture, hobby or recrea-
tional equipment, medical bills, friends or relatives, etc. This category does not include credit cards, margin loans, or loans
against life insurance or pensions.

Z. Liu and D. M. Blanchett / Financial Services Review 29 (2021) 121–145 125



To better analyze the liability management of U.S. households, we structure our theoretical

framework according to the life-cycle hypothesis (Jappelli & Pagano 1989; Modigliani 1986),

which holds that a household chooses a consumption path to maximize its lifetime utility

Fig. 4. Distribution of household loan interest rates. Source: Federal Reserve Board Survey of Consumer Finances
2016 survey wave. Notes: Weights applied. This figure shows the percentile distribution of interest rates across different types
of loans. Other consumer loans include loans for household appliances, furniture, hobby or recreational equipment, medical bills,
friends or relatives, etc. This category does not include credit cards, margin loans, or loans against life insurance or pensions.

Fig. 5. Loan prevalence and median interest rates. Source: Federal Reserve Board Survey of Consumer Finances (SCF)
2016 survey wave. Notes: Weights applied. This graph uses a subsample of 45-year-old respondents to illustrate the preva-
lence and median interest rates of different types of household debts in the SCF data.

126 Z. Liu and D. M. Blanchett / Financial Services Review 29 (2021) 121–145



subject to an intertemporal budget constraint. We start with a simple two-period life-cycle

model to understand the dynamic intertemporal choice issue. Then we generalize this model to

multiple periods to capture the households’ liability decisions for different life stages.

In the two-period model, a household maximizes its utility described as following:

U c1, c2ð Þ ¼ u c1ð Þ þ d u c2ð Þ (1)

where c1 and c2 are consumptions in periods 1 and 2, respectively.5 d is the discount factor

that depicts the household’s time preference. The assumption of 0 < d < 1 illustrates the

tendency that present consumption is always more preferable than future consumption. d is

more close to 0 when the household is more future-discounting. If d is close to 1, the house-

hold has no preference between present and future consumptions.

The two-period budget constraint that the household faces can be represented by the fol-

lowing inequalities:

c1 þ s≤ y1 (2)

c2 ≤ 1 þ rð Þ s þ y2 (3)

where y1 and y2 are the income of the household for period 1 and period 2, respectively. The

borrowing/saving factor is symbolized by s. If s > 0, then the household saves in period 1. If

s < 0, then this household borrows in period 1; thereby, forfeiting investment opportunities

and reducing the consumption in period 2. r represents the prevailing interest rate in the fi-

nancial markets. If s > 0, then r stands for the investment return from savings. If s < 0, then

r can represent the interest charged for the debt the household borrows during period 1.

Substituting out the borrowing/saving factor s, we obtain the “lifetime budget constraint.”

This constraint represents the fact that the discounted present value of all periods’ consump-

tion must be less than or equal to the discounted present value of lifetime income:

c1 þ C2

1 þ r
≤ y1 þ y2

1 þ r
(4)

Now the household’s intertemporal consumption choice model can be rewritten as:

Max

c1, c2f g U c1, c2ð Þ ¼ u c1ð Þ þ d u c2ð Þ (5)

s:t: c1 þ C2

1 þ r
≤ y1 þ y2

1 þ r

Using the Lagrangian technique, the solution to this problem is:

FOC c1ð Þ: u0 c1ð Þ=l (6)

Z. Liu and D. M. Blanchett / Financial Services Review 29 (2021) 121–145 127



FOC c2ð Þ: b u0 c2ð Þ= l

1 þ r
(7)

FOC lð Þ: c1 þ C2

1 þ r
≤ y1 þ y2

1 þ r
(8)

Putting the first order conditions together, we arrive at the Euler equation:

u0 c1ð Þ
d u0 c2ð Þ ¼ 1 þ rð Þ (9)

This equation describes the intertemporal optimal consumption choice between the current
and future period: The marginal rate of substitution (appropriately discounted by d ) is equal
to the gross interest rate, which represents the relative price between consumption in period
1 and consumption in period 2. In terms of saving (s > 0), if r is high, the price of consump-

tion in period 1 is high because the household is forgoing a high interest rate of investment
return. In the case of borrowing debt (s < 0), the interpretation still applies: If r is high, the

price of consumption in period 1 is high because the household is paying a high borrowing
cost due to the high interest rate. The Euler equation implies that the household maximizes

utility by smoothing the consumption path over the life cycle, which explains the borrowing
behavior of the household.

The two-period intertemporal consumption model can be generalized for multiple-period

analysis. Assume a household’s finite lifetime can be categorized into T different periods. In

each period t, the household has income yt, saves or borrows st, and consumes ct. Then the

household’s intertemporal consumption choice model is as follows:

Max

c1, c2, . . . , ctf g

E o
t=1

T
d �tþ1u ctð Þ

" #
(10)

s:t: RT
t=1 1 þ rð Þ�tþ1

ct ≤RT
t=1 1 þ rð Þ�tþ1yt (11)

where d is still the discounting factor measuring the households’ preference for present versus

future, and r is the rate of return on the investment (or interest rate of borrowing on the debt).

Similarly, one can derive the solution to this problem and arrive at the generalized Euler

equation:

Et

u
0
ctð Þ

d u0
ctþ1ð Þ

" #
¼ 1 þ rð Þ (12)

Notice that the Eq. (12) can be rearranged into:

128 Z. Liu and D. M. Blanchett / Financial Services Review 29 (2021) 121–145



Et

u
0
ctð Þ

u
0
ctþ1ð Þ

" #
¼ 1 þ rð Þ d (13)

Then we can interpret the Euler Eq. (13) as the marginal rate of substitution between the period

(t) and the next period (t þ 1), is equal to the product of the gross interest rate and the time dis-

counting factor. In other words, the households smooth their lifetime consumption paths based

on two factors, the interest rate (borrowing or investing) and their time discounting preference.

There have been some variations to the life-cycle model since its development. For exam-

ple, the presence of liquidity and borrowing constraints has been brought up to modify the

model for better suitability to empirical analysis. In our analysis, we assumed that U.S.

households are able to leverage from various lending sources to achieve their consumption

smoothing and combine the liquidity and dollar amount borrowing constraints into the inter-

est rate constraint (the household’s tolerance of high interest rates).7 We also consider

households’ liquid assets in our analysis to investigate their debt problems. To capture the

discounting preference of American families, we use the household’s financial planning ho-

rizon as a proxy in the empirical analysis.

The household consumption decision model provides guidance on what to expect in the

regression analysis results presented in this article. For instance, we expect to observe that

interest rates significantly affect household leverages across various debt types. Households

with relatively longer financial planning horizons are less likely to carry debt (or they have

lower debt amount, debt-to-income ratio, and debt-to-asset ratio) compared with the house-

holds whose financial planning horizons are short. Liquid asset holdings should significantly

reduce the household debt level. Detailed discussion on the regression results will be pre-

sented in the following sections.

4. Data and methodology

This article uses data from the SCF to analyze the characteristics of U.S. household finan-

ces. The SCF, conducted by the Federal Reserve Board, is a nationally representative cross-

sectional survey of U.S. households. This triennial survey collects a variety of information

on income, balance sheet, and demographic characteristics from a selection of more than

6,000 American families in each survey wave. Using the 2016 survey wave, we study the

characteristics of the balance sheets of American households, explore the factors that are

associated with high debt-to-asset ratios for certain households, and investigate the benefit

of liability management for these households.8

For our analysis, we focus on “low-to-affluent” American families, which we define as

households with less than $1 million in financial assets. Households with very high net worth

often have their own unique leveraging and investment strategies, and optimizing these strat-

egies is beyond the scope of this article.

The comprehensive perspective of the average household balance sheet (see Appendix A)
indicates that the average return on financial (i.e., investment) assets is approximately 62%
of the debt interest charges for the average U.S. household. In other words, the average U.S.

Z. Liu and D. M. Blanchett / Financial Services Review 29 (2021) 121–145 129



family is spending more on interest servicing household debt than they are earning from
investing their financial assets. This is despite a significant focus on managing the asset side
of the household balance sheet that is common within the financial advising profession.

The focus of this article is to explore how low-to-affluent American families can potentially

benefit from debt restructuring and liability management with assistance from their financial

planners and advisors. Because of the nature of the SCF data, which oversamples high-income

households (Aizcorbe, 2003; Nielsen 2015), we apply sample weights to all the empirical anal-

yses. In addition to focusing only on households with less than $1 million in financial assets,

we also restricted the opportunity set to households whose head is between 20 and 85 years old

and that had an annual family income of at least $1,000. After applying these restrictions, our

analysis sample is reduced to 4,481 households (see Appendix B for descriptive statistics of the

analysis sample). Because each household in the 2016 SCF data has five implicates, the total

number of observations in our analysis sample is 22,415.9 To cope with the dual-frame com-

plex sample design and the multiple-imputation process of the SCF data, this study use the

“SCFcombo” Stata macro designed by Nielsen (2015) to conduct our regression analyses.10

5. Results and discussion

The regression analyses used in this article follow these steps: First, we use probit and or-

dinary least squares (OLS) regressions to study what factors are associated with household

debt. We look at the economic, demographic, and behavioral factors that could potentially

impact the likelihood of carrying household debt, the total debt amount, the debt-to-finan-

cial-asset ratio, and the debt-to-income ratio. Second, we isolate what are frequently consid-

ered bad debts (represented by credit card debts) and compare them with debts that are

typically viewed as good debts (represented by mortgages) to see whether the factors associ-

ated with different debt categories are similar. Then, we utilize different interest rate meas-

ures to check the attributes that relate to high interest rates. Finally, we perform alpha-

equivalent analyses and calculated the potential savings to demonstrate the impact of liabil-

ity management and interest rate reduction from a financial asset perspective. Detailed

descriptions and summary statistics of the variables used in the regression analyses are pre-

sented in Appendix B.

Table 1 presents the results of the probit and OLS regressions to better understand what

factors are associated with household debts. The dependent variables in these regressions

include “whether the household carries debt,” “total debt amount,” “debt to financial asset

ratio,” and “debt to income ratio.” The marginal effect results of the probit regression in

Table 1 provide a general picture of what factors are associated with low-to-affluent

American families’ debt holdings. The OLS regression demonstrates the impact on house-

hold debt amounts from each of these factors. In some cases, relatively wealthier families

that are in good financial conditions still carry larger amount of debt due to their high

income or sizeable financial asset accumulations. While some financially challenged fami-

lies might not be carrying a sizable sum of debt in terms of dollar amounts, these debts are

typically detrimental to their financial well-being compared with their income and asset lev-

els. To consider these cases, we analyze the debt-to-financial-asset ratio and the debt-to-

130 Z. Liu and D. M. Blanchett / Financial Services Review 29 (2021) 121–145



income ratio in comparison with the analyses on the likelihood of having debts and the total

debt amount. The intertemporal model discussed in the Theory Section above suggests that

time discounting preference should affect households’ consumption smoothing behaviors

significantly. Therefore, we expect to see from the results in Table 1 that households with

longer financial planning horizons are less likely to carry debt, have lower debt-to-asset ratio

as well as lower debt-to-income ratio. In addition, we expect to see a significant negative

relationship between the households’ liquid asset levels and the likelihood of carrying debts,

total debt amount as well as debt-to-income ratio.

Based on the results in Table 1, married families and households with children are more

likely to carry debts. Families that own houses are much more likely to borrow, and the

more real assets a family owns, the more likely this family is to carry debts. Liquid assets

and age are negatively related to the likelihood of having debts. This is most likely because

households are less likely to borrow if they have enough liquid assets to cover their needs,

which supports the advocacy of emergency savings through liquid accounts for the general

public. Older families are less likely to have debts because they generally have had a longer

time to accumulate wealth and pay off their various household debts.

It appears to be counterintuitive that education and income level, as well as reporting hav-

ing savings, are positively related to carrying household debts. However, if we consider the

OLS results together with the marginal effects of the probit regression, the impact of these

factors on household debts becomes clear. For instance, although high-income families are

more likely to leverage and have larger debt sizes, their debt-to-income ratios are lower and

negatively related to their income level. Households that have savings demonstrate much

lower debt-to-financial-asset ratios, despite the higher likelihood to borrow, with other varia-

bles such as liquid asset levels controlled.

The combined results could indicate that these families may be more financially literate

and leverage lower-interest debts to increase their investments in financial assets and sav-

ings. When it comes to education level, more educated households are more likely to carry

debts, have higher debt balances, and have a higher debt-to-income ratio, keeping all other

factors, such as income and assets, the same. This is a strong indication of the impact of stu-

dent loans on these families. Ceteris paribus, educated families are more likely to carry stu-

dent loans compared with the less educated ones, because of the prevalence of student loans

used to finance education today.

A family’s financial planning horizon is also a strong behavioral indicator of household

debts. Households with longer financial planning horizons are much less likely to have

debts. Total debt amount, as well as debt-to-financial-assets ratio and debt-to-income ratio,

are all negatively associated with a longer financial planning horizon. This finding supports

the myopic planning hypothesis, which predicts that having a myopic financial planning ho-

rizon fuels households’ borrowing and may lead families deeper into debt. It also suggests

that promoting long-term financial planning horizons serves as a good approach to help fam-

ilies with their liability management.

“Not all debt is created equal,” as the saying goes. While good debts are typically defined

as those with lower interest rates that help households finance activities and purchases that

provide long-term benefits (e.g., mortgages), bad debts are usually associated with higher in-

terest rates and are used to purchase depreciating assets that do not generate long-term

Z. Liu and D. M. Blanchett / Financial Services Review 29 (2021) 121–145 131



T
ab
le

1
P
ro
b
it
an
d
o
rd
in
ar
y
le
as
t
sq
u
ar
es

(O
L
S
)
re
g
re
ss
io
n
s
o
n
fa
ct
o
rs

as
so
ci
at
ed

w
it
h
h
o
u
se
h
o
ld

d
eb
ts

V
ar
ia
b
le
s

P
ro
b
it
(M

E
)a

O
L
S

O
L
S

O
L
S

H
av
e
d
eb
t

T
o
ta
l
d
eb
t
am

o
u
n
t
($
)

D
eb
t-
to
-fi
n
an
ci
al
-a
ss
et
ra
ti
o
b

D
eb
t-
to
-i
n
co
m
e
ra
ti
o
c

M
ar
ri
ed

0
.1
1
5
*
*
(0
.0
3
9
)

�8
1
9
.6

(3
,5
7
7
.5
0
5
)

�1
1
9
.2

(1
8
6
.3
4
4
)

�0
.2
2
3
*
(0
.1
0
2
)

N
u
m
b
er

o
f
k
id
s

0
.0
7
8
7
*
*
*
(0
.0
1
9
)

5
,4
3
3
.1
*
*
*
(7
6
2
.8
2
6
)

1
5
1
.1

(1
4
2
.5
8
7
)

0
.0
1
7
2
(0
.0
2
5
)

E
d
u
ca
ti
o
n
le
v
el

0
.0
5
9
3
*
*
*
(0
.0
0
7
)

2
,8
3
0
.9
*
*
*
(6
5
8
.9
6
3
)

�6
2
.0
1
(3
3
.7
5
1
)

0
.1
0
2
*
*
*
(0
.0
2
0
)

R
ea
l
as
se
ts
(p
er

$
1
0
K
)

0
.0
1
0
9
*
*
*
(0
.0
0
2
)

4
,3
1
4
.9
*
*
*
(2
0
7
.0
3
0
)

6
.7
6
7
(8
.7
6
0
)

0
.0
4
1
6
*
*
*
(0
.0
0
3
)

L
iq
u
id

as
se
ts
(p
er

$
1
0
K
)

�0
.0
7
9
4
*
*
*
(0
.0
1
0
)

�4
,0
7
6
.7
*
*
*
(0
.0
5
1
)

7
.4
0
7
(0
.0
0
1
)

�0
.0
4
5
9
*
*
*
(0
.0
0
0
)

H
av
e
h
o
u
se
s

0
.5
8
3
*
*
*
(0
.0
5
7
)

6
,7
1
1
.6
*
*
(2
,5
0
9
.2
6
1
)

1
2
3
.3

(2
5
3
.9
7
5
)

0
.7
5
0
*
*
*
(0
.0
6
9
)

H
av
e
sa
v
in
g
s

0
.2
9
8
*
*
*
(0
.0
3
5
)

3
,1
1
4
.6

(1
,9
6
3
.9
5
6
)

�8
5
3
.9
*
*
*
(1
4
9
.4
5
9
)

0
.1
0
9
(0
.0
8
9
)

R
ac
e
B
la
ck

0
.1
2
6
*
(0
.0
5
2
)

8
,2
9
1
.1
*
*
*
(1
,8
6
3
.7
4
6
)

2
1
6
.5

(3
0
0
.6
1
0
)

0
.1
1
6
(0
.0
6
7
)

R
ac
e
H
is
p
an
ic

�0
.0
3
0
0
(0
.0
5
2
)

�5
2
5
.3

(4
,6
7
0
.2
0
3
)

�1
7
7
.6

(3
0
0
.4
3
0
)

0
.1
1
6
(0
.0
8
8
)

R
ac
e
o
th
er

�0
.0
1
7
8
(0
.0
6
2
)

4
,9
8
9
.8

(2
,6
8
3
.8
8
1
)

�5
1
1
.6
*
*
(1
8
4
.2
1
1
)

0
.4
4
3
(0
.2
2
6
)

In
co
m
e
(p
er

$
1
0
K
)

0
.0
5
7
8
*
*
*
(0
.0
1
1
)

3
,4
1
0
.2
*
*
(1
,1
2
8
.4
5
8
)

�3
5
.2
4
(1
8
.2
7
2
)

�0
.0
9
5
0
*
*
(0
.0
2
9
)

A
g
e

�0
.0
1
6
1
*
*
*
(0
.0
0
1
)

�1
,2
3
7
.7
*
*
*
(5
2
.3
0
0
)

�1
2
.9
4
*
*
*
(3
.9
0
4
)

�0
.0
2
3
8
*
*
*
(0
.0
0
2
)

F
in
an
ci
al
p
la
n
n
in
g
h
o
ri
zo
n

(o
m
it
te
d
b
as
el
in
e

ca
te
g
o
ry

“n
ex
t
fe
w

m
o
n
th
s”
)

N
ex
t
y
ea
r

�0
.1
5
0
*
*
(0
.0
5
2
)

�3
,8
3
9
.2

(2
,0
4
3
.2
4
8
)

�9
3
0
.4
*
*
(2
8
6
.5
3
1
)

0
.1
3
6
(0
.1
6
0
)

N
ex
t
fe
w
y
ea
rs

�0
.0
1
9
5
(0
.0
4
3
)

�5
4
3
9
.2
*
(2
,2
6
9
.3
3
3
)

�7
3
4
.6
*
(2
9
3
.1
3
4
)

�0
.1
2
3
*
(0
.0
5
7
)

N
ex
t
5
to

1
0
y
ea
rs

�0
.2
0
2
*
*
*
(0
.0
5
0
)

�1
1
,8
6
1
.2
*
*
*
(2
,5
3
0
.7
1
3
)

�9
6
9
.5
*
*
*
(2
6
2
.8
3
3
)

�0
.1
9
3
*
*
(0
.0
7
0
)

L
o
n
g
er

th
an

1
0
y
ea
rs

�0
.2
4
5
*
*
*
(0
.0
6
0
)

�6
,5
5
3
.9
*
(3
,1
5
6
.3
2
2
)

�6
3
0
.7
*
(3
0
6
.9
9
2
)

�0
.2
7
6
*
*
(0
.0
9
0
)

N
4
,4
8
1

4
,4
8
1

4
,4
8
1

4
,4
8
1

N
o
te
s:

a
T
h
is
co
lu
m
n
re
p
o
rt
s
th
e
av
er
ag
e
m
ar
g
in
al

ef
fe
ct

o
f
th
e
p
ro
b
it
re
g
re
ss
io
n
.
T
h
e
2
0
1
6
S
u
rv
ey

o
f
C
o
n
su
m
er

F
in
an
ce
s
(S
C
F
)
sa
m
p
le

w
ei
g
h
ts
w
er
e

ap
p
li
ed

to
th
e
re
g
re
ss
io
n
s.

b
T
h
e
an
al
y
si
s
sa
m
p
le

in
th
is

re
g
re
ss
io
n
in
cl
u
d
es

th
e
h
o
u
se
h
o
ld
s
w
h
o
se

d
eb
t
to

fi
n
an
ci
al

as
se
t
ra
ti
o
eq
u
al
s
to

ze
ro
.
F
o
r
th
e

re
d
u
ce
d
-s
iz
e
an
al
y
si
s
sa
m
p
le

w
h
ic
h
o
n
ly

in
cl
u
d
es

th
e
h
o
u
se
h
o
ld
s
w
h
o
ca
rr
y
d
eb
t,
p
le
as
e
se
e
th
e
re
d
u
ce
d
sa
m
p
le

re
g
re
ss
io
n
re
su
lt
s
in

A
p
p
en
d
ix

C
.
c
T
h
e

an
al
y
si
s
sa
m
p
le

in
th
is

re
g
re
ss
io
n
in
cl
u
d
es

th
e
h
o
u
se
h
o
ld
s
w
h
o
se

d
eb
t
to

in
co
m
e
ra
ti
o
eq
u
al
s
to

ze
ro
.
A
lt
er
n
at
iv
e
re
g
re
ss
io
n
w
it
h
th
e
re
d
u
ce
d
sa
m
p
le

w
h
er
e
o
n
ly

h
o
u
se
h
o
ld
s
w
it
h
d
eb
ts
ar
e
in
cl
u
d
ed

is
d
is
cu
ss
ed

in
A
p
p
en
d
ix

C
.
S
ta
n
d
ar
d
er
ro
rs
in

p
ar
en
th
es
es
.

*
p
<
.0
5
,
*
*
p
<
.0
1
,
*
*
*
p
<
.0
0
1
.

132 Z. Liu and D. M. Blanchett / Financial Services Review 29 (2021) 121–145



benefits. The costs associated with good debts are often outweighed by the benefits. Bad

debts, on the contrary, carry high interest rates with little or no long-term returns (Hanson,

2006). These types of debts can potentially negatively impact the borrower’s credit scores,

retirement goals, and financial health, as well as family relationships. In some circumstan-

ces, bad debts can create a vicious borrowing cycle for some families and cause stress and

mental as well as physical health problems (Davies, Montgomerie, & Wallin 2015). The

negative health effects of debt (i.e., the “high price of debt”) is a phenomenon noted both in

the United States (Sweet et al., 2013) and internationally (Clayton, Liñares-Zegarra, &

Wilson, 2015).

This article explores the different factors that are potentially related to households carry-

ing bad debts. We first investigate these potential factors by separating debt categories. (In

Table 2, we chose credit card revolving balances as a representation of bad debt and mort-

gages as an example of good debt.) Then, we utilize different interest rate measures to check

the relationships between these factors and high interest rates (Table 3).

Although liquid assets and interest rates are both predicted to be associated with house-

hold leveraging, we expect these factors to play different roles when it comes to “good

debts” versus “bad debt.” In particular, we want to test whether interest rate has more signifi-

cant negative relationships with mortgages due to the large size and long durations of the

debt, and whether liquid assets level is more significantly and negatively related to credit

card debts due to the “liquidity needs compromise.” In addition, we expect to observe nega-

tive significant relationships between household financial planning horizons and the amount

of both types of debts. The regression results in Table 2 indicate that although some house-

hold attributes are related to both good and bad debts, certain factors are particularly note-

worthy when it comes to explaining what kinds of households are more likely to carry bad

debts. Having more children is positively associated with both credit card loans and mort-

gages. However, other factors such as interest rate, real assets, liquid assets, and income

have different relationships with credit card debt compared with mortgages. For instance,

mortgages are more sensitive to interest rate changes, but credit card loans are more sensi-

tive to liquid assets and income. The reason behind this difference could be interpreted as

“liquidity needs” compromise. Credit card loans are often used to cover short-term liquidity

needs. Their insensitivity toward interest rates could be largely caused by a lack of liquid

assets to cover certain short-term needs (such as holiday shopping, etc.). Therefore, credit

card debts are negatively related to liquid asset levels. On the contrary, mortgages are nega-

tively associated with interest rates because of their relatively larger debt size (hence larger

interest payments) and longer investment horizon.

One interpretation of the income effect on credit card loans could be that, keeping every-

thing else (including liquid assets) equal, households with higher incomes have the ability and

resources to borrow—and pay back—more credit card loans. Age is another factor that is only

negatively related to mortgages. This finding indicates that older households are more likely

to have had a longer time to pay off their mortgages and hence reduce the size of this type of

good debt. Because houses are a major component of most households’ real assets, it is not

surprising that the real asset level is positively related to family mortgage loans. The financial

planning horizon factor is negatively associated with both credit card loans and mortgages in

Z. Liu and D. M. Blanchett / Financial Services Review 29 (2021) 121–145 133



Table 2. This is consistent with the previous regression results, indicating that families with

longer financial planning horizons are less likely to carry both kinds of debts.

A major focus of household liability management is to help the targeted families to reduce

the interest rates of their debts. The following analysis seeks to explore what kind of factors

are associated with higher household interest rates. We expect to see negative relationships

between the weighted average interest rate and certain household characteristics such as

real and liquid asset levels, household head education level, homeownership, savings, as

well as being married and having a longer financial planning horizon. In Table 3, we use

different measures to capture the households’ average interest rates as well as the percentile

ranking of the average interest rates. The weighted average interest rate takes into account

the dollar amount weighted average interest rates across all loan types. For example, for

each household, the dollar amount of different loans is multiplied by their interest rates to

calculate the overall liability cost per year. Then this liability cost is divided by the total

loan amount to acquire the weighted average interest rate for this household. The simple

average interest rate measure takes the arithmetic average of the interest rates across all

loan types. This measurement, together with the weighted average interest rate percentile

and simple average interest rate percentile measures, serves as a robustness check for the

weighted average interest measurement. Based on the OLS regression results from Table 3,

households with less education, lower levels of assets, fewer savings, and older age are

Table 2 Ordinary least squares (OLS) on different debt categories

Variables “Bad” debts (credit and
store cards balance)

“Good” debts (mortgages)

Interest rate �28.40 (18.556) �1,673.5* (677.040)
Married 265.1 (364.032) �4,993.6 (4,070.197)
Number of kids 275.3* (122.659) 5,148.3*** (1,183.437)
Education level 76.84 (54.142) 1,626.4* (771.444)
Real assets 0.00197 (0.001) 0.393*** (0.020)
Liquid assets �0.0328*** (0.003) �0.0817 (0.067)
Have houses 519.9 (388.969) Omitted
Have savings �419.8 (262.935) �823.7 (3,076.358)
Race black �528.6 (345.616) 7695.6 (4,149.247)
Race Hispanic �685.3* (314.295) 6,027.6 (9,231.032)
Race other �247.4 (311.224) 4,493.5 (5,760.525)
Income 0.0201** (0.007) 0.144 (0.090)
Age 13.30 (6.916) �1,009.6*** (106.106)
Financial planning horizon (omitted

baseline category “next few months”)
Next year �685.5 (386.068) 1,138.5 (4,312.770)
Next few years �861.4** (322.013) �6,064.8 (4,592.322)
Next 5 to 10 years �1,109.0*** (329.694) �9,363.6** (3,520.808)
Longer than 10 years �1,368.6** (483.781) �12,236.5** (4,396.354)

N 2,808 1,661

Notes: Not all of the respondents in our analysis sample reported the interest of different kinds of loans.

Therefore, the number of observations was reduced in the regressions above. The 2016 Survey of Consumer

Finances (SCF) sample weights were applied to the regressions. Standard errors in parentheses.

*p< .05; **p< .01; ***p< .001.

134 Z. Liu and D. M. Blanchett / Financial Services Review 29 (2021) 121–145



T
ab
le

3
O
rd
in
ar
y
le
as
t
sq
u
ar
es

(O
L
S
)
o
n
av
er
ag
e
in
te
re
st
ra
te

m
ea
su
re
s

V
ar
ia
b
le
s

W
ei
g
h
te
d
a
av
er
ag
e

in
te
re
st
ra
te

W
ei
g
h
te
d
av
er
ag
e

in
te
re
st
ra
te
p
er
ce
n
ti
le

S
im

p
le
b
av
er
ag
e

in
te
re
st
ra
te

S
im

p
le
av
er
ag
e

in
te
re
st
ra
te
p
er
ce
n
ti
le

M
ar
ri
ed

�1
.1
8
7
*
*
(0
.3
7
8
)

0
.0
0
5
2
4
(0
.0
0
9
)

�0
.7
8
9
*
(0
.3
5
0
)

0
.0
0
9
0
3
(0
.0
0
9
)

N
u
m
b
er

o
f
k
id
s

0
.0
3
3
7
(0
.1
0
2
)

0
.0
1
0
8
*
*
*
(0
.0
0
3
)

0
.1
6
4
(0
.1
0
2
)

0
.0
1
1
0
*
*
*
(0
.0
0
3
)

E
d
u
ca
ti
o
n
le
v
el

�0
.3
1
9
*
*
*
(0
.0
5
4
)

�0
.0
0
6
1
3
*
*
*
(0
.0
0
2
)

�0
.2
7
7
*
*
*
(0
.0
5
1
)

�0
.0
0
6
8
0
*
*
*
(0
.0
0
1
)

R
ea
l
as
se
ts
(p
er

$
1
0
K
)

�0
.0
3
8
1
*
*
*
(0
.0
0
7
)

�0
.0
0
1
2
8
*
*
*
(0
.0
0
0
)

�0
.0
4
3
1
*
*
*
(0
.0
0
8
)

�0
.0
0
1
4
2
*
*
*
(0
.0
0
0
)

L
iq
u
id

as
se
ts
(p
er

$
1
0
K
)

�0
.1
4
1
*
*
*
(0
.0
0
0
)

�0
.0
0
3
4
8
*
*
(0
.0
0
0
)

�0
.1
6
6
*
*
*
(0
.0
0
0
)

�0
.0
0
3
8
8
*
*
*
(0
.0
0
0
)

H
av
e
h
o
u
se
s

�0
.8
8
8
*
*
(0
.3
4
2
)

�0
.0
0
4
8
4
(0
.0
1
0
)

�0
.3
6
8
(0
.4
0
2
)

�0
.0
0
7
7
4
(0
.0
0
9
)

H
av
e
sa
v
in
g
s

�0
.8
2
6
*
(0
.3
6
4
)

�0
.0
1
6
4
*
*
(0
.0
0
6
)

�0
.8
5
1
*
(0
.3
5
4
)

�0
.0
1
7
4
*
*
(0
.0
0
5
)

R
ac
e
B
la
ck

�0
.5
0
6
(0
.3
8
4
)

0
.0
3
0
7
*
*
(0
.0
0
9
)

0
.0
1
2
7
(0
.3
3
6
)

0
.0
2
8
4
*
*
*
(0
.0
0
8
)

R
ac
e
H
is
p
an
ic

0
.7
4
9
(0
.4
1
3
)

0
.0
4
5
2
*
*
*
(0
.0
1
1
)

1
.3
3
2
*
*
(0
.5
0
9
)

0
.0
3
5
8
*
*
*
(0
.0
1
0
)

R
ac
e
o
th
er

0
.0
2
6
0
(0
.4
2
0
)

0
.0
1
2
6
(0
.0
1
1
)

0
.6
2
7
(0
.4
7
3
)

0
.0
1
6
6
(0
.0
1
1
)

In
co
m
e
(p
er

$
1
0
K
)

�0
.0
2
5
5
(0
.0
4
0
)

�0
.0
0
1
0
9
(0
.0
0
1
)

�0
.0
3
0
2
(0
.0
4
1
)

�0
.0
0
1
6
6
(0
.0
0
1
)

A
g
e

0
.0
4
6
2
*
*
*
(0
.0
1
0
)

0
.0
0
0
5
8
9
*
(0
.0
0
0
)

0
.0
2
9
7
*
*
*
(0
.0
0
9
)

0
.0
0
0
5
5
9
*
(0
.0
0
0
)

F
in
an
ci
al
p
la
n
n
in
g
h
o
ri
zo
n

(o
m
it
te
d
b
as
el
in
e
ca
te
g
o
ry

“n
ex
t
fe
w
m
o
n
th
s”
)

N
ex
t
y
ea
r

�1
.3
4
3
*
*
*
(0
.4
0
6
)

�0
.0
0
4
0
7
(0
.0
1
1
)

�0
.3
9
6
(0
.4
1
6
)

�0
.0
1
0
9
(0
.0
1
0
)

N
ex
t
fe
w
y
ea
rs

�0
.5
0
5
(0
.5
0
7
)

�0
.0
1
8
8
*
(0
.0
0
9
)

�0
.3
3
5
(0
.4
4
7
)

�0
.0
1
2
8
(0
.0
0
8
)

N
ex
t
5
to

1
0
y
ea
rs

�1
.3
5
9
*
*
*
(0
.3
4
6
)

�0
.0
1
8
5
(0
.0
1
1
)

�1
.1
8
8
*
*
*
(0
.2
5
3
)

�0
.0
1
9
4
*
(0
.0
0
9
)

L
o
n
g
er

th
an

1
0
y
ea
rs

�0
.6
9
2
(0
.4
3
6
)

�0
.0
1
1
1
(0
.0
1
2
)

�0
.4
7
7
(0
.3
5
6
)

�0
.0
0
9
2
4
(0
.0
1
1
)

N
3
,3
9
8

3
,3
9
8

3
,3
9
8

3
,3
9
8

N
o
te
s:

In
te
re
st
in
fo
rm

at
io
n
fo
r
so
m
e
lo
an
s
w
as

n
o
t
re
p
o
rt
ed

in
th
e
2
0
1
6
S
u
rv
ey

o
f
C
o
n
su
m
er

F
in
an
ce
s
(S
C
F
)
d
at
a,

th
er
ef
o
re

th
e
to
ta
l
n
u
m
b
er

o
f
h
o
u
se
-

h
o
ld
s
w
as

re
d
u
ce
d
fr
o
m

4
,4
8
1
to

3
,3
9
8
.
T
h
e
2
0
1
6
S
C
F
sa
m
p
le
w
ei
g
h
ts
w
er
e
ap
p
li
ed
.

a
“W

ei
g
h
te
d
”
m
ea
n
s
th
is
in
te
re
st
ra
te
m
ea
su
re

ta
k
es

th
e
d
o
ll
ar

am
o
u
n
t
w
ei
g
h
te
d
av
er
ag
e
o
f
th
e
in
te
re
st
ra
te
s
ac
ro
ss

al
l
lo
an

ty
p
es

in
to

ac
co
u
n
t.
T
h
at
is
,
fo
r

ea
ch

h
o
u
se
h
o
ld
,
th
e
d
o
ll
ar

am
o
u
n
ts
o
f
d
if
fe
re
n
t
lo
an
s
ar
e
m
u
lt
ip
li
ed

b
y
th
ei
r
in
te
re
st
ra
te
s
to

ca
lc
u
la
te

th
e
o
v
er
al
l
li
ab
il
it
y
co
st
p
er

y
ea
r.
T
h
en

th
is
li
ab
il
it
y

co
st
is
d
iv
id
ed

b
y
th
e
to
ta
l
lo
an

am
o
u
n
t
to

ac
q
u
ir
e
th
e
w
ei
g
h
te
d
av
er
ag
e
in
te
re
st
ra
te
fo
r
ea
ch

h
o
u
se
h
o
ld
.

b
™
S
im

p
le
”
m
ea
n
s
th
is
in
te
re
st
m
ea
su
re

is
b
as
ed

o
n
th
e
si
m
p
le

ar
it
h
m
et
ic

av
er
ag
e
o
f
th
e
in
te
re
st
ra
te
s
ac
ro
ss

al
l
lo
an

ty
p
es
.
T
h
is
m
ea
su
re
m
en
t
se
rv
es

as
a

ro
b
u
st
ch
ec
k
fo
r
th
e
w
ei
g
h
te
d
av
er
ag
e
in
te
re
st
m
ea
su
re
.
S
ta
n
d
ar
d
er
ro
rs
in

p
ar
en
th
es
es
.

*
p
<
.0
5
;
*
*
p
<
.0
1
;
*
*
*
p
<
.0
0
1
.

Z. Liu and D. M. Blanchett / Financial Services Review 29 (2021) 121–145 135



subject to higher interest rates. Therefore, families with these attributes are more likely to

need help with liability management and could potentially benefit significantly from inter-

est rate reductions.

Finally, we perform the alpha-equivalent analysis to determine the potential savings a

household would experience if it were able to reduce the interest rates on their existing

liabilities. For the analysis we assume the household’s interest rates are reduced based on

the distribution of household loan interest rates as noted in Fig. 3 We assume each liability

would be reduced by some percentile amount, based on the distribution for that respective

liability.

For example, let us assume a household had financial assets (i.e., a portfolio) worth

$100,000 and a single liability, which was $15,000 in credit card debt at an interest rate of

15%. A 15% interest rate on credit card debt would be in the 47th percentile of interest rates

according to Fig. 4 If the household were able to reduce the interest rate by ten percentile

points, to the 37th percentile, the interest rate would decline to approximately 13%. This

results in an interest savings of 2% (15% to 13% = 2%) that would translate into $300 of

total savings on the $15,000 total credit card debt ($15,000*2% = $300). If we divide the

estimated $300 in annual interest savings by the total financial assets, we can estimate the

“alpha-equivalent” benefit associated with liability optimization, which would be 30 bps (ba-

sis points) in this case ($300/$100,000 = 30 bps).

We conduct this analysis for all households, where the rate on each loan is assumed to be

reduced by some percentile level, based on the distribution of loan rates in Fig. 3 For the

analysis the lowest possible rate is the 1st percentile. Information about the distribution of

potential dollar savings and alpha-equivalent benefit are included in Fig. 6 in Panels A and

B, respectively.

The potential savings associated with improving loan rates can be significant, especially

for households that have higher interest percentiles. If a household’s weighted average debt

interest rate is currently in the 95th percentile, a five-percentile drop could generate 113.5%

equivalent alpha, or $1,641 in annual savings. If these households achieve a 10-percentile

reduction in loan rates, the total savings would be $2,614, which is equivalent to 237.5% of

investment alpha. Even the median household stands to benefit from even modest improve-

ments. For example, the median households would on average save $410 if they were able

to reduce their weighted average loan rates by 10 percentile points, which is equivalent to a

195 bps of investment alpha. This suggests that, for many households, making efforts to

reduce the interest rates on their liabilities is more likely to result in wealth gains than

attempting to construct portfolios that might outperform the market.

Notice that when calculating the potential savings on interest rate reductions, we use the

weighted average interest rate in the discussion. Lowering the household average interest

rate may be achieved in two different ways. First, households can make more efforts on in-

terest rate shopping and negotiate lower interest rate on their loans, if possible. Second, even

if directly lowering interest rates is not feasible, the weighted average interest rate can still

be reduced through debt restructuring. Households can substitute a higher interest loan with

lower interest borrowings to achieve the reductions of overall weighted average interest

rates. (For example, consider a household with a large revolving balance on credit card loans

who cannot reduce the total amount of household debt. This household could still potentially

136 Z. Liu and D. M. Blanchett / Financial Services Review 29 (2021) 121–145



pay off this high interest rate credit card loan with low-interest secured-personal loans or

some other type of loan. By doing so, the average interest rate of this household could be

reduced, potentially significantly.)

The analyses above reveal the significant potential benefits of liability management

and point out the characteristics that are associated with different households’ debt prob-

lems. Financial planning practitioners and financial institutions can benefit from this

research not only by recognizing the potential benefits of liability management for low-

to-affluent American families, but also by identifying the attributes associated with those

Fig. 6. Benefit of reducing interest rates on debt. Source: Federal Reserve Board Survey of Consumer Finances (SCF)
2016 survey wave. Notes: The subsample is restricted to households that carry loans, reported complete data on all loan types,
and have more than $1 in financial assets. The number of observations is 3,371. The 2016 SCF sample weights were applied.

Z. Liu and D. M. Blanchett / Financial Services Review 29 (2021) 121–145 137



households that most need debt assistance. This study can also encourage consumers to

seek for an integrated approach to making decisions about their marginal income and

benefit significantly from analyzing both sides of their balance sheet extensively and

regularly.11

6. Conclusion and implications

Debt is a significant and growing component of U.S. household balance sheets. With total

interest rate payments on loans exceeding the expected returns on household financial assets

for the average household, the impact of liability optimization should draw more focus from

financial advisors, financial firms, and consumers. In this study, we first reviewed American

families’ current financial outlook by looking at their debt situations. Using the SCF data,

we then analyzed the different economic, demographic, and behavioral factors that are asso-

ciated with household borrowing and leverage ratios. Next, we separated the good and bad

debts and investigated whether the attributes related to different debt categories are similar.

After checking the characteristics demonstrated by the households that carry high-interest

debts, we performed alpha-equivalent analyses to calculate the potential benefits of liability

management.

Our study indicates that households with lower assets, income, and education levels

need assistance the most and could significantly benefit from debt management.

Households’ time discounting preferences also play an important role in their borrowing

decisions. Families with longer financial planning horizons are less likely to carry loans.

Among the borrowers, a shorter financial planning horizon is usually an indicator of a

higher debt amount as well as higher debt-to-asset and debt-to-income ratios. Families

with myopic planning horizons are also more likely to carry a higher amount of bad debts,

such as credit card balances.

This study can also inspire advisors and financial services firms to consider alternative

approaches to helping consumers improve their financial well-being. For example, advisors

could help their clients design a road map for debt restructuring and interest rate reduction

along with building portfolio investment strategies. By reviewing both sides of the house-

hold balance sheet extensively and periodically, advisors can integrate both investment and

liability management strategies to better improve their clients’ economic outlooks. These

strategies would be particularly effective for households with lower income, education, and

asset levels.

Large retirement firms could explore the possibility of building a bridge between their

retirement plan participants and lending institutions to help their participants gain access to

loans with competitive rates. Participants could utilize these lower “group rate” loans to

restructure and reduce the interest payments on their existing debts. Financial planners could

also implement different behavior coaching strategies (such as behavioral nudging devices)

to help their clients increase their financial planning horizons and avoid the consequences of

myopic planning.

138 Z. Liu and D. M. Blanchett / Financial Services Review 29 (2021) 121–145



The integration of investment and liability management strategies prompts financial advi-

sors to help their clients to answer the question, “Where should my next dollar go?” By

designing a universal comparison mechanism between investing and paying off debt, finan-

cial advisors can help their customers to better manage their marginal income. An integrated

model or strategy can be designed to not only educate the consumers on the importance of

liability management, but also guide their decision-making process after taking each con-

sumer’s unique financial situation into account. Future studies may find it favorable to build

such an integrated methodology to help answer the age-old invest or pay off debt conundrum

faced by many households.

Notes

1 Defined as households with a net worth not exceeding $1 million, have more than

$1,000 annual income and have at least $1 in financial assets. High net worth house-

holds, defined as those with net worth over $1 million, often have their own unique

leveraging and investment strategies, and optimizing these special strategies is

beyond the scope of this paper. Our definition of “low-to-affluent” households

includes those in the middle-to-low income range because these households are

most likely to need debt management assistance. Detailed descriptions of the analy-

sis sample can be found in the data and methodology section of this article.

2 This is the lower end of average credit card and retail store installment card interest

rates. Source: 2016 SCF data weighted average credit card interest rate for low-to-

affluent households.

3 The definition of “good” and “bad” debts is discussed in both the literature review

section and the results section.

4 The Consumer Finance Monthly study is conducted by the Consumer Finance

Research Group at Ohio State University.

5 The utility function satisfies monotonicity (more is preferred to less) and concavity

(diminishing marginal utility) properties and assumes ct’s are normal goods for ev-

ery period t. The concavity property implies the preference of smoothing consump-

tion across time because of the love of diversity.

6 The budget constraint depicted by Inequality (14) is derived from the following constraints

while substituting out the borrowing factor st (8t from1 to T):

Ct þ St ≤ yt (11.1)

Ctþ1 ≤ ð1þ rÞ st þ ytþ1 (11.2)

7 Given the prevalence of “payday lending” and other short-term loans in the United

States (Caskey, 2001; Stegman, 2007), we assume that American households have

access to sufficient amount of lending sources despite the fact that some of the loans

may have unreasonably high interest rates. While we do not recommend consumers

access these short-term loans, we use their potential access abilities of these loans to

Z. Liu and D. M. Blanchett / Financial Services Review 29 (2021) 121–145 139



simplify the model and transform the borrowing constraints to interest rate con-

straints. Another reason why we do not restrict the borrowing/saving factor s in

the intertemporal consumption model is that this factor is canceled out when com-

bining the two-period budget constraints together using substitutional method and

Lagrangian technique to solve this intertemporal optimization problem.

8 The most current wave available at the time of the analysis.

9 The Survey of Consumer Finances uses “multiple imputation technique” to account

for missing data. Because each missing value in the SCF is imputed five times, each

SCF family has five separate observations (called “implicates”) in the final data.

10 The SCF data are derived from a dual-frame sample design, with one frame includ-

ing households chosen via an area probability sample and the second frame includ-

ing households selected from a list provided by the Internal Revenue Service. The

second selection frame has introduced the problem of oversampling wealthy families

(Nielsen, 2015).

11 Liability optimization includes debt restructuring, loan reduction, interest rate opti-

mization, behavior coaching, etc. There are numerous complexities associated with

liability optimization at the individual household level. The objective of this article

is not focused on the detailed liability optimization approaches, rather to better

understand which types of households have higher debts, in particular bad debts, and

the potential benefits associated with reducing the interest on those debts.

140 Z. Liu and D. M. Blanchett / Financial Services Review 29 (2021) 121–145



Appendix A

Sample balance sheet for the weighted mean value of the 2016 Survey of Consumer Finances households

Assets

Category Sub-
category

Sub-category detail Amount Sum total Percent of
population

Interest rate
(estimateda)

Total
annual
earnings

Financial assets $73,122.13 98.29%
Transaction accounts (liquid) $13,590.40 97.75% 0.20% $27.18
CDs $1,485.40 4.84% 0.96% $14.26
Pooled investment funds $4,569.21 5.66% 5.61% $256.52
Savings bonds $351.91 7.33% 2.62% $9.22
Directly held stocks $2,791.27 9.20% 7.53% $210.13
Directly held bonds $273.57 0.38% 3.70% $10.12
Cash value of whole life insurance $2,746.16 17.23% 2.20% $60.42
Other managed assets: $3,725.21 3.35% 4.10% $152.73

Annuities $2,861.44
Trusts $863.77

Quasi-liquid retirement accounts $42,112.74 48.04% 4.00% $1,684.51
Other misc. financial assets $1,476.26 8.29% 4.00% $59.05

Nonfinancial assets $159,244.50 89.76%
Vehicles (RVs, planes, boats, etc.) $17,982.83 84.47%
Primary residence $118,573.20 59.54%
Residential property excluding

primary residence
$11,071.96 9.12%

Net equity in nonresidential real
estate

$2,780.49 4.22%

Businesses $7,812.36 9.23%
Other misc. nonfinancial assets $1,023.67 4.92%

Total assets $232,366.60 99.32%
Net worth $156,181.85
Total investment assets $73,122.12 $2,484.13
Total financial assets less total debt $(3,062.63)
Mortgages (including home equity loans, HELOCs)

Debt secured by primary
residence:

$53,249.90 40.67%

Mortgages and home
equity loans secured
by primary residence

$51,818.03 39.27% 4.51% $2,336.99

Home equity lines of
credit secured by
primary residence

$1,431.87 3.59% 5.81% $83.19

Debt secured by other residential
property

$3,875.32 3.87% 5.45% $211.20

Other lines of credit (not secured
by residential real estate)

$143.90 1.80% 6.00% $8.63

Credit card balances after last payment $2,581.29 47.79% 15.09% $389.52
Installment loans $15,794.03 53.50% 16.27%

Education loans $8,390.03 24.69% 5.92% $496.69
Vehicle loans $5,764.13 35.36% 6.63% $382.39
Other installment loans $1,639.87 12.51% 6.00% $98.39

(continued on next page)

Z. Liu and D. M. Blanchett / Financial Services Review 29 (2021) 121–145 141



Appendix A (Continued)

Assets

Category Sub-
category

Sub-category detail Amount Sum total Percent of
population

Interest rate
(estimateda)

Total
annual
earnings

Other debt (e.g., loans against pensions or
life insurance, margin loans)

$540.31 5.35% 6.00% $32.42

Total debt $76,184.75 79.24% $4,039.43
Total asset return less total interest charges $(1,547.19)
Financial asset to debt ratio 0.960

Notes: Sample weights applied. Number of households: 4,481; Net worth < $1million; Income > $1,000; Age:

20–85.
a Interest rate estimation sources:

CDs: FRED, Federal Reserve Bank of St. Louis. Averaged since 2008.

Pooled investment fund: Assumes 50% stocks and 50% bonds. Uses the average for mutual fund return.

Savings bonds: US Department of the Treasury, 10-Year High Quality Market (HQM) Corporate Bond

Spot Rate [HQMCB10YR].

Directly held stocks: S&P 500 Return Calculator, with Dividend Reinvestment. (2019). Retrieved March

22, 2019.

Directly held bonds: US Department of the Treasury, 10-Year High Quality Market (HQM) Corporate

Bond Spot Rate [HQMCB10YR]. Retrieved March 22, 2019, from FRED, Federal Reserve Bank of St. Louis.

142 Z. Liu and D. M. Blanchett / Financial Services Review 29 (2021) 121–145



Appendix B
Descriptive statistics of the analysis sample

Variable Definition/explanations Mean SD Min. Max.

Have debt (yes = 1, no = 0) 0.7924 0.4056 0 1
Debt amount Dollar amount of total debt $76,185 $117,866 $0 $2,630,000
Married (yes = 1, no = 0) 0.5433 0.4981 0 1
Number of children in

household
Total number of children in

the household
0.7989 1.1334 0 7

Education level Highest level of education
completed according to
the SCF standard
categoriesa

9.2774 2.7128 0 14

Real assets Total value of real assetsb $150,409 $185,257 $0 $2,282,900
Liquid assets All types of transaction

accountsc
$13,590 $32,926 $0 $572,000

Leverage ratio Total debt/total asset 12.8096 462.9600 0 25,750
Own houses (yes = 1, no = 0) 0.5954 0.4908 0 1
Have savings Have more than $0 in

savings? (yes = 1, no = 0)
0.5043 0.5000 0 1

Race black (yes = 1, no = 0) 0.1637 0.3700 0 1
Race Hispanic (yes = 1, no = 0) 0.1140 0.3178 0 1
Race other (yes = 1, no = 0) 0.1073 0.3095 0 1
Income Household income in previ-

ous calendar year
$62,321 $59,020 $1,013 $2,531,591

Age Age of the household head 49.6815 16.5849 20 85
Financial planning horizon categorical variablesd

Next year (yes = 1, no = 0) 0.1551 0.3620 0 1
Next few years (yes = 1, no = 0) 0.2817 0.4498 0 1
Next 5 to 10 years (yes = 1, no = 0) 0.2186 0.4133 0 1
Longer than 10 years (yes = 1, no = 0) 0.1059 0.3077 0 1

Notes: Sample size is 4,481 households. Sample weights applied.
a 2016 Survey of Consumer Finances (SCF) codebook education level standard categories:

1. 1st, 2nd, 3rd, or 4th grade.

2. 5th or 6th grade.

3. 7th and 8th grade.

4. 9th grade.

5. 10th grade.

6. 11th grade.

7. 12th grade, no diploma.

8. High school graduate - high school diploma or equivalent.

9. Some college but no degree.

10. Associate degree in college - occupation/vocation program.

11. Associate degree in college - academic program.

12. Bachelor’s degree (e.g., BA, AB, BS).

13. Master’s degree (e.g., MA, MS, MENG, MED, MSW, MBA).

14. Professional school degree (e.g., MD, DDS, DVM, LLB, JD) and Doctorate degree (e.g., PHD, EDD).
b Real assets, according to the SCF Bulletin category definition, include: Houses, vehicles, residential proper-

ties excluding primary residence (e.g., vacation homes), and net equity in non-residential real estate.
c Liquid assets, according to the SCF Bulletin category definition, include: Money market accounts, checking

accounts, savings accounts, call accounts, and prepaid cards.
d Original Survey Question from SCF codebook: “In planning or budgeting your (family’s) saving and spend-

ing, which of the time periods listed on this page is most important to you (and your family living here)?”

Z. Liu and D. M. Blanchett / Financial Services Review 29 (2021) 121–145 143



References

Aizcorbe, A. M., Kennickell, A. B., & Moore, K. B. (2003). Recent changes in US family finances: Evidence

from the 1998 and 2001 Survey of Consumer Finances. Federal Reserve Bulletin, 89, 1.

Barba, A., & Pivetti, M. (2009). Rising household debt: Its causes and macroeconomic implications—a long-pe-

riod analysis. Cambridge Journal of Economics, 33, 113-137.

Bricker, J., Dettling, L. J., Henriques, A., Hsu, J. W., Jacobs, L., Moore, K. B., . . . Windle, R. A. (2017).

Changes in US family finances from 2013 to 2016: Evidence from the Survey of Consumer Finances. Federal

Reserve Bulletin, 103, 1.

Caskey, J. (2001). Payday lending. Journal of Financial Counseling and Planning, 12. (available at https://ssrn.

com/abstract=2443130)

Clayton, M., Liñares-Zegarra, J., & Wilson, J. O. (2015). Does debt affect health? Cross country evidence on the

debt-health nexus. Social science & Medicine, 130, 51-58.

Davies, W., Montgomerie, J., & Wallin, S. (2015). Financial Melancholia: Mental Health and Indebtedness.

London: Political Economy Research Centre.

Emmons, W. R., & Noeth, B. J. (2013). Economic vulnerability and financial fragility. Federal Reserve Bank of

St. Louis Review, 95, 361-388.

Appendix C

Ordinary least squares (OLS) regressions on factors associated with debt ratios (reduced sample)

Variables OLS OLS OLS
Debt-to-financial-asset ratioa Debt-to-income ratiob Debt-to-financial-asset ratioc

Married �215.9 (244.237) �0.357* (0.144) 21.86 (24.820)
Number of kids 168.7 (166.241) �0.0148 (0.029) 6.673 (6.939)
Education level �117.3** (44.613) 0.105*** (0.027) 1.151 (4.360)
Real assets (per $10K) 8.508 (9.951) 0.0435*** (0.004) 0.342 (0.526)
Liquid assets (per $10K) 13.88 (9.462) �0.0343** (0.013) �3.307** (1.085)
Have houses �109.2 (323.537) 0.770*** (0.084) 60.14*** (16.643)
Have savings �1142.8*** (198.107) 0.0345 (0.109) �69.84*** (19.396)
Race Black 314.1 (373.679) 0.113 (0.077) 65.24 (54.015)
Race Hispanic �168.1 (385.584) 0.108 (0.099) 19.05 (25.444)
Race other �662.5** (237.327) 0.576* (0.275) �1.861 (10.477)
Income (per $10K) �43.01 (22.350) �0.120** (0.040) �4.335 (2.696)
Age �13.59** (4.997) �0.0239*** (0.002) �0.517 (0.292)
Financial planning horizon (omitted baseline category “next few months”)

Next year �1184.4** (379.501) 0.234 (0.198) �48.46** (16.202)
Next few years �917.7* (371.363) �0.150* (0.071) �4.372 (29.915)
Next 5 to 10 years �1196.6*** (339.037) �0.150 (0.086) �52.28** (16.884)
Longer than 10 years �752.0 (389.874) �0.241* (0.120) �54.19** (16.834)

N 3,561 3,561 3,495

Notes: a The analysis sample in this regression has been reduced and does not include the households whose

total debt amount equals to zero. b The analysis sample in this regression only include the households whose

total debt amount is greater than zero. c The analysis sample in this regression only include the households

whose total debt amount is greater than zero. The analysis sample is further reduced by eliminating the house-

holds whose total financial assets is less than $10.

144 Z. Liu and D. M. Blanchett / Financial Services Review 29 (2021) 121–145



Federal Reserve Bank of New York Center for Microeconomic Data. (2020). Quarterly Report on Household

Debt and Credit, 2020:Q2. (available at https://www.newyorkfed.org/microeconomics/hhdc.html).

Gustman, A. L., Steinmeier, T. L., & Tabatabai, N. (2011). How Did the Recession of 2007-2009 Affect the

Wealth and Retirement of the Near Retirement Age Population in the Health and Retirement Study? (No.

w17547). National Bureau of Economic Research.

Hanson, J. (2006). Good Debt, Bad Debt: Knowing the Difference Can Save Your Financial Life. New York,

NY: Portfolio Penguin.

Jappelli, T., & Pagano, M. (1989). Consumption and capital market imperfections: An international comparison.

The American Economic Review, 79, 1088-1105.

Jiang, S. S., & Dunn, L. F. (2013). New evidence on credit card borrowing and repayment patterns. Economic

Inquiry, 51, 394-407.

Kim, H., & DeVaney, S. A. (2001). The determinants of outstanding balances among credit card revolvers.

Financial Counseling and Planning, 12, 67-77.

Lee, Y. G., Lown, J. M., & Sharpe, D. L. (2007). Predictors of holding consumer and mortgage debt among older

Americans. Journal of Family and Economic Issues, 28, 305-320.

Modigliani, F. (1986). Life cycle, individual thrift, and the wealth of nations. Science, 234, 704-712.

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.

Stango, V., & Zinman, J. (2016). Borrowing high versus borrowing higher: Price dispersion and shopping behav-

ior in the US credit card market. Review of Financial Studies, 29, 979-1006.

Stegman, M. A. (2007). Payday lending. Journal of Economic Perspectives, 21, 169-190.

Sweet, E., Nandi, A., Adam, E. K., & McDade, T. W. (2013). The high price of debt: Household financial debt

and its impact on mental and physical health. Social Science & Medicine, 91, 94-100.

Wärneryd, K. E. (1999). The Psychology of Saving. A Study on Economic Psychology. Cheltenham: Edward

Elgar Publishing.

Zinman, J. (2015). Household debt: Facts, puzzles, theories, and policies. Economics, 7, 251-276.

Z. Liu and D. M. Blanchett / Financial Services Review 29 (2021) 121–145 145