Socioeconomic Status
Transcript of Socioeconomic Status
Ichiro Kawachi, MD, PhD
Monica Wang, ScD, MS
Harvard School of Public Health
PH201x: Health and Society
Socioeconomic Status (SES)
Magazine Die Gartenlaube, Titanic Sinking, http://commons.wikimedia.org/wiki/File:St%C3%B6wer_Titanic.jpg, PD
Titanic casualties
Total on Board: 2,201 Lifeboat Capacity: 1,178 Total Deaths: 1,490 Dead
68%
Saved 32%
Chart created by Dr. Kawachi
Titanic survivors
Class Saved Total
First 203 325
Second 118 285
Third 178 706
Crew 212 885
Titanic survivors
62.5
41.4
25.2 24
0
20
40
60
80
First Second Third Crew
Class
Perc
ent S
aved
Chart created by Dr. Kawachi
Why was there a social class gradient in mortality on the Titanic?
Lord Mersey
John Bigham, 1st Viscount Mersey (1840-1929), Barclay Bros. http://en.wikipedia.org/wiki/File:1stViscountMersey.jpg, PD
Why was there a social class gradient in mortality on the Titanic?
Confounding by age and sex of passengers in different sections of the boat.
Age, sex
Class Survival
Mortality on board Titanic, by gender and class
Class Men Women/Children
First 67.4% 2.7%
Second 91.7% 11.2%
Third 83.8% 57.8%
Why was there a social class gradient in mortality on the Titanic?
1. Confounding by age and sex of passengers in different sections of the boat.
2. Upper class people were physically more fit, or quicker to respond to instructions of the crew.
3. Discrimination against 3rd class passengers / Preferential treatment of 1st class passengers.
4. Structural differences in access to life-saving resources (lifeboats).
Cutaway diagram of RMS Titanic, http://commons.wikimedia.org/wiki/File:Titanic_cutaway_diagram.png, PD
Evidence of SES gradients in health in contemporary society
§ Income § Educational attainment § Occupation
Relative risks of all-cause mortality by household income level: U.S. panel study of income dynamics
3.03
2.49
2
1.45 1.36 1
0
0.5
1
1.5
2
2.5
3
3.5
<15,000 -20,000 -30,000 -50,000 -70,000 >70,000
Rel
ativ
e ris
k
Household income (1993 $)
Source: McDonough et al. 1997
Chronic disease mortality among US adults aged 25-64 by education level
0
100
200
300
400
500
600
Men Women
Dea
ths
per
100,
000
< 12 yrs12 yrs13+ yrs
Education
Source: Health, United States, SES and Health Chartbook 1998
Chronic disease mortality among US adults aged 25-64 by education level
0
100
200
300
400
500
600
Men Women
Dea
ths
per
100,
000
< 12 yrs12 yrs13+ yrs
Education
Source: Health, United States, SES and Health Chartbook 1998
Is this cause and effect?
Mortality gradient by occupational class in the British Whitehall Study
00.5
11.5
22.5
33.5
44.5
Admin Prof/Exec Clerical Other
RR
of m
orta
lity
Marmot, MG et al. (1978). "Employment grade and coronary heart disease in British civil servants". Journal of Epidemiology and Community Health 32 (4): 244–249. doi:10.1136/jech.32.4.244
Two types of threat to causal inference
• Reverse Causation • Confounding (omitted variable bias)
Reverse Causation: Bad health compromises educational attainment, not the other way round
Education Health
Reverse Causation: Bad health compromises educational attainment, not the other way round
Education Health
It is often asserted that education is less susceptible to reverse causation (compared to income & occupation)
• Most people have completed their schooling by the time they develop chronic disease
• If you get sick, you can’t lose education (in the way that you can lose income or your job)
• But is that strictly accurate?
Evidence from 1958 British Birth Cohort (National Child Development Study, NCDS)
• Chronic health conditions during childhood do appear to have an adverse impact on educational attainment (Case, Fertig & Pason, 2005)
• Even after taking into account household and parental characteristics,
each chronic condition reported at age 7 leads to on average 0.3 fewer subjects passed on General Certificate of Education O-level examinations at age sixteen
• In short, chronic conditions during childhood – e.g. diabetes, ADHD, or mental health problems – result in children missing school
In summary …
• Reverse causation is real! • Though not every instance of SES
gradients reflects this bias
The relationship between SES and health is reciprocal and dynamic across the life course
Adler et al. Reaching for a Healthier Life, The John D. and Catherine T. MacArthur Foundation Research Network on Socioeconomic Status and Health. http://www.macses.ucsf.edu/downloads/reaching_for_a_healthier_life.pdf
Two types of threat to causal inference
• Reverse Causation • Confounding (omitted variable bias)
Confounding
Association between SES & health is spurious, and reflects the influence of omitted “third” variables.
SES Health
?
Sidebar (A bit of Epi review)
Carrying matches in your pocket
Lung cancer Relative risk = 2.0
Should we advise people to stop carrying matches around?
Sidebar (A bit of Epi review)
Carrying matches in your pocket
Lung cancer Relative risk = 2.0
smoking
How would you demonstrate that the association is confounded by cigarette smoking?
Association between carrying matches & lung cancer stratified by smoking status
Smokers Non-‐smokers
Relative risk of lung cancer among match-carriers vs. non-carriers
Association between carrying matches & lung cancer stratified by smoking status
Smokers Non-‐smokers
Relative risk of lung cancer among match-carriers vs. non-carriers
1.00
Association between carrying matches & lung cancer stratified by smoking status
Smokers Non-‐smokers
Relative risk of lung cancer among match-carriers vs. non-carriers
1.00
1.00
Is it necessary to adjust for “carrying matches” when studying the association of smoking with lung cancer?
Lung cancer
Carrying matches
Relative risk = 20 Smoking
Is it necessary to adjust for “carrying matches” when studying the association of smoking with lung cancer?
Lung cancer
Carrying matches
Relative risk = 20 Smoking
Moving beyond observational data
Identification Strategies to Assess Causality § Randomized controlled trials § Quasi-experiments
Randomized controlled trials of schooling
• Extremely sparse due to low feasibility and ethical concerns
• Two trials studied intensive preschool interventions
a. High/Scope Perry Pre-school Program b. Abecedarian Program
Perry Program • Intensive pre-school program in
Ypsilanti, MI (1962-7)
Perry Program • Intensive pre-school program in
Ypsilanti, MI (1962-7) • Enrolled N=123 disadvantaged
black children, age 3-4 years
Perry Program • Intensive pre-school program in
Ypsilanti, MI (1962-7) • Enrolled N=123 disadvantaged
black children, age 3-4 years
• Treatment: Daily dose of 2.5 hr instruction by teachers with Masters degree on weekday mornings, plus weekly 90-min home visit by teacher
Perry Program • Intensive pre-school program in
Ypsilanti, MI (1962-7) • Enrolled N=123 disadvantaged
black children, age 3-4 years
• Treatment: Daily dose of 2.5 hr instruction by teachers with Masters degree on weekday mornings, plus weekly 90-min home visit by teacher
• Duration: 30 wk/yr for 2 years • Follow-up through age 40
Perry Program • Intensive pre-school program in
Ypsilanti, MI (1962-7) • Enrolled N=123 disadvantaged
black children, age 3-4 years
• Treatment: Daily dose of 2.5 hr instruction by teachers with Masters degree on weekday mornings, plus weekly 90-min home visit by teacher
• Duration: 30 wk/yr for 2 years • Follow-up through age 40
Abecedarian Program • Intensive pre-school program in
Chapel Hill, NC (1972-7)
Perry Program • Intensive pre-school program in
Ypsilanti, MI (1962-7) • Enrolled N=123 disadvantaged
black children, age 3-4 years
• Treatment: Daily dose of 2.5 hr instruction by teachers with Masters degree on weekday mornings, plus weekly 90-min home visit by teacher
• Duration: 30 wk/yr for 2 years • Follow-up through age 40
Abecedarian Program • Intensive pre-school program in
Chapel Hill, NC (1972-7) • Enrolled N=111 disadvantaged
black children, age 4 months
Perry Program • Intensive pre-school program in
Ypsilanti, MI (1962-7) • Enrolled N=123 disadvantaged
black children, age 3-4 years
• Treatment: Daily dose of 2.5 hr instruction by teachers with Masters degree on weekday mornings, plus weekly 90-min home visit by teacher
• Duration: 30 wk/yr for 2 years • Follow-up through age 40
Abecedarian Program • Intensive pre-school program in
Chapel Hill, NC (1972-7) • Enrolled N=111 disadvantaged
black children, age 4 months • Treatment: 8 hrs/day, 5 days/wk
academic enrichment program for first 5 years. Additional activities from grade 1-3
Perry Program • Intensive pre-school program in
Ypsilanti, MI (1962-7) • Enrolled N=123 disadvantaged
black children, age 3-4 years
• Treatment: Daily dose of 2.5 hr instruction by teachers with Masters degree on weekday mornings, plus weekly 90-min home visit by teacher
• Duration: 30 wk/yr for 2 years • Follow-up through age 40
Abecedarian Program • Intensive pre-school program in
Chapel Hill, NC (1972-7) • Enrolled N=111 disadvantaged
black children, age 4 months
• Treatment: 8 hrs/day, 5 days/wk academic enrichment program for first 5 years. Additional activities from grade 1-3
• Duration: 50 wk/yr for 8 years • Follow-up through age 21
High/Scope Perry trial outcomes at age 27
Data from: http://www.highscope.org/content.asp?ContentId=219
Intervention
Control
Abecedarian program outcomes at age 21
Intervention
Control
Data from: http://abc.fpg.unc.edu/
WHY IS EARLY EDUCATION IMPORTANT FOR HEALTH?
What did Mischel find?
• Experimenter left room for up to 15 mins
• Almost no child lasted 15 mins • More than a quarter caved within 2 mins • Average child lasted 6 mins • Another 25 percent lasted more than 10 mins
Does marshmallow test predict anything of consequence?
Maybe the kids who skipped breakfast on the morning of the experiment had low blood glucose and were more likely to ring the bell…
Predicting adolescent cognitive and self-regulatory competencies from preschool delay of gratification: identifying diagnostic conditions
• Re-contacted subsample of original kids (N=185) when they were 16-18 years old
• Parental questions inquiring about adolescent’s coping styles and SAT scores
Shoda et al. (1990). Developmental Psychology, 26(6): 978-986.
Correlations between delay time on marshmallow test & adolescent outcomes
How likely is your child to yield to tempta4on? -‐.50***
How capable is your child of exhibi4ng self-‐control when frustrated? .40**
SAT Verbal .42*
SAT Quan4ta4ve .57**
Shoda et al. (1990). Developmental Psychology, 26(6): 978-986.
Correlations between delay time on marshmallow test & adolescent outcomes
How likely is your child to yield to tempta4on? -‐.50***
How capable is your child of exhibi4ng self-‐control when frustrated? .40**
SAT Verbal .42*
SAT Quan4ta4ve .57**
*4-‐year old kids who waited 5 mins. longer scored on average 300 points higher on SATs in high school (171 points higher on quan4ta4ve & 126 points higher on verbal).
Shoda et al. (1990). Developmental Psychology, 26(6): 978-986.
Broader implication
• Early education teaches children to be more patient (executive function)
• Self regulation promotes healthier lifestyles in adulthood
Early Educa4on Preven4ve behavior
Willpower
Quasi-Experiments
Life is short; most of us will not get around to doing experiments on social determinants.
If we can’t directly manipulate the exposure, we
can resort to- • Instrumental variables • Natural experiments
Instrumental variables
• Find a variable (Z) that causes variation in x, but has no direct effect on outcome y
• i.e. find exogenous source of variation (like a coin toss)
x y
u
Z
Can we find an instrument for education?
Education Risk of mortality
IQ, genes, personality,
etc.
Z
Can we find an instrument for education?
Educational attainment Risk of mortality
IQ, genes, personality,
etc.
Compulsory schooling laws in the state of one’s birth
Compulsory schooling laws (CSLs) as instruments for education
• In the U.S., the number of years a child must spend in school is determined by state law
• Requirements have historically varied between states, and states changed their CSLs repeatedly during the first half of the 20th century
• States extended mandatory schooling by: a) lowering the age at which children had to begin school, or b) raising the age at which they could drop out or get a work permit
• Hence, we can treat CSLs as “natural policy experiments”
Source: Lleras-Muney A (2005). “The Relationship between Education and Adult Mortality in the U.S.” Rev Econ Stud, vol. 72.
2 Stage least squares (2SLS)
edictorsPrOther X̂ k α+ Ζα+α= 10
ε + β+ β+β= 10 edictorsPrOther X̂Y k
Findings
• IV estimates of education on mortality suggest strong protective effect
• Each additional year of schooling lowers 5-year mortality rates by between 3.6 to 5 percent1
• Recent analyses using similar approach also suggest causal effect of education on improved cognitive functioning at older ages2
1.Lleras-Muney (2005). 2.Glymour, Kawachi, Robins & Berkman (2008) JECH, 62(2):532-437
Explanations for the association between schooling and health
1. Non causal: – Reverse causation – Confounding by IQ or other measure of ability – Less time spent in hazardous jobs (“warehousing”)
0
100
200
300
400
500
600
Men Women
Death
s per
100,0
00
< 12 yrs12 yrs13+ yrs
Source: Health, United States, SES and Health Chartbook 1998
Explanations for the association between schooling and health
2. Causal – cognitive mechanisms: – Acquisition of knowledge – Health literacy
Explanations for the association between schooling and health
2. Causal – cognitive mechanisms: – Acquisition of knowledge – Health literacy
3. Causal – non-cognitive mechanisms: – Self-regulation: patience, ability to plan for the future – Executive functioning
Explanations for the association between schooling and health
4. Indirect mechanisms: – Schooling and credentials as gateways to safer jobs, higher
income 5. Other:
– Higher prestige/status in community – More social connections (“social capital”) – Improved future prospects leads to greater incentive to invest
in health
If we accept that schooling matters, where should we invest? • Knudsen et al. (2006). Economic, neurobiological,
and behavioral perspectives on building America’s future workforce. PNAS, 103(27): 10155-10162.
“There is a striking convergence of four core concepts that have emerged from decades of mutually independent research in economics, neuroscience, and developmental psychology…” [ ]
Core concept #1 [From neuroscience]
• Skill development and brain maturation are
hierarchical processes in which higher-level functions depend on (and build on) lower level functions.
• Skills beget skills.
Core concept #2 [From developmental psychology]
• Early experiences have a uniquely powerful
influence on the development of cognitive and social skills and on brain architecture.
• There are developmentally sensitive periods for optimal learning, e.g. language acquisition.
Knudsen et al. (2006). Economic, neurobiological, and behavioral perspectives on building America’s future workforce. PNAS, 103(27): 10155-10162.
• Bowl or rice • Bowl of lice
Knudsen et al. (2006). Economic, neurobiological, and behavioral perspectives on building America’s future workforce. PNAS, 103(27): 10155-10162.
Core concept #3 [from economics]
• Early intervention lowers the cost of later
investment.
• Later remediation efforts are less effective – e.g. adult literacy services, prisoner rehabilitation programs & education programs for disadvantaged adults yield lower economic returns than early intervention.
Knudsen et al. (2006). Economic, neurobiological, and behavioral perspectives on building America’s future workforce. PNAS, 103(27): 10155-10162.
Core concept #4 • Everything is not all over by age 4!
• Early interventions are sustained best when they are followed by continued high quality learning experiences.
• Early investments must be followed by later investments to recoup maximum value.
Conclusion • The most cost effective strategy for education is
investing in the social and cognitive environments of children who are disadvantaged, beginning as early in life as possible.
• [The estimated rate of return per dollar of cost in the Perry Program is > 17%, i.e. far higher than the standard rate of return on stock market equity].