Health and mortality differentials among adults does selection matter l.jpg
This presentation is the property of its rightful owner.
Sponsored Links
1 / 77

Health and Mortality Differentials among adults: does selection matter? PowerPoint PPT Presentation


  • 154 Views
  • Uploaded on
  • Presentation posted in: General

Health and Mortality Differentials among adults: does selection matter?. Alberto Palloni Carolina Milesi Robert White Alyn Turner Center for Demography and Ecology University of Wisconsin-Madison. Some history: why did I end up here?. Mortality and health differentials over the life course

Download Presentation

Health and Mortality Differentials among adults: does selection matter?

An Image/Link below is provided (as is) to download presentation

Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author.While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server.


- - - - - - - - - - - - - - - - - - - - - - - - - - E N D - - - - - - - - - - - - - - - - - - - - - - - - - -

Presentation Transcript


Health and mortality differentials among adults does selection matter l.jpg

Health and Mortality Differentials among adults: does selection matter?

Alberto Palloni

Carolina Milesi

Robert White

Alyn Turner

Center for Demography and Ecology

University of Wisconsin-Madison


Some history why did i end up here l.jpg

Some history: why did I end up here?

  • Mortality and health differentials over the life course

  • The role of social classes

  • The argument of selection

  • Stages of research:

    • Early health and adult social class---stratification

    • Early health and adult mortality/health differentials

  • What if selection occurs? Are social classes ignorable if selection is relevant?


Empirical regularities l.jpg

Empirical regularities

  • Increasing income and wealth inequality in the US: sustained from 1985 onward, steeper after 1995

  • Persistently high intergenerational correlation of earnings, income and wealth

    • ρ = βcp * (σp /σc ) [.30-.60]

  • Stalling social mobility: increased stickiness

  • Persistent adult health and mortality differentials everywhere


Adult ses health mortality gradients l.jpg

Adult SES health/mortality gradients

  • Magnitude: They are large

  • Over time: increasing

  • By age: increase until age 50-55 then (tend to) decrease

  • Continuous: not accounted for by extremes

  • Pervasive: causes of death, health status, self-reported conditions

  • Fairly global (or almost) regardless of prevailing health care system


Age equivalence of mortality rates for high and low education groups l.jpg

Age equivalence of mortality rates for high and low education groups


Age equivalence of mortality rates for high and low education groups6 l.jpg

Age equivalence of mortality rates for high and low education groups


Age equivalence of mortality rates for high and low education groups7 l.jpg

Age equivalence of mortality rates for high and low education groups


Age equivalence of mortality rates for high and low education groups8 l.jpg

Age equivalence of mortality rates for high and low education groups


Age equivalence of mortality rates for high and low education groups9 l.jpg

Age equivalence of mortality rates for high and low education groups


Proportion with poor self reported health by education and age 1975 l.jpg

Proportion with poor self-reported health by education and age, 1975

Source: NHIS, 1975


Proportion with poor self reported health by education and age 2000 l.jpg

Proportion with poor self-reported health by education and age, 2000

Source: NHIS, 2000


Evidence is strong water tight l.jpg

Evidence is strong, water tight

  • Education differentials

  • Income (permanent)

  • Wealth:

    • Studies with PSID

    • Studies with HRS

    • Studies with NLSM (couples)

  • Occupation (prestige/social class)


  • The origins of social class differentials in health and mortality l.jpg

    The origins of social class differentials in health and mortality

    Social class traits

    Health


    Selection processes l.jpg

    Selection processes

    • Direct or “drift”

    • Reverse causality

    • Heterogeneity

    • Indirect ‘health selection proper’


    Indirect selection l.jpg

    Indirect Selection

    • Health inequalities by SES early in life translate into health inequalities by SES late in life.

    • This requires that:

      • Health inequalities early in life be a significant determinant of adult social stratification

      • Either that early health directly affect adut health OR that there be paths linking early health with conditions (exposure to risks) that are class related


    Need a unified theory l.jpg

    Need a unified theory

    • How does early health affect stratification?

      • Social stratification theory

      • Labor economics and wage determination

    • How does early health affect adult health?

      • Early effects theory (Barker-type hypotheses)

      • Life cycle trajectories (Kuh, Ben-Shlomo, Smith, Wadsworth)


    Conditions of possibility l.jpg

    Conditions of possibility

    • Allocation of health in early childhood is non-random (class dependent)

    • There must be a relation between early childhood health and acquisition of traits that determine social stratification

    • There must be a relation between early health and later exposure to risk (direct or indirect)


    A new problem relations involved in selection l.jpg

    A new problem: relations involved in “selection”

    Social class

    of origin

    (genes/environments)

    Social class

    of

    destination

    Traits, antecedent

    health status

    Adult health

    status


    What is social class l.jpg

    What is social class?


    What is child health status l.jpg

    What is child health status?


    Slide21 l.jpg

    “Emotional development during the first five years offers a window into the psychological growth of young children”(National Research Council, 2003)


    Slide22 l.jpg

    “Early social experiences, especially those with primary attachment figures, can have profound consequences for both biobehavioral and physiological functioning throughout the life span” (Suomi, 1999)


    What is adult health status l.jpg

    What is adult health status?


    Two tasks l.jpg

    Two tasks

    • Early health status affects adult social class positions

      • Early health status matters for intergenerational transmission of inequalities

    • Early health status is responsible for effects that account for some or all of adult SES health gradients


    The data l.jpg

    The data

    • British Cohort Survey 1958

      • Only males

      • Use information from 0 up to 41-42

      • Do not do multiple imputation to resolve attrition

      • We go from circa 8,000 to 1,000 cases

      • Disproportionate attrition among low class


    Procedures l.jpg

    Procedures

    • Formulate and estimated structural equation model:

      • Limited: one factor models

      • Limited: ignore attrition and non response

    • Use estimates to decompose paths

    • Use estimates to simulate life course


    Task 1 intergenerational transmission of inequalities l.jpg

    Task 1: intergenerational transmission of inequalities


    What pays out there l.jpg

    What pays out there?

    • Years of education?

    • Credentials?

    • Cognitive performance?

    • Labor market experience?

    • Personalities?

    • Parents’ wallets?

    • Luck?


    Slide29 l.jpg

    “It is common knowledge outside academic journals that motivation, tenacity, trustworthiness, and perseverance are important traits for success in life” (Heckman and Rubinstein, 2001)


    Slide30 l.jpg

    Path diagram of relations between early child health conditions and earnings

    Cognitive

    Skills

    1

    1

    Educational

    Attainment

    Parental Social Class

    1

    1

    3

    Early Health

    4

    Social Class

    2

    2

    2

    Parental Health

    Non-

    Cognitive

    Skills

    2


    Representation of relations bcs58 l.jpg

    Representation of relations: BCS58

    Cognitive

    skills

    age 11

    Educational

    attainment

    age 16

    SR Health

    30

    Parental social class

    SR Health

    40

    Chronic

    conditions

    age 7

    Chronic

    conditions

    age 16

    Low birth

    weight

    Social class

    age 40

    Non-

    cognitive

    skills

    age 11

    Parental health

    Social class

    age 30


    Slide32 l.jpg

    Main results

    Cognitive

    skills

    age 11

    Educational

    attainment

    age 16

    Parental social class

    Low birth

    weight

    Chronic

    conditions

    age 7

    Chronic

    conditions

    age 16

    Non-

    cognitive

    skills

    age 11

    Parental health

    Social class

    age 41/42


    Slide33 l.jpg

    Main results

    Cognitive

    skills

    age 11

    Educational

    attainment

    age 16

    Parental social class

    Low birth

    weight

    Chronic

    conditions

    age 7

    Chronic

    conditions

    age 16

    Non-

    cognitive

    skills

    age 11

    Parental health

    Social class

    age 41/42


    Slide34 l.jpg

    Main results

    Cognitive

    skills

    age 11

    Educational

    attainment

    age 16

    Parental social class

    Low birth

    weight

    Chronic

    conditions

    age 7

    Chronic

    conditions

    age 16

    Non-

    cognitive

    skills

    age 11

    Parental health

    Social class

    age 41/42


    Slide35 l.jpg

    Main results

    Cognitive

    skills

    age 11

    Educational

    attainment

    age 16

    Parental social class

    Low birth

    weight

    Chronic

    conditions

    age 7

    Chronic

    conditions

    age 16

    Non-

    cognitive

    skills

    age 11

    Parental health

    Social class

    age 41/42


    Slide36 l.jpg

    Main results

    Cognitive

    skills

    age 11

    Educational

    attainment

    age 16

    Parental social class

    Low birth

    weight

    Chronic

    conditions

    age 7

    Chronic

    conditions

    age 16

    Non-

    cognitive

    skills

    age 11

    Parental health

    Social class

    age 41/42


    Slide37 l.jpg

    Main results

    Cognitive

    skills

    age 11

    Educational

    attainment

    age 16

    Parental social class

    Low birth

    weight

    Chronic

    conditions

    age 7

    Chronic

    conditions

    age 16

    Non-

    cognitive

    skills

    age 11

    Parental health

    Social class

    age 41/42


    Main inference l.jpg

    Main inference

    • The effects of early health on social class are important but they only work through cognitive skills and non-cognitive traits. None are direct.


    Decomposition of effects l.jpg

    Decomposition of effects

    • Close to 50 percent of the correlation between parental and offspring social class works through early childhood health indicators (birthweight, chronic conditions at ages 7 and 11)


    Decomposition of effects of parental social class on sons social class l.jpg

    Decomposition of effects of parental social class on sons’ social class


    Decomposition of effects of parental social class on sons social class41 l.jpg

    Decomposition of effects of parental social class on sons’ social class


    Decomposition of effects of parental social class on sons social class42 l.jpg

    Decomposition of effects of parental social class on sons’ social class


    Decomposition of effects of parental social class on sons social class43 l.jpg

    Decomposition of effects of parental social class on sons’ social class


    Monte carlo matrix exercise l.jpg

    Monte Carlo matrix exercise

    Use structural equation model to produce mobility matrices

    Each iteration generates a random matrix (with random entries)

    From each matrix several measures of mobility can be extracted


    Slide45 l.jpg

    Probability of staying in LOWEST class


    Slide46 l.jpg

    Probability of staying in HIGHEST class


    Task 2 contribution of early health on adult health status social class gradients l.jpg

    Task 2: contribution of early health on adult health status social class gradients


    Slide48 l.jpg

    Figure 1: Simplified path diagram representing the relations

    between early child health conditions and adult social class

    ρ2

    $

    *

    1

    1

    PARENTAL SOCIOECONOMIC

    COGNITIVE

    EDUCATIONAL

    BACKGROUND

    SKILLS

    ATTAINMENT

    (

    *

    "

    *

    ρ0

    1

    2

    3

    1

    ,

    1

    (3

    EARLY HEALTH

    SOCIAL CLASS

    (at age 33)

    *

    4

    "

    2

    (

    ,

    2

    2

    *

    5

    PARENTAL HEALTH

    SOFT SKILLS

    SOCIAL CLASS

    (at age 41-42)

    $

    2

    *

    6

    ρ1


    Representation of relations bcs5849 l.jpg

    Representation of relations: BCS58

    Cognitive

    skills

    age 11

    Educational

    attainment

    age 16

    SR Health

    30

    Parental social class

    SR Health

    40

    Chronic

    conditions

    age 7

    Chronic

    conditions

    age 16

    Low birth

    weight

    Social class

    age 40

    Non-

    cognitive

    skills

    age 11

    Parental health

    Social class

    age 30


    Representation of relations bcs5850 l.jpg

    Representation of relations: BCS58

    Social Class 30

    SR Health

    30

    Educational

    attainment

    Cognition

    Early

    Health

    SR Health

    40

    Social Class 40


    Representation of relations bcs5851 l.jpg

    Representation of relations: BCS58

    Social Class 30

    SR Health

    30

    Educational

    attainment

    Cognition

    Early

    Health

    SR Health

    40

    Social Class 40


    Representation of relations bcs5852 l.jpg

    Representation of relations: BCS58

    Social Class 30

    SR Health

    30

    Educational

    attainment

    Cognition

    Early

    Health

    SR Health

    40

    Social Class 40


    Representation of relations bcs5853 l.jpg

    Representation of relations: BCS58

    Social Class 30

    SR Health

    30

    Educational

    attainment

    Cognition

    Early

    Health

    SR Health

    40

    Social Class 40


    Representation of relations bcs5854 l.jpg

    Representation of relations: BCS58

    Social Class 30

    SR Health

    30

    Educational

    attainment

    Cognition

    Early

    Health

    SR Health

    40

    Social Class 40


    Representation of relations bcs5855 l.jpg

    Representation of relations: BCS58

    Social Class 30

    SR Health

    30

    Educational

    attainment

    Cognition

    Early

    Health

    SR Health

    40

    Social Class 40


    Representation of relations bcs5856 l.jpg

    Representation of relations: BCS58

    Social Class 30

    SR Health

    30

    Educational

    attainment

    Cognition

    Early

    Health

    SR Health

    40

    Social Class 40


    Representation of relations bcs5857 l.jpg

    Representation of relations: BCS58

    Social Class 30

    SR Health

    30

    Educational

    attainment

    Cognition

    Early

    Health

    SR Health

    40

    Social Class 40


    Some estimates effects on log odds of being in poor health general model l.jpg

    Some estimates: effects on log odds of being in poor health: general model

    • SRH at age 40

      • Social Class at 40: -.109(.04)

      • Ed Attainment:-.155(.120)

      • Social Class at 30:-.031(.060)

      • LBW:.679(.24)

      • SRH at 30:1.148(.108)!!!!!!!

    • LBW or Chronic conditions at 11 have similar effects


    Main results relation between adult social class and health status l.jpg

    Main results: relation between adult social class and health status

    • Social class of origins has significant net effects

    • Large effects of cognition (education)

    • Persistent direct effects of birthweight

    • Persistent effects of No of chronic conditions at age 7


    Main inferences l.jpg

    Main inferences

    • The effects of lagged social class and health status at age 41-42 are mainly explained by education-cognition

    • Effects of birthweight and No of chronic conditions at age 7 are direct and important


    Main inferences cont l.jpg

    Main inferences (cont.)

    • Effects of SClass(30) on SRH(40) are strong even after controlling for early life conditions

    • Effects of LBW and chronic conditions on SRH (40) at 11 are strong

    • But, contemporaneous and lagged effects of social class persist


    Decomposition of effects62 l.jpg

    Decomposition of effects

    • Close to 60 percent of the correlation between lagged social class at age 33 and self reported health status at age 41-42 works through early childhood health indicators (birthweight, chronic conditions at ages 7 and 11)


    Decomposition of effects63 l.jpg

    Decomposition of effects

    • About 10 percent of total effects of Social Class at 30 on SRH at 40 are attributable to early conditions

    • The remainder is NOT selection

    • The remainder is NOT trivial: people of lower social classes have a relative risk of being in poor health that is 1.70 as high as that of higher classes


    Monte carlo matrix exercise64 l.jpg

    Monte Carlo matrix exercise

    Use structural equation model to produce social class-health status mobility matrices

    Each iteration generates a random matrix (with random entries)

    From each matrix several measures of social class and health mobility can be extracted


    If early health effects were five times larger than they are l.jpg

    If early health effects were five times larger than they are


    If early health effects were five times larger than they are66 l.jpg

    If early health effects were five times larger than they are


    If early health effects were five times larger than they are67 l.jpg

    If early health effects were five times larger than they are


    But they are not five times as large l.jpg

    But they are not five times as large…


    But they are not five times as large69 l.jpg

    But they are not five times as large…


    But they are not five times as large70 l.jpg

    But they are not five times as large…


    Inference from matrix exercise l.jpg

    Inference from matrix exercise

    • Early health status accounts for only 7 to 10 percent of the actual probability of immobility

    • Early health status accounts for no more than 10 percent of the association between lagged social class and self reported health status during adulthood.


    Conclusions l.jpg

    Conclusions

    • Early childhood health has a small but non-ignorable contribution to overall transmission of social inequalities

    • It is not an important mechanisms leading to health selection. If selection is important it must be due to other types of mechanisms

    • We are likely to have underestimated its effects due to:

      • Measurement problems

      • Attrition processes


    New strategies l.jpg

    New strategies

    • BCS 1958 and 1970 with Multiple Imputation

    • Artificial cohorts with multiple imputation


    Data bases l.jpg

    Data Bases

    ECLS

    or

    NLSY

    NELS


    Data bases75 l.jpg

    Data Bases

    ECLS

    or

    NLSY

    NELS


    Data bases76 l.jpg

    Data Bases

    ECLS

    or

    NLSY

    NELS

    Imputation


    Slide77 l.jpg

    END

    For the time being


  • Login