Transcript NTU presentation

Health and Financial Strain:
Evidence from the Survey of Consumer Finances
Angela Lyons
University of Illinois at Urbana-Champaign
Tansel Yilmazer
Purdue University
National Taiwan University
November 2006
The Motivation
(Recent Financial Trends in the U.S.)
• Uncertain economy and higher unemployment
• Rise in bankruptcies and delinquencies
• Large debt burdens from the 1990s
• Rising health care costs
The Research Question:
What is the impact of financial strain on health?
Previous Research
• Strong positive relationship between health and
socioeconomic status (SES).
• However, little consensus on the direction of
causality.
• Is poor health both a cause and a consequence of
socioeconomic status (SES)?
On the one hand….
Some studies find that poor health affects SES.
• Individuals who are in poor health work fewer
hours or are unemployed, limiting abiltity to
accumulate income and wealth.
• Serious health conditions have a larger effect on
SES than less serious conditions.
• Smith and Kington 1997; Zagorsky, 1999; Wu 2003
On the other hand….
Studies find that lower SES affects health.
•
Individuals health can be affected in 2 ways:
1. Financial problems creates physical or psychosocial
stress which affects health
2. Limited access to quality health care services and
preventative care
•
Caplovitz 1974; Smith 1998, 1999; Roberts et al. 1999;
Drentea and Lavrakas 2000; Meer, Miller, and Rosen 2003
Also, note….
Focus on income and wealth
• Smith and Kington (1997)
• Adams et al. (2003)
• Zagorsky (1999)
Focus on liability holdings and financial stress
• Drentea and Lavrakas (2000)
• Roberts et al. (1999)
Specific Studies
• Smith and Kington (1997)
– Health and Retirement Survey (HRS) and the Asset and Health Dynamics
among the Oldest Old (AHEAD).
– Find direction of causality primarily from health to SES.
• Adams et al. (2003)
– Panel data from AHEAD; distinguish between acute, chronic, and mental
health conditions; control for existing health conditions.
– Find some evidence that wealth increases incidence of some mental and
chronic conditions.
– But in general reject hypothesis that SES results in health problems.
• Meer, Miller, and Rosen (2003)
– Use PSID to examine changes in wealth and health.
– Control for endogeneity of SES using IV that controls for changes in
wealth (receipt of an inheritance).
– The effect of wealth on health becomes insignificant when endogeneity
of wealth is taken into account.
Contributions of this study to the literature:
1.
Moves beyond income and wealth and focuses on
the relative financial position of the household.
2.
Controls for the possible endogeneity between
health and financial burden.
3.
Uses a representative sample of the U.S.
population.
Description of the Data
Data from the 1995, 1998, and 2001 Survey of Consumer Finances
Features of the SCF:
• Cross-sectional survey that collects data every three years.
• Detailed info on financial holdings, income and demographics.
• Includes a self-reported measure of health status.
Households are identified as “financially strained” if
• Delinquent on any type of loan payment by two months or more
• Total assets/total debts < 1.0
• Liquid assets/disposable income < 0.25
Table 1
Demographic Statistics by Financial Strain and Health Status
_________________________________________________________________________________
Financial Strain____________
_Health
Delinq
Assets/debts<1.
Liq/inc<0.25
H
PH
No. of obs.
FS=1 FS=0
(552) (12,250)
FS=1 FS=0
(739) (12,063)
FS=1 FS=0
(4,065) (8,737)
H=0 H=1
(10,281) (2,521)
_________________________________________________________________________________
Poor health
32.4 23.8
27.0 24.0
31.4 19.4
-.-
-.-
Measures of Financial Strain
% delinquent
100.0 0.0
20.1 4.4
9.6 2.7
4.9 7.4
% (assets/debts) < 1.0
26.4 6.1
100.0 0.0
15.7 1.5
7.0 8.1
% (liq assets/inc) < 0.25
70.7 38.7
87.8 36.7
100.0 0.0
36.6 52.3
_________________________________________________________________________________
For each measure of financial strain, FS=1 indicates the household is financially strained and
FS=0 indicates the household is not financially strained. H represents household heads who are not
in poor health and PH represents household heads who are in poor health.
Summary of Descriptive Statistics
• Financially-strained households are significantly
more likely to be in poor health.
• Those who are financially strained by one measure
are more likely to be financially strained by other
measures.
• With respect to reverse causality, those in poor health
are more likely to be financially strained.
• HOWEVER, it is likely that health status plays a
more important role in explaining why some
households are under financial strain than vice versa.
Empirical Framework
Simultaneous two-equation probit models:
FS*i  α1H *i   1X1i  ε1i
H *i  α 2 FS*i   2 X 2i  ε 2i
where
FSi* = the degree to which the household is under financial strain
Hi* = the degree to which the head of the household is in poor
health
Probability of Financial Strain
X1i includes:
• Financial factors: income of head, liquid assets, other assets
• Demographics: head’s age, education, marital status, gender,
ethnicity, employment status, number of children, whether
household receives welfare, whether household has private
health insurance coverage
• Identification: whether household experienced negative
income shock in past year that was unrelated to health;
household’s attitudes, preferences, or values for borrowing
specific consumption goods
Probability of Poor Health
X2i includes:
• Same financial and demographic factors as X1i
• Identification: whether head currently smokes (health
behaviors), whether household expects major medical
expenses in the next 5-10 years (expectations), whether head’s
father is still living (biological)
Testing the Overidentifying Restrictions
(Hausman 1983, p. 444; Johnson and Skinner 1986, p. 465)
• Each structural equation was estimated with and without the
excluded variables from the other equation.
• Null hypothesis: Addition of excluded variables should have
little effect on explanatory power of the equation.
• Use likelihood-ratio tests.
• Tests reveal that overidentifying restrictions have not been
seriously violated.
Table 2
Two-Stage Probit Models: Effect of Poor Health on Probability of Financial Strain (N=12,802)
________________________________________________________________________________________________________
Delinquent
Assets/Debts < 1.0
Liq Assets/Income < 0.25
Variable
Coeff
SE
Coeff
SE
Coeff.
SE
Predicted value: Poor health
log (Income)
log (Liquid assets)
log (Other assets)
Age
Education (years)
Female
Black
Number of children
Divorced/Separated
Single
Widowed
Retired
Self-employed
Receives welfare
Private health insurance
Negative income shock
All right to borrow for vacation
All right to borrow when income cut
All right to borrow for fur/jewelry
All right to borrow for car
All right to borrow for education
Year 1998
Year 2001
Constant
0.742
-0.009
-0.044
0.031
-0.020
0.041
-0.002
0.092
0.084
0.155
0.052
0.029
-0.756
-0.050
-0.426
0.066
0.249
0.083
0.130
0.048
0.129
-0.019
0.052
0.016
-0.835
(0.146)***
(0.030)
(0.012)***
(0.009)***
(0.004)***
(0.013)***
(0.085)
(0.078)
(0.022)***
(0.084)*
(0.091)
(0.133)
(0.112)***
(0.069)
(0.124)***
(0.071)
(0.068)***
(0.074)
(0.051)***
(0.093)
(0.065)**
(0.074)
(0.056)
(0.053)
(0.286)***
0.324
-0.156
-.----.----0.033
0.037
0.110
0.013
-0.051
0.170
0.042
0.094
-0.107
-0.319
-0.038
-0.250
0.051
0.077
0.099
0.186
-0.026
0.094
0.119
0.058
1.301
(0.117)***
(0.031)***
(-.----)
(-.----)
(0.003)***
(0.013)***
(0.073)
(0.063)
(0.023)**
(0.084)**
(0.079)
(0.129)
(0.112)
(0.080)***
(0.101)
(0.058)***
(0.058)
(0.057)
(0.049)**
(0.084)**
(0.060)
(0.066)
(0.054)**
(0.060)
(0.311)***
0.293
-.----.----0.082
-0.029
-0.080
-0.005
0.014
0.036
0.119
-0.130
0.178
-0.170
0.033
0.118
-0.593
-0.002
0.027
0.058
0.106
0.096
-0.179
-0.136
-0.098
3.744
(0.088)***
(-.----)
(-.----)
(0.006)***
(0.002)***
(0.009)***
(0.058)
(0.048)
(0.016)***
(0.056)**
(0.051)**
(0.077)***
(0.048)***
(0.039)
(0.089)
(0.043)***
(0.039)
(0.039)
(0.032)**
(0.057)**
(0.042)**
(0.036)***
(0.031)***
(0.029)***
(0.115)***
_________________________________________________________________________________________________________________
Table 3
Two-Stage Probit Models: Effect of Financial Strain on Probability of Poor Health (N=12,802)
__________________________________________________________________________________________________________
Probability of Poor Health
Variable
Coeff
SE
Coeff
SE
Coeff.
SE
Pred value: Delinquent
0.114
(0.115)
-.---(-.----)
-.---(-.----)
Pred value: Assets/Debts < 1.0
-.---(-.----)
0.020
(0.195)
-.---(-.----)
Pred value: Liq Assets/Inc < 0.25
-.---(-.----)
-.---(-.----)
0.142
(0.185)
log (Income)
-0.061
(0.020)***
-0.131
(0.042)***
-.---(-.----)
log (Liquid assets)
-0.033
(0.011)***
-.---(-.----)
-.---(-.----)
log (Other assets)
-0.016
(0.006)***
-.---(-.----)
-0.023
(0.018)
Age
0.018
(0.002)***
0.016
(0.005)***
0.018
(0.005)***
Education (years)
-0.060
(0.005)***
-0.069
(0.006)***
-0.066
(0.019)***
Female
-0.109
(0.052)**
-0.114
(0.055)**
-0.085
(0.053)*
Black
0.101
(0.049)**
0.173
(0.050)***
0.138
(0.050)***
Number of children
-0.035
(0.013)**
-0.024
(0.017)
-0.033
(0.015)**
Divorced/Separated
-0.031
(0.054)
-0.003
(0.061)
0.012
(0.056)
Single
-0.012
(0.055)
0.013
(0.070)
0.033
(0.064)
Widowed
-0.029
(0.058)
-0.026
(0.070)
-0.005
(0.083)
Retired
0.188
(0.093)**
0.086
(0.057)
0.144
(0.055)***
Self-employed
-0.091
(0.046)**
-0.123
(0.079)*
-0.150
(0.034)***
Receives welfare
0.447
(0.065)***
0.528
(0.060)***
0.469
(0.071)***
Private health insurance
-0.096
(0.038)**
-0.173
(0.073)***
-0.100
(0.136)
Currently smokes
0.147
(0.042)***
0.192
(0.042)***
0.166
(0.051)***
Expects medical expenses
0.375
(0.051)***
0.400
(0.055)***
0.411
(0.039)***
Father still living
-0.136
(0.037)***
-0.141
(0.047)***
-0.137
(0.039)***
Year 1998
0.003
(0.035)
0.007
(0.046)
0.012
(0.041)
Year 2001
0.078
(0.033)**
0.086
(0.038)**
0.076
(0.036)**
Constant
0.409
(0.169)**
0.779
(0.346)***
-0.555
(0.667)
__________________________________________________________________________________________________________
Table 4
The Effect of a Change in Poor Health Status on the Probability of Financial Strain
and a Change in Financial Strain on the Probability of Poor Health
______________________________________________________________________________________
Models
Pred Prob
Pred Prob
of being under
of being in
Financial Strain Poor Health
ME of a change
in Health Status
on Financial Strain
ME of a change
in Financial Strain
on Poor Health
______________________________________________________________________________________
All Households
Delinquent 2 months or more
0.033
0.207
0.054***
0.033
(Total Assets/Total Debts) < 1.0
0.040
0.207
0.028***
0.006
(Liquid Assets/Income) < 0.25
0.366
0.201
0.110***
0.040
______________________________________________________________________________________
Marginal effects were calculated using the weighted sample means.
Effects by Education Level
The effect of poor health on financial strain may vary for
different income groups.
• Difficult to calculate permanent income using SCF.
• We use education groups (high school education or less, some
college, college degree) as proxies for permanent income.
• The impact that poor health has on delinquency and
assets/debts < 1.0 decreases and becomes less significant as
education level of the head increases.
• The impact that poor health has on liquid assets/income < 0.25
increases and becomes more significant as education level of
the head increases.
Elasticities
• Use marginal effects and predicted probabilities to calculate
elasticities:
E= (% financial strain / %  in poor health)
= 0.054 * [20.7 / 3.3] = 0.339
• 10% increase in percentage of households in poor health
increases percentage of delinquent households by 3.39%.
• Increases percentage of households with assets/debts < 1.0 by
1.45% and liquid assets/income < 0.25 by 0.62%.
• Poor health has the largest effect on the percentage of
delinquent households.
Conclusions
• Using a more robust conceptualization of SES,
evidence shows that the direction of causality is
primarily from health to SES than SES to health.
• Findings are robust across all 3 measures of financial
burden.
• Poor health increases the probability of financial
strain.
• Little evidence that financial strain contributes to
poor health.
Implications
• Gaps in health inequality may be
contributing to widening financial
disparities.
• Those most likely to be affected are
low-to-middle income families,
especially those already in poor
health.
• Those with lower incomes who
are in poor health may find
themselves in a vicious cycle.
• Severe health conditions may result
in larger financial burdens while
large financial burdens are unlikely
to accelerate a decline in health
status.
Decrease
in overall
Quality of life
Poor Health
Limited capacity
to increase
income
and wealth
Decreased access
to quality
health care
Increased
financial strain
Policy Implications
• May result in greater dependency on government
assistance.
• Reduction in overall household welfare.
• More affordable and quality health care services for
the poor may result in improved health outcomes and
overall economic well-being.
Limitations and Directions for Future Research
• Longitudinal data to examine in more detail the
relationship between household finances and health.
• Further investigation of the definition of financial
strain and the definition of health.
• Issues of identification and instruments.
• Additional research on the relationship between
financial burden and health across households (i.e.
income, age, gender, and race).
Where do we go from here?
No Pain, No Strain:
Impact of Health on the Financial Security of the Elderly
(with Hyungsoo Kim, University of Kentucky)
Motivation:
• U.S. population is rapidly aging.
• Rising costs of health care (insurance premiums and medical
expenses).
• Dramatic growth in household debt levels for families near or
in retirement.
• Elderly will be particularly vulnerable to financial strain from
rising health care burdens.
Description of the Data
Data from the 2004 Health and Retirement Study (HRS)
Measures of health status:
• Self-reported health status (SRH)
• Objective measures of health:
– Severe chronic health condition
– Mild chronic health condition
Households are identified as “financially strained” if
• Solvency ratio: total assets/total debts < 1.0
• Liquidity ratio: liquid assets/monthly income < 2.5
• Wealth accumulation ratio: investment assets/net worth < 0.25
Direction of Causality
• At retirement, shift from accumulating wealth to spending it
down.
• Also, there is a point where additional spending on health
services results in little improvement in health status.
• Research shows the pathway from health to financial strain is
more likely to be dominant.
 Smith (1997, 1999)
As individuals grow older, changes in economic resources have little
additional impact on health.
 Smith & Kington (1997) and Lee & Kim (2003)
Direction of causation for older populations is from health to wealth.
Empirical Framework
• Focus on effect of health on financial strain.
• Assume the effect of financial strain on health is negligible for
elderly.
Two-stage probit model:
FS*i  α  H *i    X i  ε i
where
FSi* = the degree to which the respondent is under financial strain
Hi* = the degree to which the respondent is in poor health
Probability of Financial Strain
Xi includes:
• Financial characteristics of household: income, assets,
monetary transfers
• Demographics: elderly person’s age, education, gender,
marital status, race/ethnicity, living arrangements, employment
status, health insurance coverage (Medigap, Medicare HMO,
employer-sponsored health insurance plan, Medicaid)
Instruments for H*i :
smoking and exercise (measures of health behaviors)
Key Findings
• Health problems significantly increase likelihood of financial
strain for the elderly, especially those with severe chronic
conditions.
• Findings were consistent for all measures of financial strain
and health.
• Impact of poor health was significantly larger for severe
chronic conditions than for mild chronic conditions and SRH.
• Supplementary health insurance coverage significantly
mitigated financial strain for the elderly.
• The oldest elderly (aged 80+) may be most vulnerable.
Implications
• Using financial ratios provides a more
comprehensive picture of how health
affects overall financial security of the elderly.
• Important to consider both subjective and objective health
measures to determine who is likely to bear greatest financial
burden.
• For elderly persons who have not adequately saved for
retirement, a severe chronic condition could result in rapid
wealth depletion, resulting in serious financial strain.
• The results could be devastating for low-income elderly, who
do not qualify for Medicaid and who cannot afford health
insurance.
Other directions for future research….