Longterm Effects of 19591961 China Famine: Mainland China and Hong Kong

1
Long-term Effects of 1959-1961 China Famine
Mainland China and Hong Kong

• Douglas Almond, Lena Edlund, Hongbin Li and
  Junsen Zhang

2
Overview

• Fetal Origins Hypothesis
• Conceptual Framework
• The 1959-61 China Famine
• Data
• 1990 and 2000 mainland censuses, 1 samples
• 1985/87 mainland fertility surveys
• Hong Kong natality microdata

3
Fetal Origins/Barker Hypothesis

• Textbook status in medicine
• Disruptions to early-life health has persistent
  effects on subsequent health
• Consider a health production functiondeterminants
  of health divided into
• Post-birth health investments (vaccinations,
  diet,...)
• Initial health endowment
• Barker hypothesis emphasizes 2 the health
  endowment (often proxied by birth weight)

4
Evidence on Barker

• Long-term Health Effects
• Controlled Laboratory Experiments
• Dutch Famine
• China Famine
• St. Clair, et al. 05 Luo, Mu and Zhang 06
• 1918 Influenza Pandemic
• Health Outcomes (Almond and Mazumder, 2005)
• Human Capital Outcomes (Almond, 2006)

5
1918 Influenza Pandemic in the United States
6
1980 Disability Rates by Quarter of Birth
7
Previous China Famine Papers

• Meng and Qiang
• 1990 Census
• CHNS, 1989
• Excludes in utero famine cohort
• Finds no effect in OLS
• Gorgens, Meng, Vaithianathan
• CHNS, 1989, 1991, 1993, 1997
• Control, Old, Young, control
• 1938-47, 1948-1956, 1957-1961, 1962-71
• Finds young famine cohort 2 shorter
• Chen and Zhou
• CHNS
• Regional variation in severity
• Height, labor supply, agrarian income negative

8
Conceptual Framework

• Famine raised Infant Mortality Rate (IMR) b/c
  change in
• Mortality selection
• or
• Unobserved health distribution
• Polar implication for subsequent health
• Fetal origins hypothesis assumes
• Unobserved health distribution shifted
• Some of the shift was permanent

From Almond (2006)
9
(No Transcript)
10
Empirical Framework

• Compare outcomes
• cohorts in utero during treatment (famine)
• vs.
• adjacent cohorts
• Question when was the famines peak?
• Death rate 1960
• Birth rate
• Average month of birth 1960 (61 HK)
• Conception Index 1960/61
• Sex Ratio 1960-

11
Famine Timing Death Rate
12
Famine Timing Average MOB
13
(No Transcript)
14
Famine Timing Conception Index
15
(No Transcript)
16
Famines impact the reproductive ages
17
2000 China Census DataCohort Size by Birth
Month(urban residents)
18
In Utero Cohort
19
Cohort Size by Birth Month(rural residents)
20
In Utero Cohort
21
Sex Ratios gt Famine peaked 1960/61(same answer)
22
China 2000 Census, 1 sample
23
Birth Cohort Regressions

• 1956-1964 birth cohorts only
• Cubic in birth year (flexible)
• Dummies for 12 months of birth
• Separate regressions by gender
• Indicator for whether born November 1959 to
  October 1961

24
(No Transcript)
25
(No Transcript)
26
(No Transcript)
27
(No Transcript)
28
(No Transcript)
29
(No Transcript)
30
Fixed Effect Regressions

• Include dummies for each birth cohort and in
  utero cohort
• Include provinceurban dummies
• 1959-1961 Birth Cohorts Only
• Differences-in-differences estimator
• Treatment is regional variation in the Famine as
  measured by all-age mortality rates (e.g. Anhui
  rural very bad)

31
Regional Variation in Mortality
32
(No Transcript)
33
China Fertility Surveys

• 1985
• Hebei
• Shaangxi
• Shanghai
• 1987
• Beijing
• Liaoning
• Shandong
• Guangdong
• Guizhou
• Gansu

• In utero cohort
• Fewer children
• More female offspring
• Effect for both mothers and fathers

34
11.5
10.2
15.8
41.3
6.8
12.3
52.3
15.1
35
Fraction male by Mothers YOB
36
Fraction male by Fathers YOB
37
(No Transcript)
38
Alternative Explanations

• The 1960/61 birth cohort different along other
  (non-fetal health) dimensions?
• Fertility negatively selected
• Who gives birth in famines???
• Empirical evidence of positive selection
• China fertility surveys 1985/87
• (parental education)
• Dutch famine 1944/45
• (occupation of father)
• gt other control group?

39
Positively selected
40
Positively selected
41
Alternative Explanations

• FE results suggest that famine severity predicts
  long-term damage.
• But everyone in mainland china appears touched by
  famine.
• gt Is there another control group?

42
Hong Kong

• Hong Kong unaffected by 1959-61 Famine part of
  British Empire
• Migration selection positive
• Have administrative data on births
• Identified country of birth (i.e. mainland China
  versus Hong Kong)

43
Famine Timing -- HK Mainland vs. HK born
44
Famine Timing HK Immigrants
45
Age Coverage in Hong Kong Natality Data
46
(No Transcript)
47
(No Transcript)
48
Lower SES More Daughters

• Trivers-Willard 1973

RS
RS son
RS daughter
condition
49
Pathways?

• Marital status, Darwin
• Partnership status, Norberg 04
• Infant mortality
• Male vulnerability, Wells 00
• In utero mortality
• Glucose level, Cameron 04

• Conception
• time of cycle
• Mortality
• in utero
• infant
• later

50
(No Transcript)
51
(No Transcript)
52
Conclusion

• Fetal origins hypothesis seems to apply to
  socioeconomic outcomes
• Echo effect in utero cohort have more daughters
• True in 3 large-scale datasets
• True when have a control group (HK)