Transcript Spacing between children and trends in mean age of

Spacing between children and
trends in mean age of successive birth
orders:
quite different stories!
Marion Burkimsher
Affiliated to the University of Lausanne
Assertion
Spacing≠ difference in mean age of successive birth orders
(Age at birth childx+1 – Age at birth childx) ≠ (MABx+1– MABx)
This is counter-intuitive!
Spacing: relates to behaviour of individuals
MABx: relates to mean of whole group
1. Model of spacing and Δ(MABx+1-MABx)
2. Spacing between children – empirical results
3. Mean age at birth of successive birth orders - data
4. Does the model have any relation to reality?
5. Relevance to real life
1. Model of spacing and Δ(MABx+1-MABx)
2. Spacing between children – empirical results
3. Mean age at birth of successive birth orders - data
4. Does the model have any relation to reality?
5. Relevance to real life
Scenario 1: 4 women, having 1, 2, 3 and 4 children respectively
Age
25
Mean age at 1st birth: 25 Std dev 0
Mean age at 2nd birth: 28 Std dev 0
Mean age at 3rd birth: 31 Std dev 0
Mean age at 4th birth: 34
28
31
34
Mean spacing between all birth
orders: 3 years
Mean age at last birth: 29.5
(MAB4-MAB3) = (MAB3-MAB2) = (MAB2-MAB1) = 3yrs
Scenario 2: 4 women, having 1, 2, 3 and 4 children respectively
Age
27
30
Mean age at 1st birth: 31.5 Std dev 3.9
Mean age at 2nd birth: 33 Std dev 3.0
Mean age at 3rd birth: 34.5 Std dev 2.1
Mean age at 4th birth: 36
33
36
Mean spacing between all birth
orders: 3 years
Mean age at last birth: 36
(MAB4-MAB3) = (MAB3-MAB2) = (MAB2-MAB1) = 1.5yrs
1. Model of spacing and Δ(MABx+1-MABx)
2. Spacing between children – empirical results
3. Mean age at birth of successive birth orders - data
4. Does the model have any relation to reality?
5. Relevance to real life
What is birth spacing?
The mean gap between births of successive children
It is, by definition, a cohort measure
It is a very skewed distribution: mean>median>mode
In Switzerland the mode is almost always a 2 year gap
The likelihood of progressing on to a further child has
changed much more than the interval between births
Probability distribution of 2nd child after 1st
by period when 1st child born
Life table analysis, Swiss census 2000
Probability distribution of 3rd child after 2nd
by period when 2nd child born
Life table analysis, Swiss census 2000
Probability distribution of 4th child after 3rd
by period when 3rd child born
Life table analysis, Swiss census 2000
Median duration between 1st and 2nd birth by these factors
Note: * the median spacing associated with total number of children can only be calculated for
those who have 2+ children, so is not exactly equivalent to other measures which include those
who have only one
Life table analysis, Swiss census 2000
Summary on spacing
Very small variations in birth spacing over time
No trend towards wider or narrower birth spacing
1. Model of spacing and Δ(MABx+1-MABx)
2. Spacing between children – empirical results
3. Mean age at birth of successive birth orders - data
4. Does the model have any relation to reality?
5. Relevance to real life
Mean age of mother at 1st, 2nd, 3rd, 4th birth
This can be calculated from birth registration records:
period measure
Since ~1998 biological birth order has been recorded; for
before then, biological birth order distributions have been
modelled (Burkimsher 2001)
Can calculate the difference between mean age at 1st birth
and mean age at 2nd birth, etc. (MAB2-MAB1….)
Burkimsher, M. 2011. Modelling biological birth order and comparison with census parity data in
Switzerland: a report to complement the Swiss data in the Human Fertility Collection (HFC). MPIDR
Technical Report TR-2011-005. (2011).
Mean age at 1st, 2nd, 3rd and 4th births
(MAB4-MAB3)
(MAB3-MAB2)
(MAB2-MAB1)
Increase in MAB1: 5.3 years
Increase in MAB2: 4.3 years
Increase in MAB3: 2.9 years
Increase in MAB4: 2.0 years
Swiss birth registration (BEVNAT)
Difference in mean age of successive birth orders
Swiss birth registration (BEVNAT)
Standard deviation in mean age at 1st, 2nd, 3rd, 4th births
Increases 1980-2008
Std dev MAB1: 0.8
Std dev MAB2: 0.5
Std dev MAB3: 0.2
Std dev MAB4: 0
Swiss birth registration (BEVNAT)
Summary of empirical observations
From 1950s-1990s (using census 2000 data):
• Spacing between all births has stayed ~constant
• Median spacing between births ~3 yrs (mode=2yrs)
From 1969 to 2013 (using birth registration data)
• (MAB2-MAB1) declined from 3yrs to 2 yrs
• (MAB3-MAB2) declined from 2.8 to 1.2 yrs
• (MAB4-MAB3) declined from 2.4 to 1.2 yrs
Why the difference between spacing and ΔMABx?
Because they relate to different subsets of women!
Women who start their family younger in life have a
greater likelihood of having a larger family, and v.v.
Why the trends in ΔMABx post-1969?
‘All’ women used to have their 1st child around the same
age: since 1970 there has been a divergence of behaviour
Transition from Scenario 1 to Scenario 2
1. Model of spacing and Δ(MABx+1-MABx)
2. Spacing between children – empirical results
3. Mean age at birth of successive birth orders - data
4. Does the model have any relation to reality?
5. Relevance to real life
Scenario 1: 4 women, having 1, 2, 3 and 4 children respectively
Age
25
Mean age at 1st birth: 25 Std dev 0
Mean age at 2nd birth: 28 Std dev 0
Mean age at 3rd birth: 31 Std dev 0
Mean age at 4th birth: 34
28
31
~1969
34
Mean spacing between all birth
orders: 3 years
Mean age at last birth: 29.5
(MAB4-MAB3) = (MAB3-MAB2) = (MAB2-MAB1) = 3yrs
Scenario 2: 4 women, having 1, 2, 3 and 4 children respectively
Age
27
33
30
Mean age at 1st birth: 31.5 Std dev 3.9
Mean age at 2nd birth: 33 Std dev 3.0
Mean age at 3rd birth: 34.5 Std dev 2.1
Mean age at 4th birth: 36
~2008
36
Mean spacing between all birth
orders: 3 years
Mean age at last birth: 36
(MAB4-MAB3) = (MAB3-MAB2) = (MAB2-MAB1) = 1.5yrs
How well does the model match reality?
• Changes in MABx
• Changes in (MABx+1 – MABx)
• Changes in Std dev MABx
Δ yr 1-yr2
Model
Actual
MAB1
25->31.5 = +6.5yrs
25.1->30.4 = +5.3yrs
MAB2
28->33 = +5yrs
28.0->32.3 = +4.3yrs
MAB3
31->34.5 = +3.5yrs
30.7->33.6 = +2.9yrs
MAB4
34->36 = +2yrs
32.8->34.8 = +2.0yrs
Δ yr 1-yr2
Model
Actual
(MAB2 – MAB1)
3->1.5 = -1.5yrs
3->2 = -1yr
(MAB3 – MAB2)
3->1.5 = -1.5yrs
2.8->1.2 = -1.6yrs
(MAB4 – MAB3)
3->1.5 = -1.5yrs
2.4->1.2 = -1.2yrs
Δ yr 1-yr2
Model
Actual
Std dev MAB1
0->3.9 = +3.9yrs
4.5->5.3 = +0.8yrs
Std dev MAB2
0->3.0 = +3.0yrs
4.2->4.7 = +0.5yrs
Std dev MAB3
0->2.1 = +2.1yrs
4.2->4.4 = +0.2yrs
Std dev MAB4
-
4.2->4.2 = 0yrs
1. Model of spacing and Δ(MABx+1-MABx)
2. Spacing between children – empirical results
3. Mean age at birth of successive birth orders - data
4. Does the model have any relation to reality?
5. Relevance to real life
Why is this relevant?
• Do not make incorrect deductions as in Houle &
Shkolnikov (2006)*
• Spacing data is essential for forecasting births and
predicting evolution of TFR, using method proposed by
McDonald and Kippen (2011)*
• Change in fertility behaviour has implications for type
of contraceptive requirements
Possible
permanent
contraception
Temporary,
reversible
contraception
Conclusion
A complicated, little studied aspect of fertility studies…
Feedback welcome!
*Houle, René and Vladimir M. Shkolnikov, 2006. Regularities and peculiarities of birth schedules
in industrialized countries: an analysis of FFS data. MPIDR Working Paper WP-2006-015 Rostock.
*McDonald, Peter and Rebecca Kippen. 2011. Forecasting Births, Feature Article, Cat. no.
2051.0, Australian Bureau of Statistics, Canberra
Thank you!
[email protected]
Differences in MABx by education
Highly educated women start their families later,
but differentials by family size similar
Mean age at 1st birth by educational level
5.3 years
2 years
Swiss census 2000
Variation in timing of 1st birth by educational level
Swiss census 2000
Differences in MABx (tertiary – low educated)
Swiss census 2000