Does College Education Impact Adult Morbidities? Evidence From Pre-Lottery Vietnam War Draft

1
Does College Education Impact Adult
Morbidities?Evidence From Pre-Lottery Vietnam
War Draft

• Bo MacInnisInstitute for Social
  ResearchUniversity of Michigan

2
Motivations

• Health behaviors and morbidities as potential
  mechanisms for
• Understanding education-mortality relation
• Curbing rising health care costs
• College education
• Steepest gradient observed
• Steady decline in males college enrollment

3
(No Transcript)
4
Education-Health Gradient
BA Some College HS lt HS
Ever smoked 0.45 0.57 0.57 0.68
Hypertension 0.22 0.28 0.30 0.40
Obesity 0.39 0.52 0.52 0.59
Mental Distress 0.017 0.039 0.040 0.108
Type 2 diabetes 0.036 0.054 0.059 0.095
Authors calculations from NHIS 1998-2003
5
Enrollment rate females
Card Lemieux 2000
6
Enrollment rate males
7
Education-Health Literature

• Instruments for college education
• Arkes (2001) Unemployment rate. Reduce
  work-limiting health conditions
• Currie and Moretti (2003) College opening.
  Increases infants birth weight
• Lleras-Muney (2002), Adams (2002) Compulsory
  school law. Reduces mortality increases good
  health
• Kenkel etc. (2006) K-12 education policy. High
  school graduation reduces smoking.

8

• Smoking College Education
• De Walque (2004), Grimard Parent (2005)
• Vietnam War draft Card and Lemieux (2001)
• Reduce smoking initiation
• No evidence on smoking cessation
• Contributions
• Discontinuity research design
• Endogenous effect of veteran status

9
Annual Number of Inductions
10
A Quasi-Natural Experiment

• Pre-lottery draft 1965-1969
• Most draftees were aged 19-22
• Deferments were easily obtainable
• College attendance
• Children and other family hardship
• High inductions
• Consequence for males of cohorts 1945-50

11
A Quasi-Natural Experiment (cont.)

• Draft lottery institutionalized in 1970
• Random Sequence Number (RSN)
• 1970 lottery for cohorts 1946-50
• 1971-3 lottery for cohort 1951-3 respectively
• Much reduced inductions
• College deferments effectively eliminated
• Consequence to males of cohorts 1951-1953

12
Induction Risks and College Education
13
Induction Risks and Male-Female Difference in
College Education
14
Treatment Exposure
15
Data

• National Health Interview Survey (NHIS) 1998-2003
• Personal characteristics (veteran status)
• Health outcomes chronic conditions
• Lifestyle behaviors smoking
• Body weight and height
• Base sample
• Males and females U.S. born of birth cohorts
  1942-1953, sample size 30,158

16
Smoking Initiation and Cessation
17
BMI Distributions Gender College
18
BMI Distributions Gender College
19
Estimation

• Health beduc b1veteran XB b2trend
  error
• Linear probability model for educ, veteran,
  health
• IV for educ male cohorts 1946-1950
• IV for veteran male draft age eligibility
• Difference of age in 1967 and 19 (quartic terms)
• trend no/linear/quadratic trend
• X age, gender, race, survey year, and
  interactions
• Two-stage least square estimation

20
First Stage College Education
Spec. Unadjusted No trend Linear trend Quadratic trend
College Enrollment 0.0345 (0.0118) 0.0311 (0.0179) 0.0388 (0.0178) 0.0335 (0.0190)
College Completion 0.0202 (0.0102) 0.0366 (0.0137) 0.0367 (0.0151) 0.0363 (0.0154)
sample U.S. born males and females of cohorts
1942-1953 sample size 30,158 standard errors
are in parentheses. , , significant at
1, 5, 10.
21
Estimates of the Impact of College on Morbidities
Regular vs. Never Obesity Hyper-tension Mental Distress Type 2 Diabetes
College Enrollment -0.663 (0.307) -0.42(0.38) -0.42(0.32) -0.14 (0.06) 0.10(0.14)
College Completion -0.600 (0.259) -0.70 (0.32) -1.01(1.02) -0.12 (0.06) -0.33 (0.18)
Sample size 20,131 19,143 19,631 19,631 28,715
p lt .05
22
Robustness Checks

• Control for linear, and quadratic trend
• Use various measures of induction risk as
  instruments for schooling and veteran
• Avoidance choices besides going to college
• Emigrate to Canada sample selection issue
• Become delinquent
• Obtain dependency deferments

23
Canadian Emigrants Delinquents
Year Male Male R Female Female R Delinquents
1965 1,922 656 2,592 625 -
1966 2,447 1,181 3,329 1,362 -
1967 3,032 1,766 3,750 1,783 13,500
1968 4,076 2,810 4,330 2,363 22,500
1969 4,405 3,139 5,112 3,747 30,000
1970 5,510 4,244 5,714 3,747 18,500
1971 4,778 3,512 5,477 3,510 22,500
1972 3,980 2,714 5,278 3,310 -
1973 4,120 2,854 5,356 3,389 -
24
Live Births by Year
25
Live Births by Treatment Status
26
Marriages by Year
27
Concluding Remarks

• Quasi-natural experiment RDD
• strong instrument for college education
• Return to college includes a reduction in
• smoking, obesity, and diabetes
• Policy implications
• Increase college enrollment/completion rates