Mortality and I ts Measurement Section A

Stan Becker, PhD Definitions and I ndicators

Bloomberg School of Public Health

Year of Registration vs.

De J ure vs. De Facto
Year of Occurrence
� De jure � Year of registration
– Death is registered at place of residence – Death is tabulated according to the time
� De facto of its registration
– Death is registered at place of � Year of occurrence
occurrence – Death is tabulated according to the time
it occurs

4 5

I ndicators I ndicators

� Basic indicator � Midyear population is an approximation of

– Crude death rate—Number of deaths the average population exposed to risk
per 1,000 population � Total person-years lived is a better
denominator if available
Number of deaths
∗ 1000
Midyear po pulation

Continued 6 Continued 7
 I ndicators Age Specific Death Rate:

� Typically computed for calendar year so as � Age Specific Death Rate—Number of

to eliminate the effect of seasonal or deaths per 1,000 persons of a specific age
monthly variations on the comparability of (group)
the rates Da
= ∗1000
� There is a problem if the mid-point of the Pa
time reference for the numerator does not
correspond to the time of the midpoint � Where Da = Number of deaths in age
population (group) a
P a = Midyear population in age
(group) a
8 9

Egypt ASDR Graph 1990 I nfant Mortality “Rate”

120 � I nfant Mortality Rate—Number of infant
100 deaths per 1000 births
Males
80
Females

60 D0
40
IMR = ∗ 1000
B

ASDR per 1000

20

0
0 1 5 10 15 20 25 30 35 40 45 50 55 60 65+ � Where D0 = Number of infant (< 1 year)
Age group
deaths
Source: UN, Demographic year book 1990 10
B = Number of births Continued 11

I nfant Mortality “Rate” Exercise

I nfant Mortality “Rate”

� Calculate the conventional infant mortality

rate for the U.S. (1990), based on the
following data
United States
Year Birth Cohort Births Deaths I nfant Deaths
1989 1989 4,040,958 39,655 33,645
1990 1989 -- 5,861
1990 1990 4,158,212 38,351 32,490
1991 1990 -- 5,657
1991 1991 4,110,907 36,766 31,109

You have 15 seconds to calculate the answer. You

may pause the presentation if you need more time.
12 Source: Vital Statistics of the United States 1989-1991 13
 Exercise Answer I nfant Mortality “Rate”
I nfant Mortality “Rate”

� The correct I MR for 1990 is as follows: � Good index of child mortality in low
– 9.22 infant deaths per 1,000 births mortality populations; less good in high
mortality populations
United States
Year Birth Cohort Births Deaths I nfant Deaths � Because of the very high level of mortality
1989 1989 4,040,958 39,655 33,645
1990 1989 -- 5,861 in the first hours, days, and weeks of life,
1990 1990 4,158,212 38,351 32,490 I MR is broken down into even more
1991 1990 -- 5,657
1991 1991 4,110,907 36,766 31,109 specific rates
� Not a true rate

14 15

Adjusted I nfant Mortality Rate Adjusted I nfant Mortality Rate

� To be used when the number of births

fluctuates sharply between years and
within a year
� I nfant deaths in a year occur to births of
that year and also to births of the previous
year

Continued 16 17

Three Techniques for Adjusting Adjusted I nfant Mortality Rate

I nfant Mortality Rate Cohort Probability

� Cohort probability � The portion of deaths under one in year y

� Data by year and cohort and the portion of deaths under one in
� Separation factors year y+1 occurring to births in year y are
combined and divided by the births in
year y

Dy' + Dy'' +1
IMRC = ∗1000
By
18 Continued 19
 Adjusted I nfant Mortality Rate Exercise
Cohort Probability Cohort Probability

� Calculate the cohort probability of infant

death for the U.S. (1990)
United States
Year Birth Cohort Births Deaths I nfant Deaths
1989 1989 4,040,958 39,655 33,645
1990 1989 -- 5,861
1990 1990 4,158,212 38,351 32,490
1991 1990 -- 5,657
1991 1991 4,110,907 36,766 31,109

You have 15 seconds to calculate the answer. You

may pause the presentation if you need more time.
20 Source: Vital Statistics of the United States 1989-1991 21

Exercise Answer Adjusted I nfant Mortality Rate

Cohort Probability Data by Year and Cohort

� The correct I MRC for 1990 is as follows: � Each portion of the infant deaths occurring
– 9.17 infant deaths per 1,000 births in a given year is related to the births in
the appropriate year and cohort
United States
Year Birth Cohort Births Deaths I nfant Deaths
1989 1989 4,040,958 39,655 33,645 ⎛ D 'y'
1990 1989 -- 5,861
D 'y ⎞
1990 1990 4,158,212 38,351 32,490 +
IMR B = ⎜ ⎟ ∗1000
1991 1990 -- 5,657 ⎜B ⎟
1991 1991 4,110,907 36,766 31,109 ⎝ y −1 B y ⎠

22 23

Adjusted I nfant Mortality Rate Exercise

Data by Year and Cohort Data by Year and Cohort

� Calculate the adjusted infant mortality rate

using data by year and cohort for the U.S.
(1990)
United States
Year Birth Cohort Births Deaths I nfant Deaths
1989 1989 4,040,958 39,655 33,645
1990 1989 -- 5,861
1990 1990 4,158,212 38,351 32,490
1991 1990 -- 5,657
1991 1991 4,110,907 36,766 31,109

24 Source: Vital Statistics of the United States 1989-1991 25

 Exercise Adjusted I nfant Mortality Rate
Data by Year and Cohort Separation Factors

� The correct I MRB for 1990 is as follows: � I nfant deaths in year y are divided by a
– 9.26 infant deaths per 1,000 infant weighted average of births in years y and
births y-1
� The weights are called separation factors
United States
Year Births Infant Deaths ⎛ Dy ⎞
IMR W = ⎜ '' ⎟ ∗1000
1989 4 040 958 39 655 ⎜ f B + f 'B ⎟
⎝ y −1 y ⎠
1990 4 158 212 38 351
1991 4 040 958 36 766 D 'y
'
f = and f '' = 1 − f '
Dy
26 Continued 27

Adjusted I nfant Mortality Rate Exercise

Separation Factors Separation Factors

� Calculate the adjusted infant mortality rate

using the separation factors technique for
the U.S. (1990)
United States
Year Birth Cohort Births Deaths I nfant Deaths
1989 1989 4,040,958 39,655 33,645
1990 1989 -- 5,861
1990 1990 4,158,212 38,351 32,490
1991 1990 -- 5,657
1991 1991 4,110,907 36,766 31,109

You have 15 seconds to calculate the answer. You

may pause the presentation if you need more time.
28 Source: Vital Statistics of the United States 1989-1991 29

Exercise Answer Summary

Separation Factors

� The correct I MRW for 1990 is as follows: � Mortality is one of the demographic
– 9.26 infant deaths per 1,000 births phenomena most commonly studied
United States
� I t is important to take into consideration
Year Birth Cohort Births Deaths I nfant Deaths the definitions in the data collection
1989 1989 4,040,958 39,655 33,645
1990 1989 -- 5,861 systems when interpreting and comparing
1990 1990 4,158,212 38,351 32,490 different rates
1991 1990 -- 5,657
1991 1991 4,110,907 36,766 31,109 � There are several ways of calculating infant
mortality rates; the results can be
somewhat different
30 31
 Neonatal Mortality Rate

� Neonatal Mortality Rate—Number of deaths

of newborns in the first month of life per
1,000 births
Section B D 0 − 3 weeks
= ∗ 1000 or
Events Occurring During Pregnancy B
or Around the Time of Birth
D < 1 month
= ∗ 1000
B
33

Post-Neonatal Mortality Rate I nfant Mortality Rate (I MR)

� Post-Neonatal Mortality Rate—Number of � Note:

deaths of babies between the 1st and 12th – I MR = Neonatal mortality rate +
month of life per 1,000 births Post-neonatal mortality rate
D 4 - 51 weeks
= ∗ 1000 or
B
D1-11 months
= ∗ 1000
B
34 35

Fetal Death Fetal Death

� Early fetal loss: < 20 weeks � Let Df = Number of fetal deaths

� I ntermediate: 20– 27 weeks B = Number of births
� Late: 28+ weeks
1) Fetal death ratio 2) Fetal death rate

Df Df
= =
B B + Df

Continued 36 37
 Perinatal Mortality Rate Maternal Mortality

� Perinatal Mortality Rate—Number of � Maternal Mortality—Death of a woman

stillbirths and deaths of babies up to one while pregnant or within 42 days of
week old per 1,000 births termination of pregnancy, irrespective of
the duration or site of the pregnancy, from
(late fetal deaths + any cause related to or aggravated by the
early neonatal deaths) pregnancy or its management but not from
= ∗ 1000 accidental causes (WHO)
Live births

38 Continued 39

Maternal Mortality Maternal Mortality Ratio

� Let Dmc = Number of deaths due to � Maternal Mortality Ratio—Number of

maternal causes deaths due to maternal causes per 100,000
W 15-49 = Number of women of births
reproductive age D mc
B = Total live births = ∗ 100000
B

40 41

Maternal Mortality Rate Exercise

Maternal Mortality Ratio and Rate

� Maternal Mortality Rate—Number of deaths � Calculate the maternal mortality ratio and
due to maternal causes per 1,000 women rate for the U.S. (1990), based on the
of reproductive ages following data
D mc United States, 1990
= ∗ 1000 Births 4158212
W15 − 49 Maternal deaths 343
W15-49 65 624
� Note: Maternal mortality ratio is more You have 15 seconds to calculate the answer. You
widely used may pause the presentation if you need more time.
42 Source: Vital Statistics of the United States 1989-1991 43
 Exercise Answer Cause- Specific Morbidity
Maternal Mortality Ratio and Rate and Mortality
� The correct answer for the maternal � Let Dc = Number of deaths from cause
mortality indicator is as follows: (disease) c
Ratio: Rate: Cc = Number of cases of cause c
8.25 per 100,000 5.23 per 1,000 Nc = Number of new cases of cause
United States, 1990 (disease) c
Births 4158212 D = Total number of deaths
Maternal deaths 343 P = Mid-point population
W15-49 65 624
44 45

Cause-Specific Death Ratio Cause-Specific Death Rate

� Cause-Specific Death Ratio —Proportion of � Cause-Specific Death Rate—Number of

all deaths attributable to cause c deaths attributable to cause c per 100,000
population
Dc
= ∗ 100 Dc
D = ∗ 100000
P

46 47

Table: Selected CSDR in Egypt and U.S. 1987 Table: Selected CSDR in Egypt and U.S. 1987
Causes Egypt 1987 U.S. 1987 Causes Egypt 1987 U.S. 1987
Total 950.4 874.4
Cholera - 0.0 Other ischaemic heart diseases 16.0 106.5
Typhoid fever 0.2 0.0 Cerebrovascular disease 18.9 61.7
Other intestinal infectious diseases 82.1 0.2 Other diseases of circulatory system 243.9 101.4
Tuberculosis 2.6 0.7 Bronchitis, emphysema and asthma 29.5 9.2
Tetanus 7.2 0.0 Abortion 12.5 1.2
Septicemia 0.7 8.2 Direct obstetric causes 46.8 5.0
Malignant neoplasm of stomach 0.9 5.7 Indirect obstetric causes 5.9 0.4
Malignant neoplasm of colon 0.2 19.9 Congenital anomalies 9.0 5.1
Malignant neoplasm of rectum, 0.6 3.3 Birth trauma 0.5 6.8
rectosigmoid junction and anus Other conditions originating in the
Malignant neoplasm of trachea, 1.7 53.5 perinatal period 639.5 471.5
bronchus and lung Motor vehicle traffic accidents 6.6 19.5
Malignant neoplasm of female breast - 41.2
Suicide and self-inflicted injury 0.0 12.7
Malignant neoplasm of cervix uteri - 4.5
Homicide and injury purposely
All other malignant neoplasms 14.6 88.2 inflicted by other persons 0.5 8.6
Acute myocardial infarction 0.2 104.4 Other violence 20.4 1.4

Source: UN, Demographic Year book 1996 Continued 48 Source: UN, Demographic Year book 1996 49
 Case Fatality Rate Due to Cause
I ncidence of Cause (Disease) c
(Disease) c
� Ι(c)—Proportion of new cases of cause � Case Fatality Rate Due to Cause (Disease)
(disease) c in a population c—Proportion of persons with cause
Number of new cases of cause (disease) c who die from it
(disease) c in time (t, t+1)
= ————————————————— Dc
Mid-point population
=
Cc
Nc
=
P
50 51

Acute Causes (Diseases) Cause-Specific Death Rate

� Note: For acute causes (diseases), case � Therefore:

fatality = cause specific death rate =
Dc
incidence * case fatality
Nc � This relationship works well for acute
diseases but not for chronic ones
� So Dc N D
= c ∗ c
P P Nc

52 53

Calculation of potential years of life lost between ages 1 and 70 (PYLL), Rate and Age-
Person Years of Life Lost Adjusted Rate, Ontario, I schemic Heart Disease, Males 1974 (Source: Hetzel BS. I n: New
Developments in the Analysis of Mortality and Cause of Death. 1986
PYLL and Rate Standardized PYLL and Rage
Remaining No. of Age-adjusted
From Cause c (PYLL(c))
Age yrs deaths PYLL Correcting factor PYLL
1 to 4 67 0 0 1.08 0
5 to 9 62.5 0 0 1.02 0
10 to 14 57.5 1 57.5 1.03 59.2
70 15 to 19 52.5 1 52.5 1.05 55.1
20 to 024 47.5 3 142.5 1.03 146.8
25 to 29 42.5 9 382.5 0.97 371
PYLL ( c ) = ∑ ( 70 − a )m ac ∗ 1000
a=0 30 to 34 37.5 26 975 0.96 936
35 to 39 32.5 89 2,892.5 0.96 2,776.8
40 to 44 27.5 198 5,445 0.95 5,172.8
� Where c = Cause of death 45 to 49 22.5 489 11,002.5 0.94 10,342.4
50 to 54 17.5 772 13,510 0.95 12,834.5

a = Age at death 55 to 59 12.5 1,015 12,687.5 1 12,687.5

60 to 64 7.5 1,419 10,642.5 1 10,642.5
65 to 69 2.5 1,630 4,075 1.01 4,115.8
mac = Age-cause specific death rate Total (1 to 70) 5,652 61,865 60,140.4

61,865 / 3,791,600 X 1,000 60,140.4 / 3,791,600 X 1,000

= 16.3 per 1,000 = 15.9 per 1,000

54 55
 Summary

� Several indicators have been developed to

measure fetal/newborn mortality
� Maternal mortality can be measured as a
rate or as a ratio; the two indicators give Section C
different perspectives on the problem
� Cause-specific indicators are important in Relationships of Death Rates and
singling out the contribution of a particular Probability of Death, and Differential
disease or cause of death in a population in Mortality Between Populations

56

Death Rate and Probability Death Rate and Probability

of Death of Death
� Deriving probability of dying ( nqx) from observed � Also assume that
mortality rate ( nMx), using the actuarial method
– nDxt are constant over the years
Let nDxt = D = Deaths in age group (x, x+n) – Deaths are linearly distributed
in year t throughout the year
x
nP t = P = Mid-point population in age
group (x, x+n) in year t
n M x = M = Mortality rate in age group (x,
t
x+n) in year t
n M x = Dx / P x
t n t n t

Continued 58 Continued 59

Death Rate and Probability Death Rate and Probability

of Death of Death
t
n ∗D
nPx n qx = and
n
P+ D
2
n n n
D P+ D 1+ M
2 1 1 1 2
= 2 = + = 2 + nM =
q
n x n ∗ D nM 2 2nM nM
n
D
2 n∗M
so n qx =
n
t-4 t-3 t-2 t-1 t t+1 t+2 t+3 t+4 1+ M
2
Continued 60 61
 Percentage of Error in a Ratio for Given Levels of
Under and Overcount in Numerator and Denominator Differentials in Mortality

� I mportant variations in mortality are

associated with a number of socio-
economic characteristics:
– Ethnicity / race
• Blacks versus whites in the U.S.
– Marital status
• Single versus married

Source: Stan Becker, University of Dacca, August, 1978 (Not published) 62 Continued 63

Differentials in Mortality Differentials in Mortality

� Educational attainment � I mportant for understanding the physical

– Higher education consistently associated and sociological factors in health and for
with better health outcomes program planning
� Occupation
– Miner versus a white collar
� I ncome
– Rich versus poor

Continued 64 65

Summary
� The life table probability of dying ( nqx) can
be calculated from the observed mortality
rates ( nMx)
� One of the most common methods used to Standardization
derive nqx from nMx is the actuarial method
which assumes that deaths are linearly
Stan Becker, PhD
distributed throughout the year
Bloomberg School of Public Health
� I mportant variations in mortality are
associated with a number of socio-
economic characteristics
66
 Purpose of Standardization Purpose of Standardization

� Purpose of Standardization—Procedure of � This is necessary because:

adjustment of crude rates to eliminate from – Rates are affected by the demographic
them the effect of differences in population composition of the population for which
composition with respect to age and/or they are calculated
other variables – Age composition is a key factor
affecting crude rates

Continued 3 Continued 4

Purpose of Standardization Notation

� For purpose of comparison of rates over � ri = Rate for ith group in study population
time or from area to area, it is important to ni = Number of persons in ith group in study
determine the difference between the rates population
after taking into account the differences in n = Total number of persons in study
the composition of the populations population = ∑ni
pi = Proportion of persons in ith group in study
� Adjusted rates have no direct meaning in
population = ni/n = weight
themselves; they must be compared with
e = Number of events in study population =
the original crude rates or with other ∑ri* ni
adjusted rates using the same standard
5 Continued 6

Notation Direct Standardization

Ri = Rate for ith group in standard � Simplest and most straightforward
population technique
Ni = Number of persons in ith group in � Provides the best basis for determining the
standard population difference between two crude rates
N = Total number of persons in standard
population = ∑Ni � The rates from two or more study
populations are applied to a common
P i = Proportion of persons in ith group in
standard population = N i/N = weight population distribution (standard
population)
C = Crude rate in standard population

7 Continued 8
 Direct Standardization Standard Population

� Often directly standardized rates are � Standard Population—The standard

calculated for a series of populations using population can be any one of the study
the same standard populations, their average, or any other
� The series could be the same population at population distribution; the choice of the
different time points standard is arbitrary

9 Continued 10

Standard Population Direct Standardization

Formulas

� I f one of the populations is chosen, its

crude rates = its standardized rates ∑ ri Ni ∑ ri Ni ∑ Ri Ni
DSDR = ∑ ri Pi = = ∗ =
– The rates from the other study N ∑ Ri Ni N
population(s) are applied to the
distribution of the standard expected events in study population
� I f the average is chosen, the rates from = ∗C =
actual events in standard population
each study population are applied to that
new standard population
= comparative mortality figure (CMF) ∗ C
11 12

Direct Standardization of Mauritius Island:

Direct Standardization (Study) (Standard) Expected
Age group Rates M.I. per Population number of
1000 Mali births, M.I.
� Example—Direct Standardization Birth Rate
15-19 18 725719 13063
(DSBR) of Mauritius I sland’s (M.I .) 1985 20-24 58 574357 33313
crude birth rate using Mali's 1987 data as 25-29 57 536226 30565
30-34 36 443702 15973
standard
35-39 19 379184 7204
40-44 6 325824 1955
Total 2985012 102073

Total number of births, Mali: 375117

Total number of births, M.I.: 18247
CBR Mali: 48.7
CBR M.I. 18.5
13 Source: U.N. Demographic Yearbook 1986, 1992, and 1996 14
 Direct Standardization I ndirect Standardization

� I ndirect Standardization—The rates from a

Expected births in M.I.
DSBRM.I. = ∗ CBR standard population are applied to the
Actual births in Mali Mali
distribution of one or more study
populations
� Choice of standard rates is up to the
102073 demographer
= ∗ 48.7=13.3
375117

Source: U.N. Demographic Yearbook 1986, 1992, and 1996 15 16

I ndirect Standardization I ndirect Standardization

Formulas Standard Mortality Rate (SMR)

∑ ri ni � Notes on SMR:
e – The absolute value of the standardized
IDSR = ∗C = i ∗C =
∑ R i ∗ ni ∑ R i ni mortality ratio depends on the chosen
i i set of standard rates
– SMR has no meaning by itself—it should
actual events in study population be compared with other SMRs (relative
= ∗C =
expected events in study population values indicate higher or lower
standardized incidences of events)
= standardized mortality ratio (SMR) * C
17 18

Indirect Standardization of Mauritius Island:

I ndirect Standardization
Crude Birth Rate (Standard) (Study) Expected
Age group Rates Mali Population number of
per 1000 M.I. births, M.I.
� Example—I ndirect Standardization Birth 15-19 79 105764 8355
Rate (I SBR) of Mauritius I sland’s (M.I .) 20-24 159 109914 17476
1985 crude birth rate using Mali's 1987 25-29 171 94576 16172
30-34 140 81144 11360
data as standard 35-39 107 60063 6427
40-44 50 45825 2291
Total 497286 62082

Total number of births, Mali: 375117

Total number of births, M.I.: 18247

CBR Mali: 48.7

CBR M.I. 18.5
19 Source: U.N. Demographic Yearbook 1986, 1992, and 1996 20
 I ndirect Standardization I ndirect Standardization
Formulas

� I ndirect standardization can be deceptive

Observed births in M.I . � I t is to be used if:
I SBR M.I . = ∗ CBR
Expected births in M.I . Mali – There are no rates available for study
populations, i.e., only counts available
– Rates for study populations are not
18247
= ∗ 48.7 = 14.3 reliable because of small numbers of
62082 events or population

Source: U.N. Demographic Yearbook 1986, 1992, and 1996 21 Continued 22

I ndirect Standardization I ndirect Standardization

� Example— Comparison of crude and

Study Indirectly standardized rates
indirectly standardized rates for four Population CDR 1962 England and Wales Mexico
populations, using England and Wales or Rate Rank Rate Rank Rate Rank
Czechoslovakia 10.38 3 12.27 1 8.97 2
Mexico as standard Iceland 14.78 4 22.84 3 13.52 4
Poland 8.15 1 14.35 2 8.56 1
Thailand 9.15 2 23.90 4 10.49 3

Continued 23 Continued 24

I ndirect Standardization I ndirect Standardization

� Example— Comparison of directly and Standardized

indirectly standardized crude death rates Death Rates
for four countries, using the U.S. Country Crude Direct Indirect CMF SMR
population and crude death rate as Kuwait 5.15 11.7 13.64 1.34 1.56
New Zealand 7.93 9.16 9.13 1.04 1.04
standard (U.S. CDR = 8.75)
Singapore 5.18 9.46 9.10 1.07 1.05
Sri Lanka 6.00 11.6 11.80 1.27 1.34

Continued 25 26
 Comparing the Two Methods Comparing the Two Methods

� I n direct standardization, the weights are

⎡ ⎛ ⎞⎤ constant across study populations
DSDR = ∑ ri ∗ ⎢Ni ⎜ ∑ R ∗ N ⎟⎥ ∗ C
⎜ i i⎟ � I n indirect standardization, the weights (ni)
i ⎣⎢ ⎝i ⎠ ⎥⎦
are influenced by the distributions of the
study populations
⎡ ⎛ ⎞⎤
IDSR = ∑ ri ∗ ⎢ni ⎜ ∑ R ∗ N ⎟⎥ ∗ C
⎜ i i⎟
i ⎢⎣ ⎝ i ⎠ ⎦⎥

Continued 27 28

Exercise Exercise Answer

Direct and I ndirect Standardization Direct and I ndirect Standardization

� Calculate directly and indirectly standardized � The correct answers are as follows:
crude death rates for populations one and two by
using the standard population Population 1 Population 2
Population1 Population2 Standard Pop.
Age group Rate Prop. in group Rate Prop. in group Rate Prop. in group Directly standardized crude death rate:
(i) ( ri ) ( ni/n ) ( ri ) ( ni/n ) ( Ri ) ( Ni/N )
1 30 0.8 32 0.3 20 0.6 30* 0.6+15* 0.4=24.0 32* 0.6+16* 0.4=25.6
2 15 0.2 16 0.7 35 0.4
Indirectly standardized crude death rate:
CDR: 30*0.8+15*0.2=27 32*0.3+16*0.7=20.8 20*0.6+35*0.4=26
You have 15 seconds to calculate the answer. You (27/(20* 0.8+35* 0.2))* 26=30.5 (20.8/(20* 0.3+35* 0.7)* 26=17.7
may pause the presentation if you need more time.
Source: Vital Statistics of the United States, 1989– 1991 29 30

Adjustment for Two Factors Simultaneous Direct Adjustment

Distribution of mongoloids and total live births by maternal age and
birth order.
I ncidence rates of discovered mongolism by maternal age and birth order

Adapted from J oseph L. Fleiss, Statistical methods for rates and proportions, Second Edition, J ohn Wiley and Sons,
I nc. 1981. Data from Stark and Mantel (1966)

Adapted from J oseph L. Fleiss, Statistical methods for rates and proportions, Second Edition, J ohn Wiley and
Sons, I nc. 1981. Data from Stark and Mantel (1966)

31 32
 Summary

� Both methods of adjustment are used, but

it is preferable to use the direct method
(when possible)
� An adjusted rate has no meaning by itself; Life Table
it is only used for purpose of comparison
� The direct and the indirect methods can Stan Becker, PhD
lead to different interpretations: Be Careful Bloomberg School of Public Health

33

I ntroduction
� Although life tables were developed for the
study of mortality, they can be applied to
any other “failure” process so long as:
Section A – The process is measured in time, i.e.,
varies with age or some other measure
of duration
Use, Types, Graphs, and
I nterpretation of the Columns – The starting point of exposure can be
defined unambiguously
– The failure is an unambiguous event
with a clear location in time
Continued 4

I ntroduction I ntroduction

� Life tables are used by demographers, � Life tables are of particular value in the use
public health workers, actuaries, and many of data that may be affected by censoring
others in studies of mortality, longevity, or loss to follow-up
fertility, marriage, migration, and � The life table approach can make efficient
population growth, as well as in making use of partial data by including
projections of population, and in many observations up to the time of censoring
other areas

Continued 5 6
 History History

� J ohn Graunt’s Natural and Political � Antoine Deparcieux’s series of life tables
Observations upon the Bills of Mortality for annuitants and monastic orders (1746)
(1662) � Lowell Reed and Margaret Merrell’s new
– I mplications of information obtained technique to estimate the life table
from christening and death lists in probability of dying within an interval from
London the age-specific death rates (1939)
� Edmund Halley’s presentation of the
Breslau (Wroclaw, Poland) life table (1693)

Continued 7 Continued 8

History Types of Life Tables

� T.N.E. Greville’s development of another � Single decrement

technique to estimate the life table
probability of dying within an interval from
the age-specific death rate (1943)

9 Continued 10

Types of Life Tables Types of Life Tables

� Multiple decrement � Multi-state and increment-decrement

Continued 11 12
 Conventional Life Table Columns of a Life Table

� The life table conceptually traces a cohort � The basic columns of a life table are:
of newborn babies through their entire life
under the assumption that they are subject Age
to the current observed schedule of age- interval
specific mortality rates nqx lx ndx nLx Tx ex
(years)
� As age increases, the number of survivors
(x, x+n)
of the original group declines, the decline
being more rapid at ages where mortality
rates are high
13 14

Conventional Life Table Description of Columns

� Note on life table notation: � Age interval

– The subscript before a symbol defines – Age interval from exact age x to age
the width of the interval x+n, i.e., [x, x+n]
– The subscript after the symbol defines � n qx
the starting point of the interval – (Conditional) probability of dying in the
– lx, T x and ex never have a left subscript interval [x, x+n], given survival to age x
as they refer to exact age x

15 16

Conventional Life Table Conventional Life Table

Proportion Dying in Age Interval—nqx � lx or l(x)
Mexico, Females, 1983
– The first entry, for age 0, is called the
1.0
0.8 radix
0.6
0.4 – The lx column indicates the probability
0.2
Proportion

0.0
of survival to exact age x if the radix is
0
5
15
25
35
45
55
65
75

set at 1, or if the value of lx is divided

85 +

by the radix
Age Group
– i.e., p(x) = l(x) / l(0)

Source: Adapted from Keyfitz and Fleiger, 1990 Continued 17 Continued 18

 Conventional Life Table Conventional Life Table
Number Living at Beginning of Age Interval—lx
� Alternatively, the lx column can also Mexico, Females, 1983
indicate survivorship if the radix is set to
represent the number of initial births—the 100
80
column then shows the number of the 60
radix still alive by any exact age x 40
Thousands 20
– l20 = Number of survivors at exact age 0
0
5
15
25
35
45
55
65
75
85
95
105

20 (their 20th birthday)

Age Group
� I n this case, the radix is usually set at
100,000
Continued 19 Source: Adapted from Keyfitz and Fleiger, 1990 Continued 20

Conventional Life Table Conventional Life Table

� or d(x,n) Number Dying During Age Interval—ndx

ndx Mexico, Females, 1983
– Number of deaths to the radix between
15
exact ages x and x+n
– Since all members of the radix die 10

sooner or later, the sum of the ndx over 5

Thousands

all ages is equal to the radix 0

0
5
15
25
35
45
55
65
75
85
95
105

– I f radix = 1, ndx = (unconditional)

probability of dying between ages x and Age Group
x+n
Continued 21 Source: Adapted from Keyfitz and Fleiger, 1990 Continued 22

Conventional Life Table Conventional Life Table

� or L(x,n)
nL x � Note: nLx and T x are not expectations in the
– Person-years lived between exact ages usual sense since deaths before age x are
x and x+n implicitly included
• I f radix = 1, nLx = “expected” years
lived between ages x and x+n
� T x or T(x)
– Person-years lived above age x
• I f radix = 1, T x = “expected” years
lived after age x
Continued 23 Continued 24
 Conventional Life Table Conventional Life Table

� ex or e(x)
– Expectation of life at age x = Average
remaining lifetime for a person who
survives to age x
• e.g., e(0) = Life expectancy at birth =
number of years a person can expect
to live in his/her life

Source: Adapted from Keyfitz and Fleiger, 1990 Continued 25 Continued 26

Conventional Life Table Conventional Life Table

Average Remaining Lifetime—e(x)
Mexico, Females, 1983
� Notes:
– The six functions are generally
77.4
calculated and published for every life
51.6 table; however, some columns may be
omitted without a significant loss of

Years
25.8
information since the functions are
0.0 interrelated and some can be directly

0
5
15
25
35
45
55
65
75
85
95
105

calculated from the others

Age Group

Source: Adapted from Keyfitz and Fleiger, 1990 Continued 27 Continued 28

Mathematical Derivation and

Conventional Life Table
Relationships
� Notes: � Mathematical derivation and relationships
– I n general, the conditional probability of (cohort life table)
death ( nqx) is the basic function in the � Let X = Random variable for age at death
life table f(X) = Probability density function of X
F(X) = Cumulative density function of X
x
= ∫ f ( y )dy
0

29 Continued 30
 Mathematical Derivation and Mathematical Derivation and
Relationships Relationships
∞
� P{X> x} P{x < X < x + n} d(x, n)
= ∫ f ( y )dy = 1 − F(x ) � P{x<X<x+n|X>x} = =
x P{X > x} l(x)
= l(x) = q( x,n)
x +n
� P{x<X<x+n} = ∫ f ( y )dy = F( x + n) − F( x )
x � P(x,n) = 1-q(x,n) =
l(x, n)
= l(x) − l(x + n) l(x)
= d( x, n) � Note: P(x,0) = l(x)
Continued 31 Continued 32

Mathematical Derivation and Mathematical Derivation and

Relationships Relationships
∞ ∞
T(x) = ∫ l(y)dy
x
∫ l(y)dy
x
E(X | X > x) = x +
x +n l(x)
L(x, n) = ∫ l(y)dy = T(x) − T(x + n)
x
T(x)
∞ =x+
T( x ) = ∑ L( y, n) l(x)
y=x
T(x)
∞ ∞ E(X − x | X > x) = = e 0 (x)
0 l(x)
E(0) = ∫ xf(x)dx = ∫ l(y)dy = T0 = e (0)
0 0
Continued 33 Continued 34

Mathematical Derivation and Average Life Lived in I nterval

Relationships by Those Dying in Age I nterval [x,x+n]
x +n
∫ tf ( t )dt n dx
E( X − x | x < X > x + n) = t=x −x ∑ ( Xi − x )
x +n i =1
∫ f ( t )dt a( x,n)=n a x =
t=x n dx
L(x, n) − n ∗ l(x + n) � Where Xi = Exact age at which
=
d(x, n) a person dies
= a( x,n) x = Age at beginning of interval

35 Continued 36
 Average Life Lived in I nterval Mathematical Derivation and
by Those Dying in Age I nterval [x,x+n] Relationships

n L x = n ∗ l x + n + n ax ∗n dx
� Assuming that deaths are linearly
distributed in age interval, then:

n
L
n x = n ∗ l x+n + ∗n dx
2

37 Source: Adapted from Keyfitz and Fleiger, 1990 Continued 38

Mathematical Derivation and

Summary
Relationships
Age l0 = 10 1d6 =2
l7 � There are three main categories of life
cross l6 = 8 l7 = 6 tables: Decrement only vs. increment-
line decrement, single decrement vs. multiple
7
(6.5 − 6 ) + (6.7 − 6 ) decrement, and single state vs. multi-state
1a 6 = � The conventional life table traces a cohort
2
of newborn babies as they go through life
6 = 0 .6
assuming that they are subject to the
l6 cross
= 6 + 0 .6 ∗ 2 current observed schedule of age-specific
line 1L 6
mortality rates
= 7 .2 39 Continued 40

Summary

� There are seven basic columns in a life

table
� The six functions are generally calculated
and published for every life table, however, Section B
some columns may be omitted without
significant loss of information since the Construction of a Life Table and Use
functions are interrelated and some can be of Survivor Ratios
directly calculated from the others

41
 Construction Construction
Conventional Life Table Conventional Life Table

� Conventional life table � Excessive smoothing of the data by

– Assumption: Data on births, deaths, and mathematical methods would eliminate or
population are accurate reduce true variations in the age pattern of
– However, one of the most important mortality rates
aspects of the preparation of a life table � Therefore, it must be decided whether the
is the testing of the data for possible emphasis in the life table should be on its
biases and other errors; the level of
accuracy that can be tolerated depends closeness to the actual data or on its
mostly on the intended use of the life presentation of the underlying mortality
table picture after fluctuation has been removed
Continued 43 Continued 44

Construction Steps
Conventional Life Table Age-Specific Death Rates

� Life table functions, presumed for cohorts, � Steps

are derived from period mortality data – Derive the probability of dying ( nqx)
from the age-specific death rates by one
of various methods
– Calculate each entry of the l(x) and
d(x,n) columns
– Derive nLx, T x and ex columns

45 46

Observed Age-Specific
Relation of nMx and nmx
Death Rate
� Let D = Observed number of deaths in age � nMx = Observed age-specific death rate in
group x, x+n age group x, x+n
D
P = Mid-point population in age group =
P
x, x+n
ndx = Number of deaths in life table
� nmx = Life table death rate in age group x,
x+n
nLx = Life table population in age group d
x, x+n =n x
nL x

47 48
 Finding nqx from nmx Construction
since n L x =n ax ∗ l x + (n−n ax ) ∗ l x + n
� Finding nqx from nmx
ndx n mx ∗[ n ax ∗ l x + (n−n ax ) ∗ l x + n ]
n qx = n qx =
lx lx

n dx l x +n
n mx = ⇒ n dx =n m x ∗n L x =n mx ∗n ax + n mx ∗ (n−n ax ) ∗
nL x
lx
n mx ∗n L x l x +n
so n qx = since n qx = 1 −
lx lx
49 Continued 50

Construction Construction

n qx =n m x ∗n a x +n m x ∗ (n−n a x ) ∗ (1−n qx ) � When n=1 and nax = 0.5,

2 ∗ mx
Rearranging and bringing q to other side, qx = ⇒ qx < 1mx
2 + mx
= n m x ∗n a x + n∗n m x −n m x ∗n a x
n qx � When n=5 and nax = 2.5
[1+n m x ∗ (n−n a x )]
n∗n m x 5 ∗ mx
= 5 qx = ⇒ 5qx > 5mx
1 + (n−n a x )∗n m x 1 + 2 .5 * m x
Continued 51 52

Abridged Life Table Constant Risk Method

� The abridged life table is less burdensome � Finding nqx from Mx

to prepare and more convenient to use – Assumption: The force of mortality is
� Contains data by intervals of five to ten constant in (x, x+n) and equal to nMx
years of age
– Except the zero to one and one to four
age groups, which are retained n qx = 1 − e −n∗( n Mx )
separately
– Note: Constant risk of deaths ⇒ more
� Values for five or ten year intervals are deaths at beginning of interval
sufficiently accurate for most purposes
53 54
 Actuarial Method Reed-Merell’s Method
� Assumption: Population distribution is the � Assumption:
same as life table distribution and life table
nM x
= n mx
distribution is linear in (x, x+n)
empirical fit to find “a”
n∗n Mx
n qx = 3
n − n∗n m x − a∗n ∗n m 2x
1+ ∗n Mx n qx = 1 − e
2
� Note: a = 0.008 takes lack of fitness into
� Note: Same number of deaths through the
account
interval
55 56

Greville’s Method Sirken’s Method

� Assumption: nMx=nmx � Assumption: ngx is derived from a complete

l(x) is fit by cubic equation life table
n∗n Mx
= n mx n qx =
n qx 1+ n gx ∗n Mx
1 ⎡1 n ⎤
+n m x ∗ ⎢ + ∗ ( n m x − loge c )⎥
n ⎣ 2 12 ⎦
� Note: Since this method obtains the new
table by reference to a standard table, it
� Note: logec could be assumed to be about should only be used when mortality in both
0.095 tables is of a comparable level
57 58

Relation of ngx to nax Chiang’s Method

n 1 n 1 � Assumption: Population distribution is the
n gx = − = −
n qx n mx n dx n dx same as life table distribution
lx nL x
� Need nax

n ∗ l x n L x n ∗ l x −n L x n∗n m x
= − = n qx =
n dx n dx n dx 1 + (1−n a x ) ∗ n∗n m x
= n−n a x � Note: Chiang defines nax as the fraction of
an interval lived by those dying
Note : n−n gx =n a x
59 60
 Survival Ratios Survival Ratios

� One can use the life table to either � Project or look forward
“project” or “reverse survive” populations � Proportion surviving from birth to an age
interval [x, x+n]:
nLx l x +n 2
or
nl 0 l0
� e.g., Proportion surviving from birth to age
20– 24
5L 20 l 22.5
or
5l 0 l0
Continued 61 Continued 62

Survival Ratios Survival Ratios

� Proportion surviving from an age interval � Proportion surviving from an age interval
[x, x+n] to a fixed age “a”: [x, x+n] to an age interval [y, y+n]:
n ∗ la nL y

nL x
nL x
� e.g., Proportion surviving from age group
� e.g., Proportion of 60-year-olds who survive to 30– 34 to age group 35– 39
their 65th birthday
l 65 5 L 35
5 L 30
1 L 60 Continued 63 Continued 64

Survival Ratios Survival Ratios

� Reverse survive to recover a previous � Births represented by children now age two
population from a present one:
– i.e., Persons age y now (P y) represent
l0
∗ P2
how many persons age x, y– x years ago L
1 2

� Births in one year represented by children

nL x ∗ Py age zero to four
nL y l0
∗ P0 − 4
L
5 0
Continued 65 Continued 66
 Survival Ratios Summary

� Persons age 60– 64, 10 years ago � I n general, the conditional probability of
represented by persons now 70– 74: death, nqx, is the basic function from which
all other functions are derived
5 L 60 � nqx is derived from the age-specific death
∗ P70 −74 rates (there are various methods for
5 L 70 performing this derivation)
� Life tables can be used to project or
reverse survive population (in a projection,
the survival ratio is <1, whereas in reverse
survival, it is >1)
67 68

Stationary Population

� Characteristics
– A stationary population is defined as a
population whose total number and
Section C distribution by age do not change with
time
Stationary Population and Model Life
Tables

Continued 70

Stationary Population Stationary Population

� Assuming no migration, such a hypothetical � The characteristics of a stationary
population can be obtained if the number population are given by the corresponding
of births per year remained constant life table
(usually assumed at 100,000) for a long
period of time and each cohort of births
experienced the current observed mortality
rates throughout life
� The annual number of deaths would thus
equal 100,000 also, and there would be no
change in size of the population
Continued 71 72
 I nterpretation of Columns I nterpretation of Columns

� The columns “age interval,” “nqx,” and “ex” � n dx

have the same interpretation as in the = Number of persons who die each year
regular life table within the age interval
� lx � nL x
= Number of persons who reach the = Number of persons in the population
beginning of the age interval each year who at any moment are living within the
age interval

Continued 73 Continued 74

I nterpretation of Columns Examples of Calculation

� Tx � Crude birth rate = crude death rate

= Number of persons who at any moment
are living within the age interval and all l0
higher age intervals =
� T0 T0
= Total population � Since l0= Number of births, but also
number of deaths

75 76

Proportion of Population
Dependency Ratio
in Age Group
� Proportion of population in age group [x, x+n]
among those above age “a” P + P65 +
= <15
– Single-year life table P15 − 64
=
l x + 0.5
= 1L x
Ta Ta L + L
= 15 0 w 65
– Abridged life table 50 L15

= 5Lx � Where w = last age group

Ta 77 78
 Mean and Median Age of
Mean and Median Age at Death
Population
� Mean age of population � Mean age at death
w
w
∑ (x + 0.5) ∗ l x + 0.5 T0
x =0
= ∑ (x + 0.5) ∗ dx = = e0
= x =0 l0
T0
� Median age of population—age x such that � Median age at death—age x such that lx =
Tx = T0 / 2 0.5

79 80

Model Life Tables Model Life Tables

� Model life tables were first developed in the � A number of more flexible model systems
1950s by the United Nations have since been developed
� These first models were single parameter � Among them are the following:
models with no variation other than – Coale-Demeny Model Life Tables
mortality level – Brass Logit Life Table System
– United Nations Model Life Tables for
Developing Countries

Continued 81 82

Coale-Demeny Model Coale-Demeny Model

Life Tables Life Tables
� Based on 192 observed and evaluated life � Each level of mortality has tables for males
tables from the 19th and 20th centuries, and females
largely for European or European-origin � Thus, a given life table for one sex implies
populations a corresponding table for the other sex, the
difference in e(0) being an average of the
� A one-parameter system with 25 mortality differences observed in the observed tables
levels from expectations of life from 20
� The female advantage in e(0) ranges from
years to 80 years by 2.5 year intervals for about two years for the heaviest mortality
four distinct mortality age patterns called (lowest levels) to about 3.5 years for the
North, South, East, and West lowest mortality (highest levels)
Continued 83 Continued 84
 Coale-Demeny Model Coale-Demeny Model
Life Tables Life Tables
� Regional family—North � Regional family—South
– Based on northern European tables – Based on southern European tables
– For given level of expectation of life at – High mortality under age five (and high
birth, have low infant and high child child mortality relative to infant
mortality, below expected mortality mortality), low mortality from 40 to 65,
above age 50 high mortality above 65
– Based on only nine life tables – Based on 22 observed life tables

Continued 85 Continued 86

Coale-Demeny Model Coale-Demeny Model

Life Tables Life Tables
� Regional family—East � Regional family—West
– Based on eastern European tables – Based on all other tables i.e., 130
tables, including some from non
– Very high mortality in infancy relative to European populations (e.g. East Asia)
child and adolescent mortality, high
– Standard against which other families
mortality again after age 50 are compared
– Based on 31 observed tables – As the residual pattern, ‘West” is
sometimes recommended as the family
to choose if no other basis for choice
exists
Continued 87 88

Brass Logit Life Table System Brass Logit Life Table System

� Based on the expectation that, on a � Thus:

suitably transformed age scale, the ⎧1 - p(a) ⎫ ⎧⎪1 − p* (a) ⎫⎪
survivorship function p(a) of any life table 0.5 ∗ ln⎨ ⎬ = α + β ∗ 0 . 5 ∗ ln ⎨ * ⎬
⎩ p(a) ⎭ ⎪⎩ p (a) ⎪⎭
should be linearly related to the
� Where α effectively determines the level of
survivorship function of any other life table,
mortality in the observed life table relative
or of a standard, p* (a), by a level
to the standard
parameter α and a shape parameter β
� β determines the relative importance of
� The transformation used is the logit
child versus adult mortality in the observed
0.5* ln{[1-p(a)]/p(a)} relative to the standard tables
Continued 89 Continued 90
 Brass Logit Life Table System Brass Logit Life Table System

� The standard has α=0 and β=1 � Brass provides two standards:
� An overall mortality that is lower than the – A general standard
standard gives a negative α – “African” standard with heavier child
� Heavier adult relative to child mortality mortality (and particularly heavier child
gives β>1.0 mortality relative to infant mortality)
� And vice versa in both cases

Continued 91 Continued 92

Brass Logit Life Table System Brass Logit Life Table System

� Any reliable life table can be used as a � The logit system is very easy to use, and very
standard flexible given the range of standards that can be
adopted
� The model has two parameters (given a
� However, in most developing countries, one has
particular standard) and can be fitted to no basis on which to select a standard, and the
any two points of an observed life table choice of an inappropriate one can give poor
results
� Further, old age mortality is sensitive to the value
of β, and values very different from one can
produce implausible levels of old age mortality
Continued 93 94

United Nations Model Life United Nations Model Life

Tables for Developing Countries Tables for Developing Countries
� Developed in the early 1980s � Tabulated values of the models for each
� Five regional groupings of mortality regional grouping are available for males
patterns: and females separately for e(0) ranging
– Latin America from 35 to 75 years
– Chile
– South Asia
– Far East
– General
Continued 95 Continued 96
 United Nations Model Life United Nations Model Life
Tables for Developing Countries Tables for Developing Countries
� The models were developed using principal � The main advantages of the U.N. models
components analysis, and the mortality are that they are based on recent
level e(0) is merely the first component experience in developing countries and that
� Thus the models can be used as two- or they are flexible with the second and third
three-parameter models (the second parameters
parameter being the relation of mortality
under age five to over age five, and the
third being a childbearing parameter for
females and diverse factors for males)
Continued 97 Continued 98

United Nations Model Life United Nations Model Life

Tables for Developing Countries Tables for Developing Countries
� The main disadvantages are that the � I t should also be noted that the regional
regional patterns may reflect residual data groupings are not very precise
problems such as age misreporting, that
they provide a limited range of e(0)
(though adequate for most current
applications) and that the second and third
parameters may not be very useful since
their estimation requires heavy reliance on
uncertain data
Continued 99 100

Summary

� A stationary population has the same

number of births and deaths, so its size
does not change; the interpretation of the Measures of Marriage and
columns is slightly different than in the
conventional life table
Divorce
� Model life tables have been developed, the Stan Becker, PhD
most renown of which are those of the Bloomberg School of Public Health
Coale-Demeny, Brass, and the United
Nations
101
 Marriage

� Marriage—Legal union of persons of

opposite sex
� The legality of the union may be
Section A established by civil, religious, or other
means as recognized by the laws of each
I ndicators of Marriage country

Consensual Union Marriage

� Consensual Union—Establishment of a � Let M = Number of marriages

marital union without recorded legal P = Mid-year population
sanction P 15+ = Mid-year population age 15
and older
m
P 15+ = Number of men age 15 and
older
Pf15+ = W 15+ = Number of women age
15 and older
5 6

Crude Marriage Rate (CMR) General Marriage Rate (GMR)

� Crude Marriage Rate—Number of � General Marriage Rate—Number of

marriages per 1,000 population marriages per 1,000 population age 15 and
older
M
= ∗ 1000 M
P = ∗ 1000
P15 +

7 8
 General Marriage Rate for
Marriage
Women and Men
� I mportant note: � General marriage rate for women (GMR f)
– Most measures of marriage are sex- M
GMR f = f
∗ 1000 ≈ 2 ∗ GMR
specific P15 +
– I n the remaining measures, whenever � General marriage rate for men (GMR m)
the rates are calculated for women, the
same calculations can be done for men M
GMR m = m ∗ 1000
P15 +
� Note: GMRf ≠ GMRm
9 10

Exercise Exercise Answer

General Marriage Rate General Marriage Rate

� Calculate the general marriage rate for � The correct answers for the general marriage
women and compare it to the general rates are as follows:
marriage rate for Brazil (1988) based on GMR GMRf
the following data 10.24 20.37
Brazil, 1988
Brazil, 1988
Number of marriages : 951 236
Total population 15+ : 92 852 000 Number of marriages : 951 236
Total female population 15+: 46 706 000 Total population 15+ : 92 852 000
Total female population 15+: 46 706 000
You have 15 seconds to calculate the answer. You
may pause the presentation if you need more time.
Source: U.N. Demographic Yearbook, 1989 and 1990 11 12

Age-Specific Marriage Rate Exercise

(ASMR) Age-Specific Marriage Rate

� ASMR—Number of marriages per 1,000 women � Calculate the age-specific marriage rate for
(or men) of age “a”
women 25– 29 for Brazil (1988) based on
Ma the following data:
= ∗ 1000 Brazil, 1988
Wa
Number of marriages, women 25-29: 161 086
� Where M a = Number of marriages to women Female population 25-29: 6 362 000
of age “a”
W a = Mid-year population of women You have 15 seconds to calculate the answer. You
of age “a” may pause the presentation if you need more time.
13 Source: U.N. Demographic Yearbook, 1990 14
 Exercise Answer Order-Specific Marriage Rate
Age-Specific Marriage Rate (OSMR)
� The correct answer for the age-specific � OSMR—Number of order “i” marriages per 1,000
marriage rate for women 25– 29 for Brazil persons age 15 and older with marriage order
(1988) is as follows: “i-1”

– 25.32
Mi
= i−1 ∗ 1000
Brazil, 1988 P15+
Number of marriages, women 25-29: 161 086 � Where Mi = Number of marriages of order
Female population 25-29: 6 362 000 “i”
i-1
P 15+= Mid-year population age 15+
with marriage order “i-1”
15 Continued 16

Order-Specific Marriage Rate First Order-Specific

(OSMR) Marriage Rate
� Note: � First order-specific marriage rate is used
– OSMR is normally sex-specific for a nuptiality life table
1
– I t can also be age-specific M
s
∗ 1000
OSMR for women OSMR for men P15 +
Mi Mi � Where M1 = Number of first marriages
∗ 1000 m,i −1
∗ 1000
i−1 P s15+ = Mid-year population of
W15+ P15 + never-married (single) persons
w � Note: I s usually calculated sex-specific
∑ OSMR ≠ GMR
i =1
17 18

Age-Order Specific Marriage Age-Order Specific Marriage

Rate (AOSMR) Rate (AOSMR)
� Let Mia = Number of marriages of order “i” � AOSMR—Number of marriages of order “i”
to women (or men) of age “a” per 1,000 women of age “a” who are at
i-1
P a = Mid-year population of women their “i-1th” marriage
(or men) of age “a” who are at
marriage order “i-1” Mia
= ∗ 1000
Wai-1

Continued 19 20
 Some Relationships Total Marriage Rate (TMR)

� Note that: ∑ AOSMR ≠ ASMR � Total Marriage Rate—Total number of

i marriages a person will have at the end of
Mia Wai-1 his/her marriageable age if he/she follows
– Since: ASMR = ∑ i-1
∗ the given schedule of marriage
i Wa Wa
65 +Ma
= ∑
� Also that: ∑ AOSMR ≠ OSMR a =15 Pa
a

– Since: OSMR = ∑
Mia
∗
Wai-1 � Note: I s sex-specific
i-1 W15 +
a Wa
21 22

Total First Marriage Exercise

Rate (TMFR) Total First Marriage Rate (TMFR)

65 + M1 � Calculate the total first marriage rate for women

TFMR = ∑ a from Brazil (1988) based on the following data
a Pa Age 1st
Group Marriages Population
TFMR ≤ 1.0 for cohort rate, -20 311 233 32 885 000
but period rate can be above 1.0 20-24 346 361 6 875 000
25-29 161 086 6 362 000
30-34 61 085 5 443 000
� Note: TMRm > TMR f 35-39 29 653 4 479 000
40-44 16 503 3 590 000
TMR - TFMR = TRMR (Total remarriage rate) 45+ 25 315 12 735 000
You have 15 seconds to calculate the answer. You
may pause the presentation if you need more time.
23 Source: U.N. Demographic Yearbook, 1990 24

Exercise Answer Singulate Mean Age

Total First Marriage Rate (TMFR) at Marriage (SMAM)
� The correct answer for the age-specific � Estimate of the mean age at first marriage
marriage rate for women 25– 29 for Brazil approximated by indirect method from
(1988) is as follows: cross-sectional data on marital status by
– 0.55 age
Age 1st
Group Marriages Population – Mean age at marriage of women
-20 311 233 32 885 000 marrying before they reach 50
20-24 346 361 6 875 000
25-29 161 086 6 362 000
30-34 61 085 5 443 000
35-39 29 653 4 479 000
40-44 16 503 3 590 000
45+ 25 315 12 735 000
25 Continued 26
 Singulate Mean Age Singulate Mean Age
at Marriage (SMAM) at Marriage (SMAM)
� Basic assumptions: � Let Si, Sj = Proportions of women single at
– The change in the proportion single ages i, j, then
from age “x” to age “x+1” is a measure ⎡⎛ 49 54 ⎞ ⎤
⎢ ⎜ ∑ Si + ∑ S j ⎟ ⎥
of the proportion of a birth cohort who ⎡⎛ 49 ⎞ ⎤ i= 45 j=50 ⎟
⎜ ∑ Si ∗ 5 ⎟ + 1,500 − ⎢⎜ ∗ 50 ⎥
married at that age if no woman dies ⎢⎣⎝ i=15 ⎠ ⎥⎦ ⎜
⎢ 2 ⎟ ⎥
⎟
between her 15th and 55th birthday SMAM =
⎢⎣⎜⎝ ⎠ ⎥⎦
⎛ 49 54 ⎞
– The risk of marriage has remained ⎜ ∑ Si + ∑ S j ⎟
i = 45 j = 50 ⎟
constant (otherwise we estimate the 100 − ⎜
⎜ 2 ⎟
mean for some average cohort) ⎜ ⎟
⎝ ⎠
Continued 27 28

Other Measures Standardization

� Mean and median ages (of first marriage, � Most of the marriage measures can and
of remarriages) should be standardized for comparisons
� Percent single � Can be done by direct or indirect method
� Age of groom by age of bride � Used for comparative marriage analyses
� Duration of marriage � Example: Age-standardized first marriage
rate of the U.S. (1960), using England and
Wales (1961) as standard-direct method

29 30

Male Single Male Expected

First Population First
Gross Nuptiality Table
Age Marriage England Marriages
Group Rates U.S. and Wales U.S.A � Assumes no person dies before passing
15-19 0.031 1,604,910 49,752 through the marriageable ages
20-24 0.210 989,661 207,829
25-29 0.190 425,643 80,872 � Same techniques as in life table
30-34 0.111 262,656 29,155 construction
35-44 0.060 373,918 22,435 � Allows one to determine what proportion of
45-64 0.019 498,549 9,472 a cohort of single persons would be
65+ 0.011 158,846 1,747
married at various ages assuming that the
Total Crude = 70.7 4,314,183 401,262
marriage rates used continues to prevail
U.S. age-standardized first marriage rate= and there is no mortality
401,262 / 4,314,183 x 1,000 = 93
31 32
 Net Nuptiality Table Net Nuptiality Table

� Takes into account mortality as well as � Also gives the probability of a single person
marriage marrying at each year of age according to
� I ndicates the pace at which a group of the current nuptiality and mortality rates
single persons is decreased annually by � Provides information on the average age at
marriage and death marriage
� Multiple-decrement life table techniques
are needed

Continued 33 34

Summary Summary

� Marital status provides a static � Marriage rates directly measure changes in

representation of the population with population composition characteristics
respect to its marital composition rather than changes in population size
� Measures on marriage focus on the � But if the change in population size is
considered a function of the broader
dynamic aspect process of reproduction—which in most
� Marriage information are mainly derived societies occurs through formation of
from vital statistics whereas marital status families—then the rates of family formation
is primarily analyzed from census and and dissolution are part of population
survey data dynamics
Continued 35 36

Divorce

� Divorce—Separation of the husband and

wife by a judicial decree which confers on
the parties the right to remarriage
Section B

I ndicators of Divorce

38
 Crude Divorce Rate General Divorce Rate

� Crude Divorce Rate—Number of divorces � General Divorce Rate—Number of divorces

per 1,000 population per 1,000 persons age 15 and more
Div Div
= ∗ 1000 = ∗ 1000
P P15 +

� Where Div = Number of divorces

� Where Div = Number of divorces P 15+ = Mid-year population 15 and
P = Mid-year population older
39 40

Exercise Exercise Answer

Crude and General Divorce Rates Crude and General Divorce Rates

� Calculate the crude divorce rate and the � The correct answers for the crude divorce rate for
general divorce rate for Brazil (1988) based Brazil (1988) are as follows:
on the following data: – Crude Divorce Rate = 6.45 divorces per
1,000 population
Brazil, 1988 – General Divorce Rate = 10.03 divorces
Number of divorces : 33,437 per 1,000 persons age 15+
Total population 15+ : 92,852,000 Brazil, 1988
Total population : 144,428,000
Number of divorces : 33,437
You have 15 seconds to calculate the answer. You Total population 15+ : 92,852,000
may pause the presentation if you need more time. Total population : 144,428,000
Source: U.N. Demographic Yearbook, 1990 41 42

Divorce Rate Among Married

Age-Specific Divorce Rate
Couples
� Divorce Rate Among Married Couples— � Age-Specific Divorce Rate—Number of
Number of divorces per 1,000 married divorces per 1,000 women (or men) of age
persons “a”
Div Div f
= ∗ 1000 = f a ∗ 1000
Pm Pa
� Where P m = Mid-year population of
� Where Divfa = Number of divorces by
married persons women age “a”
� Note: Can be sex-specific f
P a = Mid-year population of
women age “a”
43 44
 Exercise Exercise Answer
Age-Specific Divorce Rates Age-Specific Divorce Rates

� Calculate the age-specific divorce rates for � The correct answers for the age-specific divorce
women 30– 34 and 40– 44 for Brazil (1988) rate are as follows:
based on the following data – 1.23 divorces per 1,000 women 30– 34
Brazil, 1988 – 1.34 divorces per 1,000 women 40– 44

Age Group Divorces Population Brazil, 1988

30–34 6,689 5,443,000 Age group Divorces Population
40–44 4,825 3,590,000 30–34 6,689 5,443,000
You have 15 seconds to calculate the answer. You 40–44 4,825 3,590,000
may pause the presentation if you need more time.
Source: U.N. Demographic Yearbook, 1990 45 46

Duration-Specific Divorce Rate Duration-Specific Divorce Rate

� Let Divd = Number of divorces to � Duration-Specific Divorce Rate—Number of

persons who have been divorces per 1,000 persons who have been
married for a duration d married for a duration “d”
� P dmar = Mid-year married population
of persons who have been Div d
= d
∗ 1000
married for a duration d Pmar

Continued 47 48

Order-Specific Divorce Rate Order-Specific Divorce Rate

� Let Divi = Number of divorces of order � Order-Specific Divorce Rate—Number of

“i” divorces of order “i” per 1,000 persons who
P imar = Mid-year married population have been married “i” times
of persons who have been
married “i” times Div i
Pi-1div = Mid-year population of = i ∗ 1000
persons divorced “i-1” times Pmar
� Note: I t may be hard to get the
denominator
Continued 49 Continued 50
 Standardization of
Order-Specific Divorce Rate
Divorce Rates
� Alternatively: � Crude and general divorce rates can be
– Number of divorces of order “i” per standardized
1,000 persons who have divorced “i-1” � Can be done by direct or indirect method
times � Used for comparative divorce analyses
i
Div
= i-1 ∗ 1000
Pdiv

51 52

Estimated Marital Duration

Comparison of Crude Annual Divorce Rates
Effects for Divorce

Adapted from J oshua R. Goldstein, The Leveling of Divorce in the Adapted from J oshua R. Goldstein, The Leveling of Divorce in the
United States, Demography, Volume 36 (3), August 1999 53 United States, Demography, Volume 36 (3), August 1999 54

Proportion of Marriages that

U.S. Annual Divorce Rate
Will End in Divorce

Adapted from Andrew J . Cherlin, Marriage, Divorce, Remarriage. Adapted from Andrew J . Cherlin, Marriage, Divorce, Remarriage.
Harvard University Press, 1992 55 Harvard University Press, 1992 56
 Summary

� Measures on marriage and divorce focus on

the dynamic aspect
� Divorce information is mainly derived from
vital statistics Fertility and I ts Measurement
� Divorce rates measure changes in
population composition characteristics Stan Becker, PhD
rather than changes in population size Bloomberg School of Public Health

57

Definitions

� Fecundity—Physiological capacity to
conceive
� I nfecundity (sterility) —Lack of the capacity
Section A to conceive
– Primary sterility—Never able to produce
I ndicators of Fertility Based a child
on Vital Statistics
– Secondary sterility—Sterility after one or
more children have been born

Continued 4

Definitions Definitions

� Fecundability—Probability that a woman � Reproductivity—Extent to which a group is

will conceive during a menstrual cycle replacing its own numbers by natural
� Fertility (natality) —Manifestation of processes
fecundity � Gravidity— Number of pregnancies a
� I nfertility—I nability to bear a live birth woman has had
� Natural fertility—Fertility in the absence of � Parity—Number of children born alive to a
deliberate parity-specific control woman

Continued 5 Continued 6
 Definitions Crude Birth Rate (CBR)

� Birth interval—Time between successive � Let B = Number of births

live births � Let P = Mid-year population
� Pregnancy interval—Time between � Let W 15-44 = Number of women of
successive pregnancies of a woman reproductive ages

7 Continued 8

Crude Birth Rate (CBR) Exercise

Crude Birth Rate (CBR)

� Crude Birth Rate—Number of births per � Use the following data to calculate the CBR
1,000 population per 1,000
B Island of Mauritius, 1985
= ∗ 1000
P Total Births: 18,247
B W15 − 44 Total female population: 491,310
= ∗ ∗ 1000 Total male population: 493,900
W15 - 44 P
You have 15 seconds to calculate the answer. You
may pause the presentation if you need more time.
9 Source: U.N. Demographic Yearbook, 1986 10

Exercise Answer Crude Birth Rate (CBR)

Crude Birth Rate (CBR)

� The correct answer is as follows: � Crude birth rates can be standardized using
– 18.5 births per 1,000 population the direct or the indirect method
� Example: Direct (DSBR) and indirect (I SBR)
standardization of the I sland of Mauritius
Island of Mauritius, 1985 (I .M.) 1985 crude birth rate using Mali's
Total Births: 18,247 1987 data as standard
Total female population: 491,310
Total male population: 493,900

11 12
 Direct Standardization of Birth I ndirect Standardization of
Rate for Mauritius I sland Birth Rate for Mauritius I sland
(Study) (Standard) Expected (Standard) (Study) Expected
Age group Rates I.M. per Population number of Age group Rates Mali Population number of
1000 Mali births, I.M. per 1000 I.M. births, I.M.
15-19 18 725719 13063 15-19 79 105764 8355
20-24 58 574357 33313 20-24 159 109914 17476
25-29 57 536226 30565 25-29 171 94576 16172
30-34 36 443702 15973 30-34 140 81144 11360
35-39 19 379184 7204 35-39 107 60063 6427
40-44 6 325824 1955 40-44 50 45825 2291
Total 2985012 102073 Total 497286 62082

Total number of births, Mali: 375117 Total number of births, Mali: 375117
Total number of births, I.M.: 18247 Total number of births, I.M.: 18247

CBR Mali: 48.7 CBR Mali: 48.7

CBR I.M. 18.5 CBR I.M. 18.5

Source: U.N. Demographic Yearbook, 1986, 1992, 1996 13 Source: U.N. Demographic Yearbook, 1986, 1992, 1996 14

Expected births in I.M. General Fertility Rate (GFR)

DSBR I.M. = ∗ CBR
Actual births in Mali Mali
102073 � General Fertility Rate—Number of births
= ∗ 48.7 = 13 .3
375117 per 1,000 women of reproductive ages
B
Observed births in I.M. = ∗ 1000
ISBR I.M. = ∗ CBR Mali W15 - 44
Expected births in I.M .
18247
= ∗ 48 .7 = 14 .3 GFR ≈ 4 ∗ CBR
62082
CBR I.M. = 18.5
Source: U.N. Demographic Yearbook, 1986, 1996 15 16

Exercise Exercise
General Fertility Rate (GFR) General Fertility Rate (GFR)

� Use the following data to calculate the GFR � The correct answer is:
per 1,000 women aged 15– 44: – 73.7 births per 1,000 women aged
Island of Mauritius, 1985
15-44
Age Group Women Island of Mauritius, 1985
15-19 52 013 Age Group Women
20-24 54 307 15-19 52 013
25-29 46 990 20-24 54 307
30-34 40 211 25-29 46 990
35-39 30 401 30-34 40 211
40-44 23 496 35-39 30 401
Total births: 18 247 40-44 23 496
You have 15 seconds to calculate the answer. You may pause the Total births: 18 247
presentation if you need more time.
Source: U.N. Demographic Yearbook, 1986 Continued 17 18
 Age-Specific Fertility Age-Specific Fertility
Rate (ASFR) Rate (ASFR[a, n])
� Let Ba = Number of births to women of � ASFR(a, n) —Number of births per 1,000
age (group) “a” women of a specific age (group)
W a = Number of women of age
(group) “a” B
= Fa = a ∗ 1000
n = Number of years in age group Wa

� I f n = 1 , then write ASFR(a)

� Example: ASFR Poland 1984

19 20

Exercise
Age-Specific Fertility Rates Age-Specific Fertility Rate (ASFR[a, n])
Poland, 1984
� Use the following data to calculate the
ASFR per 1,000 for women age 20– 24 and
200 25– 29
150 Island of Mauritius, 1985

100 Age Group

of Mother Women Births
50 15-19 52 013 1884

Rate (per
thousand)
20-24 54 307 6371
0
25-29 46 990 5362
15-19 20-24 25-29 30-34 35-39 40-44 45-49 30-34 40 211 2901
Age Group 35-39 30 401 1170
40-44 23 496 268
21 Source: U.N. Demographic Yearbook, 1986 22

Exercise Answer Total Fertility Rate (TFR)

Age-Specific Fertility Rate (ASFR)

� The correct answers are: � Total Fertility Rate—Number of children a

– ASFR(20,5) = 117.3 births per 1,000 women 20– 24
– ASFR(25,5) = 114.1 births per 1,000 women 25– 29
woman will have if she lives through all the
reproductive ages and follows the age-
Island of Mauritius, 1985 specific fertility rates of a given time period
Age Group (usually one year)
of Mother Women Births
15-19 52 013 1884
20-24 54 307 6371
25-29 46 990 5362
30-34 40 211 2901
35-39 30 401 1170
40-44 23 496 268
23 Continued 24
 Total Fertility Rate (TFR) Total Fertility Rate (TFR)

� For single-year age groups � Example: ASFR and TFR—Poland, 1984

44 B
TFR = ∑ a ∗ 1000 = ∑ ASFR(a) = ∑ F Age group Ba Wa ASFR
a 15-19 43807 1230396 35.60
a = 15 Wa 20-24 257872 1390077 185.51
25-29 236088 1653183 142.81
� For five-year age groups 30-34 115566 1608925 71.83
35-39 38450 1241967 30.96
40− 44 B 40- 44 40-44 6627 941963 7.04
TFR = 5 ∗ ∑ a ∗ 1000 = 5 *
∑ ASFR(a,5) 473.74
a=15−19 W a=15-19 TFR= (5 * 473.74) / 1000 = 2.4
a
Continued 25 26

Exercise Exercise Answer

Total Fertility Rate (TFR) Total Fertility Rate (TFR)

� Use the following data to calculate the TFR � The correct answer is as follows:
per 1,000 – TFR = 1.9 children per woman
Island of Mauritius, 1985 Island of Mauritius, 1985
Age Group Age Group
of Mother Women Births of Mother Women Births
15–19 52,013 1,884 15–19 52,013 1,884
20–24 54,307 6,371 20–24 54,307 6,371
25–29 46,990 5,362 25–29 46,990 5,362
30–34 40,211 2,901 30–34 40,211 2,901
35–39 30,401 1,170 35–39 30,401 1,170
40–44 23,496 268 40–44 23,496 268

Source: U.N. Demographic Yearbook, 1986 27 28

Mean of Age of Childbearing Variance of Age of Childbearing

� For single-year groups 2

44 44⎛ _⎞
⎜
_ ∑ (a + 1 2 ) Fa ∑ a − a ⎟ Fa
a = 15 ⎟
a= 44 2 a = 15 ⎜⎝ ⎠
∑ F s = 44
a = 15 a
∑ Fa
� For five-year age groups a = 15
_ ∑ (a + 2 . 5 ) Fa
a=
∑ Fa
29 30
 Exercise Exercise Answer
Mean and Variance of Age of Childbearing Mean and Variance of Age of Childbearing

� Use the following data to calculate the � The correct answers are as follows:
mean and variance of age of childbearing – Mean age of childbearing = 27.9 years
Island of Mauritius, 1985
– Variance of age of childbearing = 38.2
Age Group years
of Mother Women Births Island of Mauritius, 1985
15Ğ19 52,013 1,884 Age Group
20Ğ24 54,307 6,371 of Mother Women Births
25Ğ29 46,990 5,362 15Ğ19 52,013 1,884
30Ğ34 40,211 2,901 20Ğ24 54,307 6,371
25Ğ29 46,990 5,362
35Ğ39 30,401 1,170 30Ğ34 40,211 2,901
40Ğ44 23,496 268 35Ğ39 30,401 1,170
Source: U.N. Demographic Yearbook, 1986 31 40Ğ44 23,496 268 32

Mean and Median Age of Exercise

Mothers Mean and Median Age of Mothers

� Mean: Median x such that: � Use the following data to calculate the
β x mean and median age of mothers
1 ∑ Ba Island of Mauritius, 1985
a a =15 = 0 .5 Age Group
∑(a + 2) ∗ B
a =α 44 of Mother Women Births
∑ Ba 15–19 52,013 1,884
∑B a
a =15 20–24 54,307 6,371
25–29 46,990 5,362
� Ba = Number of births to women age a 30–34 40,211 2,901
35–39 30,401 1,170
40–44 23,496 268
33 Source: U.N. Demographic Yearbook, 1986 34

Exercise Answer Parity

Mean and Median Age of Mothers

� The correct answers are as follows: � Mean Median x such that:

– Mean age of mothers = 26.9 years
– Median age of mothers = 24.7 years I=max( i) x
Island of Mauritius, 1985
∑ Mi ∑ mi = 0 .5
Age Group
i =1 i =1
of Mother Women Births
15Ğ19 52,013 1,884 � Mi = Proportion of women at or above parity
20Ğ24 54,307 6,371
25Ğ29 46,990 5,362 i
30Ğ34 40,211 2,901
35Ğ39 30,401 1,170 � mi = Proportion of women at parity i
40Ğ44 23,496 268
35 36
 General Marital Fertility
Marital Fertility Rate (MFR)
Rate (GMFR)
� Let Bm = Number of marital births � General Marital Fertility Rate—Number of
Bu = Number of non-marital births births per 1,000 married women of
m
W 15– 44 = Number of married women reproductive ages
of age 15– 44 B
= m
∗ 1000
W u15– 44 = Number of unmarried W15
women of age 15– 44 - 44

37 38

“Out-of-Wedlock” (Non-Marital)
Marital Fertility Rate (MFR)
Fertility Rate
� Marital Fertility Rate—Number of marital � “Out-of-Wedlock” (Non-Marital) Fertility
births per 1,000 married women of Rate— Number of non-marital births per
reproductive ages 1,000 unmarried women of reproductive
ages
Bm Bu
= m
∗1000 = ∗1000
W15 u
-44 W15-44

39 40

Age-Specific Marital
Some Relationships
Fertility Rate (ASMFR)
� Let Bma = Number of marital births to
Bm Bu B women of age group “a”
m
+ u
≠
m
W a = Number of married women in
W15-44 W15-44 W15-44
age group “a”
m
B (d) = Number of marital births to
� But
women married for “d” years
m u
Bm W15 -44 Bu W15-44 B W m(d) = Number of women married for
m
∗ + u ∗ =
W15-44 W15-44 W15-44 W15-44 W15-44 “d” years

41 Continued 42
 Age-Specific Marital Duration (of Marriage)—
Fertility Rate (ASMFR) Specific Fertility Rate (DSFR)
� Age-Specific Marital Fertility Rate— Number � DSFR—Number of marital births per 1,000
of marital births per 1,000 married women women who have been married for
of age (group) “a” duration “d”
Bmla Bl(d)
= ∗1000 = ∗ 1000
Wm m
a W (d)

43 44

Order-Specific Fertility Order-Specific Fertility

Rate (OSFR) Rate (OSFR)
� Let Bi = Number of births of order “i”, � Order-Specific Fertility Rate—Number of
i>0 order “i” births per 1,000 women of
Bia = Number of order “i” births to reproductive ages
women in age group “a”
Wa = Number of women in age Bi
= ∗ 1000
group “a” W15 - 44
W 15-44 = Number of women of age
15– 44 (or 15– 49) � Example: OSFR Poland 1984
Continued 45 46

Exercise
Order-Specific Fertility Rates Order-Specific Fertility Rate (OSFR)
Poland, 1984
� Use the following data to calculate the
30 OSFR for birth orders 1 and 3
20 Island of Mauritius, 1985

10
Age Group Number of Birth Order
of Mother Women 1 3
Rate (per
thousand)

0 15Ğ19 52,013 1,521 40

t d d h h h h h h + 20Ğ24 54,307 3,317 678
1s 2n 3r 4t 5t 6t 7t 8t 9t 0th
1 25Ğ29 46,990 1,638 1,132
30Ğ34 40,211 496 665
Birth Order 35Ğ39 30,401 142 215
40Ğ44 23,496 24 30
47 Source: U.N. Demographic Yearbook, 1986 48
 Exercise Answer Age-Order Specific Fertility
Order-Specific Fertility Rate (OSFR) Rate (AOSFR [a,I ])
� The correct answers are as follows: � AOSFR(a,i) —Number of order “i” births per
– OSFR(1) = 28.8 births of order 1 per 1,000 women of age (group) “a”
1,000 women 15– 44
– OSFR(3) = 11.2 births of order 3 per
Bia
1,000 women 15– 44 = ∗ 1000
Island of Mauritius, 1985 Wa
Age Group Number of Birth Order
of Mother Women 1 3
15–19 52,013 1,521 40 � Example: AOSFR Poland 1984
20–24 54,307 3,317 678
25–29 46,990 1,638 1,132
30–34 40,211 496 665
35–39 30,401 142 215
40–44 23,496 24 30 49 50

Exercise
Age-Order Specific Fertility Rates Age-Order Specific Fertility Rate (AOSFR)
Poland 1984 1st birth
2nd
3rd � Use the following data to calculate the AOSFR
120
4th +
for birth order 3 in age group 25– 29
100

80 Island of Mauritius, 1985

60 Age Group Birth order
of Mother Women 1 3
40
15Ğ19 52,013 1,521 40
20 20Ğ24 54,307 3,317 678
25Ğ29 46,990 1,638 1,132
0
30Ğ34 40,211 496 665
-20 20-24 25-29 30-34 35-39 40-44 45+
35Ğ39 30,401 142 215
Age 40Ğ44 23,496 24 30
51 Source: U.N. Demographic Yearbook, 1986 52

Exercise Answer Age-Order Specific Fertility

Age-Order Specific Fertility Rate (AOSFR) Rate (AOSFR[a,i])
� The correct answer is as follows: � Note:
– AOSFR = 24.1 births of order 3 per ∞
1,000 women 25– 29 ∑ AOSFR (a, i ) = ASFR ( a )
i =1
Island of Mauritius, 1985
Age Group Birth order
of Mother Women 1 3
15–19 52,013 1,521 40 44 Wa
20–24 54,307 3,317 678 ∑ AOSFR (a, i ) ∗ = O SFR(i)
25–29 46,990 1,638 1,132 a =15 W15 - 44
30–34 40,211 496 665
35–39 30,401 142 215
40–44 23,496 24 30
53 54
 Cumulative Order Specific
Birth Probability
Birth Rate (COSFR[i,a])
� Cumulative up to age “a” � Let Bi = Number of births of order “i”, i>0,
� COSFR(i,a) —Total number of order “i” in year “t”
births per 1,000 women of age (group) less W i-1 = Number of women of parity
than or equal to “a” “i-1” at beginning of year “t”

a
= ∑ AOSFR(i, x)
x=0

55 56

Birth Probability, BP(i) Birth Probability

� Birth Probability—Probability of having an � May be age-specific as well

“ith” order birth in a given year for women � Birth probabilities are the most sensitive
who already have “i-1” births indicators of temporal change in the pace
of childbearing
Bi
=
W i-1

57 58

General Fertility Rate of

Paternal Fertility Rate
Men (GFRm)
� Let B = Number of births � General Fertility Rate of Men—Number of
M15-54 = Number of men of age 15– 54 births per 1,000 men age 15 to 54
B
= ∗ 1000
M15 - 54

59 60
 Summary

� Fertility data are collected from vital

statistics, censuses, or surveys
� Vital statistics principally provide birth
statistics Section B
� Many indicators have been developed to
understand and explain the fertility I ndicators of Reproduction Based on
patterns in populations based on vital Vital Statistics
statistics

61

Gross Reproduction Rate (GRR) Gross Reproduction Rate (GRR)

� Let Bf = Number of female births � Gross Reproduction Rate—Number of

Bm+f = Number of male and female daughters expected to be born alive to a
births, i.e., all births hypothetical cohort of women (usually
1,000) if no one dies during childbearing
years and if the same schedule of age-
specific rates is applied throughout the
childbearing years

Continued 63 Continued 64

Gross Reproduction Rate (GRR) Exercise

Gross Reproduction Rate (GRR)

B f (a) � Use the following data to calculate the GRR

GRR = ∑ ASFR (a) ∗ m + f United States, 1990
B (a) Age
Group of Births
Mother Women Total Males
GRR = TFR ∗ (Proportion of female births) 15-19 8 651 522 267
20-24 9 345 094 560
25-29 10 617 1277 653
� If the sex ratio at birth is assumed constant 30-34 10 986 886 454
across ages of women 35-39 10 061 318 163
40-44 8 924 49 25
Numbers are in 1,000s

You have 15 seconds to calculate the answer. You

may pause the presentation if you need more time.
65 Sources: U.S. Census 1990 and Vital Statistics of the U.S. Vol. I 66
 Exercise Answer Net Reproduction Rate (NRR)
Gross Reproduction Rate (GRR)

� The correct answer for the gross � Let Lfa = Life table person-years lived
reproduction rate is as follows: by women in age group “a”
– 1.01 daughters per woman l0 = Radix of life table
United States, 1990
Age
Bf = Number of female births
Group of Births
Mother Women Total Males
Bm+f = Number of male and female
15-19 8 651 522 267 births
20-24 9 345 094 560
25-29 10 617 1277 653
30-34 10 986 886 454
35-39 10 061 318 163
40-44 8 924 49 25
Numbers are in 1,000s
67 Continued 68

Net Reproduction Rate (NRR) Net Reproduction Rate (NRR)

� Net Reproduction Rate—Average number � For single-year age groups

of daughters expected to be born alive to a
hypothetical cohort of women if the same Lf Bf
NRR = ∑ ASFR ∗ 1 f a ∗ m+ f
schedule of age-specific fertility and l0 B
mortality rates applied throughout the
childbearing years � For five-year age groups
f
L
5 a Bf
f
NRR = ∑ ASFR ∗ ∗ m+ f
l0 B

Continued 69 70

Exercise Exercise Answer

Net Reproduction Rate (NRR) Net Reproduction Rate (NRR)

� Use the following data to calculate the NRR � The correct answer for the NRR is as follows:
United States, 1990 – 1.00 daughter per woman
Age
United States, 1990
Group of Births
Age Stationary
Mother Women Total Males
Group of Births Population
15-19 8 651 522 267
20-24 9 345 094 560
Mother Women Total Males 5Lx

25-29 10 617 1277 653 15-19 8 651 522 267 493 629
30-34 10 986 886 454 20-24 9 345 094 560 492 399
35-39 10 061 318 163 25-29 10 617 1277 653 490 989
40-44 8 924 49 25 30-34 10 986 886 454 489 203
Numbers are in 1,000s
35-39 10 061 318 163 486 812
40-44 8 924 49 25 483 465
Numbers are in 1,000s
Sources: U.S. Census 1990 and Vital Statistics of the U.S. Vol. I 71 72
 Live Births and Fertility Rates:
Net Reproduction Rate (NRR) United States, 1920– 1988

_
NRR ≈ l ( a ) ∗ GRR
� Where l( a) = Life table probability of
surviving beyond a
a = mean age of childbearing
usually 20< a < 30

Source: U.S. Department of Health and Human Services, Monthly Vital

73 Statistics Report, Vol. 39, No. 4, Supplement, August 15, 1990 74

Completed Fertility Rates for

Reproductivity
Birth Cohorts of Women
� Reproductivity is usually studied in terms of
mothers and daughters because of the
following reasons:
– The fecund period for females is shorter
than it is for males
– Characteristics such as age are much
more likely to be known for the mothers
of illegitimate babies than for their
fathers
Source: U.S. Bureau of the Census, Current Population Reports, Series
P-25, No. 381, tables A-1 and A-2 75 76

Summary

� Reproductivity data are collected from vital

statistics
� Many indicators have been developed to
understand and explain reproductivity Section C
patterns in populations
I ndicators of Fertility Based on
Censuses and Surveys

77
 Child-Woman Ratio (CWR) Child-Woman Ratio (CWR)

� Child-Woman Ratio—Number of children P0 − 4

CWR = ∗ 1000
zero to four years old relative to number of W15 − 44
women of reproductive ages
� Where P 0-4 = Mid-year population of
persons age 0-4 years
W 15-44 = Number of women of
reproductive ages

Continued 79 80

Exercise Exercise Answer

Child-Woman Ratio (CWR) Child-Woman Ratio (CWR)

� Use the following data to calculate the � The correct answer is as follows:
CWR – 687.4 children 0-4 per 1,000
Household population composition women 15-44
by age and sex, Kenya 1998
Household population composition
Age group Male Female Total
by age and sex, Kenya 1998
0-4 268873 256705 525578
Age group Male Female Total
15-19 189272 192067 381340 0-4 268873 256705 525578
20-24 130899 158825 289723
25-29 116747 142204 258951 15-19 189272 192067 381340
30-34 100827 99727 200555 20-24 130899 158825 289723
35-39 88445 103421 191866 25-29 116747 142204 258951
40-44 67218 68332 135550 30-34 100827 99727 200555
35-39 88445 103421 191866
Source: Demographic and Health Survey, Kenya 1998 81 40-44 67218 68332 135550 82

Parity Progression Ratios

Children Ever Born (CEB)
(PPR[i])
� Children Ever Born—Total number of � Let N = Random variable for number
children a woman has ever given birth to of births
� The survival status of the children is not mi = Proportion of women of parity “i”
considered here Mi = Proportion of women at or
� Almost all censuses tabulate mean CEB by above parity “i”
marital status and by age of the mother

83 Continued 84
 Parity Progression Ratios Exercise
(PPR[i]) Parity Progression Ratios (PPR)

� Parity Progression Ratios—Probability that � Given the following percentages of women at

a woman has an (i+1st) birth given that parity “i” for women 45-49 in 1995 in Colombia,
she already has had “i” births calculate PPR(2)

Parity Percentage
= ai 0 9.2
Mi + 1 1 7.5
= 2 13.9
Mi 3 18.6
4 15.6
= Prob (N ≥ i + 1 N ≥ i) 5+ 35.2
85 Source: Demographic and Health Survey, Colombia 1995 86

Exercise Answer Parity Progression Ratios (PPR)

Parity Progression Ratios (PPR)

� The correct answer for the PPR(2) is as � Note: a0 = Prob (ever give birth)
follows: 1-a0 = Prob (never have a birth)
– 0.83 ( i.e., there is an 83% chance that a M0 = 1
woman 45– 49 has a second birth given she � ai need not decrease with increasing I
already has had a first birth)
� Mi does decrease
Parity Percentage
0 9.2
1 7.5
2 13.9
3 18.6
4 15.6
5+ 35.2 87 88

Proportion of Women of Parity Proportion of Women of Parity

i(mi) i(mi)
� I s an unconditional probability � Given the following percentages of women at
parity “i” for women 45– 49 in 1995 in Colombia,
= Mi - Mi+1 verify the relationships indicated on the previous
slide
= Prob (N ≥ i) - Prob (N ≥ i + 1)
Parity Percentage
= a0 ∗ a1 ∗ ... ∗ ai-1 ∗ (1 − ai ) 0 9.2
1 7.5
2 13.9
3 18.6
4 15.6
5+ 35.2

Continued 89 Source: Demographic and Health Survey, Colombia 1995 90

 Mean Parity Exercise
Mean Parity

� Given the following percentages of women at parity

“i” for women 45– 49 in 1995 in Colombia, calculate
the mean parity using both formulas on the previous
max(i) max(i)
slide (assume that women at parity 10+ are on
= = average at parity 11)
∑M i ∑ i∗ m i
i=1 i=1 Parity Percentage Parity Percentage
0 9.2 6 8.0
1 7.5 7 5.9
2 13.9 8 3.2
3 18.6 9 3.2
4 15.6 10+ 4.8
5 10.1
91 Source: Demographic and Health Survey, Colombia 1995 92

Exercise Answer Estimation of GFR from Census

Mean Parity Data on Ratio of Children to Women
� The correct answer is as follows: � Let P 0-4 = Enumerated number of
– Mean parity = 4.01 children under 5
W 15-44 = Enumerated number of
women of age 15– 44
Parity Percentage Parity Percentage
W 20-49 = Enumerated number of
0 9.2 6 8.0 women of age 20– 49
1 7.5 7 5.9
2 13.9 8 3.2
3 18.6 9 3.2
4 15.6 10+ 4.8
5 10.1

93 Continued 94

Estimation of GFR from Census Estimation of GFR from Census

Data on Ratio of Children to Women Data on Ratio of Children to Women
� Also let l0 = Radix of life table
l0
nLa = Life table person-years P0- 4 ∗
lived between the ages “a” L
5 0
GFR est =
and “a+n” (female life table) W15− 44 + W20 − 49 L
30 15
∗
2 30 L 20 + 30 L15
2

Continued 95 Continued 96
 Estimation of GFR from Census Estimation of GFR from Census
Data on Ratio of Children to Women Data on Ratio of Children to Women
� Note: � The life table survivorship functions are
– Assumes that a life table is available inverted in order to estimate the number of
– Estimated GFR may be used in the persons at the mid-point of the preceding
absence of birth statistics to compare five-year period
fertility levels in various areas � A lexis diagram makes it easier to
understand this calculation

Continued 97 98

Summary Summary

� Fertility data are collected from vital � Censuses provide the following:
statistics, censuses, or surveys – Data on births and fertility
– Statistics on children by family status of
the parents
– Population data on fertility-related
variables
– Population bases for calculating various
types of fertility measures

Continued 99 Continued 100

Summary Summary

� Surveys provide the following: � Many indicators have been developed to

– Same type of data as censuses understand and explain the fertility
– Additional detailed data on special patterns in populations
aspects of fertility, including number
and timing of births, marriages,
pregnancies, birth intervals, and birth
interval components

Continued 101 102

 Relations among I ndicators

� At age 50:
max(i)
∑ COSFR(i,50 ) = Completed cohort
Section D i =1 fertility rate
Relationship among I ndicators, = Mean CEB
I ndicators and Models of Birth max(i)
I ntervals, and Fertility Models = ∑ Mi
i =1
Continued 104

Pregnancy Histories and

Relations among I ndicators
Birth Histories
max(i) max (i ) Bi � Data needed
∑ OSFR ( i ) = ∑
i=1 i = 1 W 15 − 44 – Age or birth date of woman
– Dates of pregnancy terminations
B – Type of termination (live birth or not)
=
W 15 − 44

= GFR
105 Continued 106

Pregnancy Histories and Pregnancy Histories and

Birth Histories Birth Histories
� Two ways of collecting data � Problems
– Forward, i.e., from first to last birth – Dating of events
– Backward, i.e., from last to first birth – Forgetting events
– Age misreporting

Continued 107 Continued 108

Cohort fertility (births per 1000 Age 35
women) calculated from survey Birth I nterval Measures
data for forward and backward 253 and Models
questionnaires by five-year age 277 30
groups and five-year periods � Let NS = Time from previous pregnancy
before the survey 332 317 outcome to resumption of
Bangladesh, 1984 313 311 25 ovulation (postpartum non-
susceptibility subinterval)
Legend: 274 269 241 � For first order births use dates of marriage
277 278 258 20 or beginning of sexual relations
xx: Forward
xx: Backward 84 110 58 112
77 89 88 100 15
20 15 10 5 0
Years before the survey 109 Continued 110

Birth I nterval Measures Pregnancy and Live

and Models Birth I ntervals
� Also let C = Time from resumption of � Pregnancy interval
ovulation to next conception – PI = I nterval between two pregnancies
(conception-wait subinterval) = NS + C + G
G = Time from conception to next � Live birth interval
pregnancy outcome
(gestational subinterval) – LBI = I nterval between two live births
W = Waiting time due to non-live = PI + W
births

111 112

Observed Birth I ntervals Birth I nterval

� Closed intervals � Note:

� Left censored intervals – Mean birth interval for women is
� Right censored intervals different from mean birth interval for
� Life table methods can be used to births in a given time period
incorporate the different intervals and – The latter is shorter
calculate median birth interval lengths

113 114
 Birth I nterval Mean Birth I nterval
Mathematical relationships

� I n populations with nothing changing 1

Let p = Fecundability (assumed fixed) Mean birth interval = +s
( p (1 - ue) )
e = Effectiveness of contraception
u = Proportion using contraception
1
s = Non-susceptible period (constant) = Fertility rate for
Mean B irth Interval
fecund women

115 116

Mean Conception-Wait
Heterogeneous Fecundability
Subinterval (MC)
� Mean Conception-Wait Subinterval—Mean time it � Probability of conception in month “k”
takes to get pregnant under natural fertility
� “p” is assumed to vary between women
� Constant fecundability
with distribution “f(p)”
2
MC = 1 ∗ p + 2(1 - p)p + 3(1 - p) p + ...
1
= = ∫01 p qk -1 f(p) dp
p
p = Monthly probability of conception – Where q = 1-p
(assumed fixed in time and for all
women)
117 118

Probability of Conceiving Fertility Models

I n Month “k” Given No Conception Before Then Coale and Trussell

1 k � Let r(a) = Observed fertility schedule at

∫0 q f(1 - q) dq age “a”
= 1- n(a) = Natural fertility pattern at age “a”
1 k -1 f(1 - q) dq
∫0 q
� Note M = Level of fertility (estimated)
– Here the probability of conception m = Degree to which fertility control
decreases over time is practiced (estimated)
– This has important implications for the v(a) = Estimated from U.N.
study of infertility Demographic Yearbook 1965;
data for 43 fertility schedules
119 Continued 120
 Fertility Models Fertility Models
Coale and Trussell Bongaarts

� Bongaarts model of intermediate fertility variables

Let Cm = I ndex of marriage
Cc = I ndex of contraception
r (a )
= M ∗ e (m ∗ v(a)) Ca = I ndex of induced abortion
n( a ) Ci = I ndex of post-partum infecundability
K = Estimated total natural fertility
rate (15.3 [Bongaarts 1982])
� Each “C” varies between 0.0 and 1.0

121 Continued 122

Fertility Models Summary

Bongaarts

� Methods of estimation of each “C” are � I nformation on women's fertility can be

provided by Bongaarts (1982) obtained by asking them to report their
pregnancy and birth histories
TFR = C m ∗ C c ∗ C a ∗ C i ∗ K � I ndicators and models have been
developed to understand and explain the
fertility patterns in populations

123 124

Measurement of Migration Section A

Stan Becker, PhD Definition of Migration, Types of

Bloomberg School of Public Health Migration, and Estimation of Net
Migration
 Definitions Definitions

� Mover—A person who changes residence � Migration— Geographic or spatial mobility

� Migrant— A person who moves from one involving a relatively permanent change in
political area to another usual residence between clearly defined
� Non-migrant— Non-movers and local political or statistical units; has dimensions
movers of time and space

Continued 4 Continued 5

Definitions Definitions
� I n-migrant—A person who moves in a � Net Migration—I n-migrants - Out-migrants
political area within the same country � Net I mmigration—I mmigrants - Emigrants
� I mmigrant—An international migrant who
enters the area from a place outside the – Note: Net migration for an area often
country includes both international and internal
� Out-migrant—A person who moves out of a migration
political area within the same country � Gross Migration— I n-migrants + Out-
� Emigrant—An international migrant migrants = Migration turnover
departing to another country by crossing
the international boundary
Continued 6 Continued 7

Definitions Estimating Net Migration

Residual Method

� Migration stream—A group of migrants � Let I = Number of in-migrants

having a common origin and destination in O = Number of out-migrants
a given migration period P0 = Population at time “0”
� Migration counterstream—I n opposite Pt = Population at time “t”
direction of stream
B = Number of births
D = Number of deaths

8 Continued 9
 Estimating Net Migration Estimating Net Migration
Residual Method Residual Method

(I − O) = (Pt − P0 ) − (B − D) � Because the census counts and vital

� Estimates net migration as difference in statistics are subject to unknown and
usually differing degrees of error, the
population counts at two time points and residual estimate of net migration may be
net vital events in-between, i.e., subtracts in substantial error
an estimate of natural increase during the � The relative error in net (in-) migration
period from the net change in population may be considerable when the amount of
during the period migration is small
� Also known as the bookkeeping method or � The residual method can be used to
balancing equation estimate net migration for sex, race,
Continued 10
etc . . . 11

Estimating Net Migration Estimating Net Migration

Cohort-Component Method Cohort-Component Method

� I nvolves the calculation of estimates by � Survival rates are used instead of death
age groups on the basis of separate rates because the compilation of death
allowances for the components of statistics is often laborious, even when
population change basic statistics on death are available
� Let P 0a = Population in age group “a” at
� Estimates net migrants as the difference time “0”
between actual population at time “t” and t
P a+t = Population in age (group) “a+t”
the population at time “0” survived to time at time “t”
“t” (must do age-specific and then add) L
s = Life table survival ratio = a+t
La
Continued 12 13

Three Ways of Estimating Three Ways of Estimating

Net Migration Net Migration
� Forward estimate Age � Reverse estimate Age

Pta+t Pat + t Pta+t

M1 = Pat + t − sPa0 M2 = − Pa0
a+t s a+t

P0 a P0 a
N ote M 2 > M1
a a

0 t time 0 t time
Continued 14 Continued 15
 Three Ways of Estimating
Forward Estimation
Net Migration
� Average estimate � All migrants come at the end of the time
M1 + M 2 interval (or, none of the migrants die for
M3 = the period between time “0” and “t”)
2

16 17

Reverse Estimation Average Estimation

� All migrants come at the beginning of the � All migrants come at the middle of the time
time interval (or, all migrants are subjected interval (or, all migrants are subjected to
to population’s mortality for the entire population’s mortality for the half of the
period between time “0” and “t”) period between time “0” and “t”)

18 19

I mplicit Assumption Regarding

Calculating the Estimates
Time of I mmigration
Forward Reverse Average � When calculating the estimates, it is
Age Age Age important to remember that the survival
a+t Pta+t a+t Pta+t a+t Pta+t rates for children born during the period
are of a different form from those for the
older ages, i.e., for the case of five years
a a a
between population counts:
P0a P0a P0a 5L0 L
S0−4 = S5 −9 = 5 5
0 t time
5l 0 5 L0
0 t time 0 t time

20 21
 Cohort-Component Method Cohort-Component Method

� Does not require accurate vital statistics � Unlike the residual method, none of the
but does require a life table variants of the cohort-component method
� This method yields estimates of net measure net migration exactly even when
migration by age and sex without nearly as there are no errors in the underlying
much labor as is involved in the use of population and vital statistics
deaths by age � The average estimate implies a more
meaningful assumption regarding the
timing of net migration than does either
the forward or reverse method
Continued 22 23

Estimating Net Migration Summary

� The amount of difference between the � Migration is an important element in the

migration estimates from the forward and growth of the population and the labor
reverse methods depends on the amount force of an area
of net migration and on the level of the � I nternational and internal are the two
survival rate broad types of migration
� Many indicators have been developed to
measure migration under its multiple facets

24 25

Crude I n-Migration Rate

� Crude I n-Migration Rate—Number of in-

migrants per 1,000 population

Section B I
= ∗ 1000
P
I ndicators and Ways of
Studying Migration

27
 Crude Out-Migration Rate Crude Net Migration Rate

� Crude Out-Migration Rate—Number of out- � Crude Net Migration Rate—Difference

migrants per 1,000 population between the number of in-migrants and
the number of out-migrants per 1,000
O population
= ∗ 1000
P I-O
= ∗ 1000
P

28 29

Exercise Exercise Answer

Crude Net Migration Rate Crude Net Migration Rate

� Calculate the crude net migration rate for � The correct answer is as follows:
Zimbabwe in 1987, based on the following – Crude net migration rate = -0.16
data
Zimbabwe, 1987
Zimbabwe, 1987
Long-term immigrants: 3 925
Long-term emigrants : 5 330 Long-term immigrants: 3 925
Total population : 8 640 000 Long-term emigrants : 5 330
Total population : 8 640 000
You have 15 seconds to calculate the answer. You
may pause the presentation if you need more time.
Source: U.N. Demographic Yearbook, 1989 30 31

Specific Rates Age-Specific Migration Rates

� Let Ma = Number of in-migrants (or out- � Age-Specific Migration Rates—Number of

migrants) in age group “a” migrants of age group “a” per 1,000
P a = Midyear population in age group population of age group “a”
Ms = Number of in-migrants (or out-
migrants) of sex “s” Ma
P s = Midyear population of sex “s”
= ∗ 1000
Pa

32 33
 Age-Sex-Specific
Sex-Specific Migration Rates
Migration Rates
� Sex-Specific Migration Rates—Number of � Age-Sex-Specific Migration Rates—Number
migrants of sex “s” per 1,000 population of of migrants of age “a” and sex “s” per
sex “s” 1,000 population of age “a” and sex “s”

Ms M as
= ∗ 1000 = ∗ 1000
Ps Pas

34 35

Exercise Exercise Answer

Age-Sex-Specific Migration Rates Age-Sex-Specific Migration Rates

� Calculate the in- and out-migration rate for � The correct answers are as follows:
men age 20– 24 from Zimbabwe in 1987, – 0.77 in-migrants age 20– 24 per
based on the following data 1,000 men 20– 24
Zimbabwe, 1987
– 1.70 out-migrants age 20– 24 per
Men age 20-24
1,000 men 20– 24
Long-term immigrants: 302 Zimbabwe, 1987
Long-term emigrants : 666 Men age 20-24
Male population 20-24: 391 968
Long-term immigrants: 302
You have 15 seconds to calculate the answer. You
Long-term emigrants : 666
may pause the presentation if you need more time.
Male population 20-24: 391 968
Source: U.N. Demographic Yearbook, 1989 36 37

Ratios Ratios
� Various types of ratios can be computed to � Ratio of out-migration to in-migration =
indicate the relative magnitude of in- magnitude of out-migration compared to
migration, out-migration, net migration, in-migration
and gross migration to or from a country: O
I
� Let I = Number of in-migrants
O = Number of out-migrants � Ratio of net migration to in-migration
I +O = Gross migration I−O
I -O = Net migration , where I > O
I
Continued 38 39
 Exercise Exercise Answer
Ratios Ratios

� Calculate the ratio of out- to in-migration � The correct answers are as follows:
and the ratio of net to in-migration for – Ratio of out- to in-migration = 1.36
Zimbabwe in 1987 – Net to in-migration is not
Zimbabwe, 1987 calculated because I <O
Long-term immigrants: 3 925
Long-term emigrants : 5 330 Zimbabwe, 1987
Total population : 8 640 000
Long-term immigrants: 3 925
You have 15 seconds to calculate the answer. You Long-term emigrants : 5 330
may pause the presentation if you need more time. Total population : 8 640 000
Source: U.N. Demographic Yearbook, 1989 40 41

Ratios Exercise
Ratios

� Ratio of net (out-) migration to out- � Calculate the ratio of net to out-migration
migration and the ratio of in- to gross migration for
O −I Zimbabwe in 1987
, where O > I
O Zimbabwe, 1987

� Ratio of in-migration to gross migration = Long-term immigrants: 3 925

magnitude of in-migration to the overall Long-term emigrants : 5 330
Total population : 8 640 000
migration movement
I You have 15 seconds to calculate the answer. You
may pause the presentation if you need more time.
I+ O
42 Source: U.N. Demographic Yearbook, 1989 43

Exercise Answer Ratios

Ratios

� The correct answers are as follows: � Ratio of out-migration to gross migration =

– Net to out-migration = 0.26 magnitude of out-migration to the overall
– I n- to gross migration = 0.42 migration movement
O
Zimbabwe, 1987 I+ O
Long-term immigrants: 3 925
Long-term emigrants : 5 330
Total population : 8 640 000

44 Continued 45
 Ratios Exercise
Ratios

� Ratio of net migration to gross migration = � Calculate the ratio of out- to gross
migration effectiveness (magnitude of the migration and the ratio of net to gross
effective addition [or loss] through migration for Zimbabwe in 1987
migration to the overall gross movement) Zimbabwe, 1987
I− O
Long-term immigrants: 3 925
I+ O Long-term emigrants : 5 330
Total population : 8 640 000
You have 15 seconds to calculate the answer. You
may pause the presentation if you need more time.
46 Source: U.N. Demographic Yearbook, 1989 47

Exercise Answer Partial Migration Rate

Ratios

� The correct answers are as follows: � Let I ij = Number of in-migrants to area “i”
– Out- to gross migration = 0.58 from area “j”
Oij = Number of out-migrants from
– Net to gross migration = -0.15 area “i” to area “j”
Zimbabwe, 1987 I ij = Oji (by definition)
Long-term immigrants: 3 925
Long-term emigrants : 5 330 � Also let
Total population : 8 640 000 P i = Midyear population in area “i”
Pj = Midyear population in area “j”
48 Continued 49

Gross Rate of Population

Partial Migration Rate
I nterchange
� Partial Migration Rate—Number of migrants � Gross Rate of Population I nterchange—
to an area from a particular origin, or from Proportion of in-migration (or out-
an area to a particular destination, per migration) between two populations
1,000 of the population at either origin or
destination Iij + I ji Oij + O ji
= or
Iij Iij Pi + Pj Pi + Pj
∗ 1000 or ∗ 1000
Pi Pj

Oij Oij
or ∗ 1000 or ∗ 1000
Pi Pj 50 51
 Distribution Distribution
From Census Data From Administrative Data

� Migration can be studied by considering the � Migration can be tabulated by the

distribution of the following: following:
– Place of birth – Nation of origin
– Place of residence at some specified – Destination
earlier date – Age
– Nationality – Sex
– Duration at current residence

52 53

Measures from Surveys with Measures from Surveys with

Migration Histories Migration Histories
� Number of moves � Note:
– Mean/median – The reference population is selected for
– Moves per unit of time current presence in area and survival
� Probability of ever moving – There may be problems of recall error
– Can be calculated by age, sex, race,
etc.

Continued 54 55
 Summary

� Migration is an important element in the

growth of the population and the labor
force of an area Population Change and
� I nternational and internal migration are the Projection
two broad types of migration
� The measurement and analysis of Stan Becker, PhD
migration are important in the preparation Bloomberg School of Public Health
of population estimates and projections

58

Population Change
� Book-keeping equation
� Let Pt = Population at time “t”
P0 = Population at an earlier time “0”
Section A B = Births between time “0” and time
“t”
Rate of Change, Doubling Time, and D = Deaths between time “0” and “t”
the Relationship between Age I = In-migration / immigration
Distinction and Demographic Rates between time “0” and “t”
O = Out-migration / emigration
4
between time “0” and “t” Continued

Population Change Linear Change

� The population changes by adding births � The population size changes by exactly the
and in-migrants and subtracting deaths and same amount “b” during each time period
out-migrants “t”
Pt = P0 + B − D + I − O Pt = P0 + bt
� Let b = Number added per time unit
Pt − P0
b=
t
5 6
 Linear Annual Rate of Change Linear Annual Rate of Change
Beginning of Period Approximation Arithmetic or Mid-Point Approximation

� ri = linear annual rate of change � rm = linear annual rate of change

(beginning of period approximation) (arithmetic or mid-point approximation)
P − P0
ri = t b Pt − P0
tP0 rm = =
1/ 2(Po + Pt ) t / 2(Po + Pt )
Pt = P0 + ri ∗ t ∗ P0

7 8

Exercise Exercise Answer

Linear Annual Growth Rate Linear Annual Growth Rate

� The Mexican population was estimated at � The correct answers are as follows:
77,938,288 at the 1985 census and at ri rm
91,158,290 at the 1995 census 1.70% 1.56%
� Calculate the linear annual growth rate
using both types of approximations

You have 15 seconds to calculate the answer. You

may pause the presentation if you need more time.
Source: U.N. Demographic Yearbook, 1992 and 1997 9 10

Problem with the Linear- Problem with the Linear-

Change Approach Change Approach
� Adding a fixed amount in each time period � So
⇒ the rate of change decreases because
P −P P1 − P0
b / Pt -> 0 r1 = 1 0 = 1 rm1 =
1∗ P0 1 2(P0 + P1)
� Say b = P0
P0 2
� Then P1 = P0 + P0 = 2P0 = =
3 2 P0 3
P2 = P1 + P0= 3P0

Continued 11 Continued 12
 Problem with the Linear-
Geometric Change
Change Approach
� And each continues declining � The population changes by step, i.e., the
increment added (or the decrement is
P2 − P1 P2 − P1
r2 = rm2 = subtracted) periodically
1∗ P1 1 2(P1 + P2 )
=
3P0 − 2P0 1
= P0 2
= = Pt = P0 (1 + rg )t
2P0 2 5 2 P0 5
� Where rg = geometric rate of growth

13 14

Geometric Annual Rate Exercise

of Change Geometric Annual Rate of Change

� The Mexican population was estimated at

1 ⎛ Pt ⎞
ln⎜ ⎟
77,938,288 at the 1985 census and at
t ⎜⎝ P0 ⎟⎠ 91,158,290 at the 1995 census
t Pt
� Calculate the geometric annual growth rate
rg = −1 = e − 1
P0

You have 15 seconds to calculate the answer. You

may pause the presentation if you need more time.
15 Source: U.N. Demographic Yearbook, 1992 and 1997 16

Exercise Answer Exponential Change

Geometric Annual Rate of Change

� The correct answer for rg is as follows: � The population changes instantaneously

– 1.58% and continuously

Pt = P0 e rt

17 18
 Exponential Growth Rate Exercise
Exponential Growth Rate

� Exponential growth rate � The Mexican population was estimated at

77,938,288 at the 1985 census and at
1 ⎛ Pt ⎞
r= ln ⎜ ⎟ 91,158,290 at the 1995 census
t ⎜⎝ P0 ⎟⎠ � Calculate the exponential growth rate
� The advantage of the exponential method is that
it’s easy to solve for any parameter

⎛P ⎞
ln ⎜⎜ t ⎟⎟ P
P P0 = rtt You have 15 seconds to calculate the answer. You
t= ⎝ 0⎠ e may pause the presentation if you need more time.
r
19 Source: U.N. Demographic Yearbook, 1992 and 1997 20

Exercise Answer Relation Between Rates of Change

Exponential Growth Rate Geometric and Exponential

� The correct answer for “r” is as follows:

rg = e r − 1
– 1.57%
x x2 x3
since e = 1 + x + + + ...
2! 3!
r2
then rg = r + + ...
2!
⇒ rg > r
21 22

Rate of Natural I ncrease Rate of Natural I ncrease

� Let B = Births during a calendar year � Most direct indication of how rapidly a
D = Deaths during same year population actually grew as the result of
vital processes
P = Midyear population
b = Birth rate B −D r < 0 ⇒ D>B
r= ∗ 1000
d = Death rate P r = 0 ⇒ B =D
=b-d r >0 ⇒ B >D

� Is influenced by age structure

Continued 23 24
 Doubling Time Doubling Time

� Calculation based on exponential growth � Since ⎛P ⎞

(“r” fixed in time) ln ⎜⎜ t ⎟⎟
P
� Allows determination of how long it will t= ⎝ 0⎠
take a population to double in size (i.e., r
time for Pt / P0 = 2)
� Then for doubling time

ln( 2)
t=
r
Continued 25 Continued 26

Some Examples of
Doubling Time
Doubling Times
� Approximation Rate of Number of Rate of Number of
70 Growth Years Growth Years
t= 1.0 69.3 2.0 34.7
r ∗ 100 1.1 63.0 2.1 33.0
1.2 57.8 2.5 27.7
1.3 53.3 3.0 23.1
1.4 49.5 3.5 19.8
1.5 46.2 4.0 17.3
27 28

Number of Years Necessary for the Population to Exercise

Double at Given Fixed Annual Rates of Growth Doubling and Tripling Time

� The Mexican population was estimated at

140
77,938,288 at the 1985 census and at
91,158,290 at the 1995 census
90
� Assuming Mexico's population is growing
40 exponentially, what is its doubling time?
� In what year will the mid-1975 population
Number of Years

-10 have tripled?

0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0 You have 15 seconds to calculate the answer. You
Rate of Growth may pause the presentation if you need more time.
29 Source: U.N. Demographic Yearbook, 1992 and 1997 30
 Exercise Answer The Relationship between Age
Doubling and Tripling Time Distribution and Demographic Rates
� The correct answers are as follows: � Age distribution of a population gives a
– Doubling time = 44.2 years record of the demographic history of that
– Tripling time = 70.1 years population
– Year will triple = 2045

31 Continued 32

The Relationship between Age

Summary
Distribution and Demographic Rates
� Age distribution is sensitive to changes in � Population change is measured by the
demographic rates, particularly fertility and, difference between population sizes at
for small areas, migration rates different dates
� In a closed population (i.e., no migration) � As time goes by, the population in an area
can do the following:
the age distribution is determined only by
fertility and mortality – Grow by adding births or people moving
in the area (in-migrants)
– Or it can decrease because of deaths or
people moving out of the area (out-
migrants)
33 Continued 34

Summary Summary

� The simplest method to estimate the � Mathematical models have been developed
population at a later time from that of an to estimate the average annual rate of
earlier date is the bookkeeping method, change based on different assumptions on
which adds the births and in-migrants to the way the population has grown during
the initial population and subtracts the the period
deaths and out-migrants

Continued 35 36
 Population Projection

� Population Projection—Forecast of
population change using estimates of
fertility, mortality, and migration
Section B � Projections may extend for varying
numbers of years into the future
Projection � Note:
– Extrapolation = projection
– Interpolation = estimation

Continued 38

Population Projection Component Method

� Long-term projections (over 25 years) are � Has become the standard methodology for
used in connection with the development projection
of natural resources, planning for provision � Makes explicit the assumptions regarding
of food, for transportation and recreational the components of population growth—
facilities, etc. mortality, fertility, and net migration
� Middle-range projections (10–25 years) are � Gives insight into the way population
used for planning educational and medical changes
facilities and services, housing needs,
etc.
39 Continued 40

Component Method General Principles

� Allows the user to estimate the effect of � Start with the population distributed by age
alternative levels of fertility, mortality, or and sex at base date
migration on population growth � Apply assumed survival rates and age-sex
� Used to obtain projections of age-sex specific fertility rates to obtain number of
structure persons alive at the end of a unit of time

41 Continued 42
 General Principles Population Projection

� Make allowance for net migration by age � Projection assuming closed population, constant fertility,
and constant mortality
and sex, if desired
– Let x = Age 5, 10, , w
� Generally five-year age groups are used
t = Time “t”
� Projection interval must be integer multiple
5Lx = Life table number of persons
of age interval between ages “x” and “x+5” at time
“t”
5L x = Survival ratio from age group
L
5 x −5
(x-5, x) to age group (x, x+5)

43 Continued 44

Treatment of Open-Ended
Population Projection
I nterval

t+5 t 5
Lx T
P = P ∗ 5 Lw
x , x +5 x −5 , x P t +5 = P t ∗ + P t ∗ (w+5)
5 L x −5 w+ w −5 w+
5 L w −5 Tw
� Projection of the population aged 55 to 59
from 1990 to 1995 � Where w+ refers to the oldest age group

1990 5L 60
P 1995 = P ∗
60 − 64 55 − 59
5L 55

45 Continued 46

Treatment of Open-Ended Treatment of Open-Ended

I nterval I nterval
� Keyfitz’s method – Substituting

t +5 t 5 L 85 t T90 t +5 t
⎛ L T 85 T ⎞
P85 + =5 P ∗ + P85 + ∗ P 85 + =5P ∗ ⎜⎜ 5 85 + ∗ 90 ⎟⎟
80 80
5 L 80 T85 ⎝ 5 L 80 5 L 80 T 85 ⎠
– In stationary population ⎛ L + T90 ⎞
= 5 P80 ∗ ⎜⎜ 5 85 ⎟⎟
⎝ 5 L 80 ⎠
5P80 L 5 P80 ∗ T 85
= 5 80 ⇒ P85 + = T 85
P85 + T 85 5 L 80
= 5 P 80 ∗
5 L 80
Continued 47 48
 Getting P0-4 at Time t+5 Getting P0-4 at Time t+5

� Let 5Lx = Number of survivors between � Also let α = Beginning age of reproduction
ages “x” and “x+5” β = End of reproductive age
f
5F x = Female fertility rate for age 5Fx = Female fertility rate for age
group (x, x+5) group x, x+5
x = Age “x”

Continued 49 Continued 50

Getting P0-4 at Time t+5 -

Getting P0-4 at Time t+5
Derivation
t +5 = Survivors to time “t+5” of births � Births in one year =
5 P0 that occurred between “t” and ⎛ ⎞
5 L 15
“t+5” (1/2 ) ∗ ⎜⎜ 5 P15 + ∗ 5 P10 ⎟⎟ ∗ 5 F 15 +
⎝ 5 L 10 ⎠
β −5
5
L0 ⎡ ⎛ L ⎞ ⎤
= ∗ ⎢ ∑ ⎜⎜ 5 Fxf + 5 x + 5 ∗5 Fxf + 5 ⎟⎟ ∗5 Px ⎥ ⎛ 5 L 40 ⎞
... + (1/2 ) ∗ ⎜⎜ 5 P 40 + ∗ 5 P35 ⎟⎟ ∗ 5 F 40
2l 0 ⎣ x = α −5 ⎝ 5L x ⎠ ⎦ L
⎝ 5 35 ⎠
β −5
⎛ 5 L x +5 ⎞
= (1/2 )∗ ∑ ⎜⎜ 5 F x + ∗ 5 F x + 5 ⎟⎟ ∗ 5 P x
x = α −5 ⎝ 5L x ⎠
51 Continued 52

Getting P0-4 at Time t+5 -

Population Projection
Derivation
� Births in five years = � To get P0-4, births must survive an average
⎡ β −5 ⎤ of 2.5 years
⎛ L ⎞
5 ∗ ⎢ (1/2 ) ∗ ∑ ⎜⎜ 5 Fx + 5 x + 5 ∗ 5 Fx + 5 ⎟⎟ ∗ 5 Px ⎥ t+5 5 5 L 0 ⎡ β −5 ⎛ L ⎞ ⎤
⎣ x = α −5 ⎝ 5L x ⎠ ⎦ 5 P0 = ∗ ∗ ⎢ ∑ ⎜⎜ 5 Fx + 5 x + 5 ∗ 5 Fx + 5 ⎟⎟ ∗ 5 P x ⎥
2 5l 0 ⎣x =α−5 ⎝ 5L x ⎠ ⎦

5L0 ⎡ β −5 ⎛ 5 L x+5
⎞ ⎤
= ∗ ⎢ ∑ ⎜⎜ 5 F x + ∗ 5 F x + 5 ⎟⎟ ∗ 5 P x ⎥
2l 0 ⎣x =α−5 ⎝ 5 L x ⎠ ⎦
β − 5
L 5 0
⎡ ⎛ 5 L x + 5
⎞ ⎤
= ∗ ⎢ ∑ ⎜⎜ 5 Px + ∗5 Px + 5
⎟⎟ ∗ 5 F x ⎥
2l 0 ⎣ x = α − 5 ⎝ 5 L x ⎠ ⎦
53 54
 Matrix Representation Matrix Representation
⎡ ⎤
� 1st row represents fertility rates and 0–5 ⎢ ⎥
⎢0 0 L0
⎛ L15
⎞ L0
⎛ L20
⎞ ⎥
⎜⎜ ∗F15⎟⎟ ⎜⎜F15 + ∗F20⎟⎟ ... 0 0
survival ⎢ 2l0 ⎝L10 ⎠ 2l0 ⎝ L15 ⎠ ⎥
⎢ ⎥
L
� Sub-diagonal represents survival ratios ⎢5 0
⎢L0
0 0 ... 0 0⎥
⎥
from one age group to another ⎢ ⎥
L10
L = ⎢0 0 0 ... 0 0⎥
� Note: The rest of the matrix (L) has zeros ⎢ L5 ⎥
⎢ ⎥
⎢. . . . .⎥
⎢. . . . .⎥
⎢ ⎥
⎢. . . . .⎥
⎢ ⎥
⎢0 0 L T
0 0 ... w-5 w ⎥
⎣⎢ Lw-10 Tw-5 ⎥⎦
Continued 55 56

Population Projection Population Projection

� To get the population at a later date � To project the total population at a later
date
– First, use the female projection and
L ∗Pt = Pt+5 fertility rates for both boy and girl births
t+5 t + 10 t – Then, put the male births into male
L ∗P = P = L2 ∗ P projection
– Note: The male matrix only contains
etc. survival ratios

Continued 57 58

Projection Adjusting for Net Projection Adjusting for Changes

Migration (NM) in Fertility and Mortality
� Let Li = Projection matrix for ith interval
Pt +5 = L ∗ Pt + NM
L0 = Projection matrix for first interval
� Where NM = Net migration
� It is easiest to take net migration into
account at the end of each time interval
� Getting estimates of net migration by age
and sex, and projected numbers by age
and sex, may be difficult

59 60
 Population Projection Population Projection

� Ex: At time t+10 � Note: Stable population results from

unchanging rates of births and deaths
Pt+10 = L1 ∗ L0 ∗Pt
( ) � Since Ln “stabilizes” when “n” gets large
� Or more generally (i.e., for every element), its value in “Ln+1”
divided by its value in “Ln” converges to a
constant “λ”
Pt +5∗n = Ln−1 ∗Ln−2...∗ L0 ∗Pt
� Where Li = Projection matrix for ith time
interval
Continued 61 Continued 62

Population Projection Population Projection

� Therefore: � Several variations of projection exist for

changing rates
Ln+1 ∗ Pt = λ ∗ Ln ∗ Pt – Change fertility rates by x% and/or
mortality rates by y%, at each step
so λ ∗ P t = P t + 5 , and – Fix final levels of fertility and mortality
and interpolate between the two time
Pt +5 periods
λ = t = e5 r in continuous notation
P – Note: Net migration can also be easily
made to vary

Continued 63 64

Ratio-Correlation or Regression
Projection for Small Areas
Methods
� Internal migration (e.g., rural-urban � Define:
migration) may be important
� Population projection for political units with
relatively few boundary changes (e.g.,
states) and for statistical subdivisions with
boundaries changing (e.g., urban/rural) are � Write equation (can use different number
two different problems of Xs):

Yi = α + β1X i1 + β2 X i2 + ...
65 66
 Projection for Small Areas Projection for Small Areas

� For example: � Estimate α, β1, β2 statistically, and use

these to predict Yi for some future time t2
– Note: Yi is a percent and if all areas are
done, the percentages may not add to
100
� So, they must be adjusted to 100 and then
multiplied by the estimated total population
at t2 to get the projection for area “i” at
that time
Continued 67 68

Administrative Records Combined Methods

Pt = Ht x PPHt + GQt � Average of several methods

� Method used by the U.S. Census Bureau
� Pt = Total population at time “t”
Ht = Occupied housing units at time “t”
PPHt = Average number of persons per
household at time “t”
GQt = Population in group quarters at
time “t”
69 70

J udging Projections J udging Projections

� The evaluation of projections requires � The concept of “accuracy” becomes less

some standard by which to judge their meaningful where several series of
quality projections are offered as reasonable
� One may reasonably compare a projection possibilities, and particularly where none is
with the population actually recorded later offered as a “forecast”
using percent differences to indicate how � The United Nations uses high, medium,
far it deviates from the actual figure low, and constant (unchanging fertility and
mortality) projections

Continued 71 Continued 72
 J udging Projections Proportional Error

� Where possible, it may be more profitable � Hindsight, calculate the proportional error:
to compare the actual components of Pactual − Pproj
population change (i.e., births, deaths, and Proportional error =
net migration) with their projected figures Pactual
because it provides insight into the � Where Pactual = Actual population
reasonableness of the various assumptions (e.g., census count at
time “t”)
Pproj = Population projection
for time “t”
73 74

Population Projection Population Projection

� One can also assess projections by looking � This width of the range depends on the
at the width of the range from the highest regularity of the following:
and the lowest series in a set of principal – Past demographic trends, knowledge
projections regarding past trends
– Ability to measure them accurately
– Analyst’s judgment of the likely course
of future change

Continued 75 Continued 76

Error in Projected Population by Target

Population Projection Year: Means across Countries and
Forecasts
� As the range widens, the analyst is
indicating that he/she has less and less
confidence in the projections

Adapted from John Bongaarts and Rodolfo A. Bulatao, Editors, Beyond six billion, National Academy Press, 2000

77 78
 Proportional Error by Projection
Uses of Projections
Length: Percentile for All Countries
� Forecasting future population change
� Warning of population increase or decrease
� Sensitivity analyses to see how population
will change with changes in assumptions
about fertility, mortality, and/or migration

Source: John Bongaarts and Rodolfo A. Bulatao, Editors, Beyond six

billion, National Academy Press, 2000 79 80

Examples of Applications Computer Programs

� Projection of population by age and sex to � Examples:

estimate the age distribution of the – DEMPROJ (Futures Group International) to
population at a later date create projections for policy presentations or
planning exercises and to produce the inputs
� Projection of children of school-age and required by the other programs of the
apply school enrollment ratios to evaluate integrated package SPECTRUM
the needs in teachers and schools – IPSS (SENECIO Software Inc.) for
� Projection for the labor force interactively exploring population dynamics,
population projection, and the life table
program for Macintosh

81 Source: UN , Statistics Division—UNFPA—Population Information Network 82

Population Projection Summary

� Contact: � Population projection is a forecast of
Director
population change using estimates of
Population Division, United Nations
United Nations Plaza, Rm. DC2-1950
fertility, mortality, and migration
New York, NY 10017 USA � No projection is 100 percent accurate, i.e.,
there is always some degree of uncertainty
Fax: (212) 963-2147
Email: [email protected] � The United Nations uses low, medium,
high, and constant projections
Population Information Network (POPIN) web site:
http://www.undp.org/popin

83 Continued 84
 Summary

� Various computer programs have been

developed to help the demographer
simulate the effects on population size of
different rates of fertility and mortality

85