LECTURE 11

GRADUATION OF DATA ON AGE DISTRIBUTION

Introduction
Age data usually contain irregularities these errors must be detected and corrected
before such age data is used for any meaningful demographic purpose. Upon detection
of irregularities, most age distributions are usually corrected by methods of graduation
or smoothing which includes
 graphic methods
 moving average method
 curve-fitting formulae
A number of graduation methods use mathematical approaches to derive data for 5-
years age groups that are corrected primarily for net reporting errors. We shall look at
two methods:

CARRIER - FARRAG GRADUATION METHOD

This methods consists essentially of splitting the total numbers reported in two adjacent
five years age groups (quinary age groups) by first grouping the quinary age groups
into 10-years age groups (denary age groups). The graduated population is then obtain
by the manipulation of simple algebraic formulae ie
U2 = 1/2U2 + 1/16(U0 - U4)
U3 = 1/2U2 - 1/16(U0 - U4)
U4 = 1/2U4 + 1/16(U2 – U6)
U5 = 1/2U4 - 1/16(U2 – U6)
U6 = 1/2U6 + 1/16(U4 – U8)
U7 = 1/2U6 - 1/16(U4 – U8)
NEWTONS QUADRATIC METHOD
The graduation formula is given as
1 1
5 Px =
2 10
Px+
16
( 10 Px −10−10 Px +10 )

This formula is applied after re-grouping the data from quinary to denary
Limitations of the mathematical methods
1. They assume that the types and magnitude of errors are the same or similar from age
group to the next or between the two sex groups or even over a period of time.
2. Although these techniques tends to smoothen the data vertically (age ratios), they do
not guarantee horizontal smoothness (sex ratios) and diagonal smoothness (survival
ratios).
DEM0GRAPGHIC MODELS
Demographic models are generalizations about a demographic phenomenon or process,
and are used as tools of estimation. In this role, the use of models becomes particularly
important when available data are limited and defective, when relevant demographic
data are easily available, there is little justification for using such models, but they are
indispensable in checking and adjusting data, in filling gaps in the available records and
in deriving reliable estimate from fragmentary pieces of evidence, each of which may be
defective if taken in isolation.
The three types of models that are widely used and applied in demographic estimation
are -
 Stable population models
 Quasi stable population model
 Model life Tables e.g. UN models life tables, Reed -Merell model life table,
Regional (Coale and Demeny) model life table, Brass model life tables and
Carrier and Hobcraft system
Life tables are more accurate if constructed from mortality data of current population.
In usual construction of life tables data are age distributions, number of births, number
of deaths in each age group if a current population over a certain period. It is important
to note that the death rate in an age group (say 60-65) say for a current population
cannot be expected to be the same when newly borne individuals reach that age. We
therefore must make assumptions which although almost untrue helps us to appreciate
the nature of the information in the life table. These assumptions are the basis of two
concepts used when studying mortality: Stationary and Stable populations.
STATIONARY POPULATION

Suppose there exist a population in which the following are true:

l0
i. there are exactly births per calendar year and these births are uniformly
tl 0
distributed over the year , so that the number of births in a period of t years is
dx
ii. the mortality experienced is exactly that of the life table and the deaths are
uniformly distributed in the same way as births. In such a population the age
composition never changes and the total size remains constant since
l 0 =d 0 + d 1 +d 2 +⋯
deaths each year as the number of deaths is exactly matched
 by the number of births. Such a population is called a STATIONARY
POPULATION.
In a Stationary population the number of individuals aged x last birthday in a
Lx
stationary population is which is easily seen by noticing that at every moment,
individuals aged x last birth day are survivors of births between x and x+1 years.
x +1
L x= ∫ l y dy
x Tx
Therefore this yields: . As a result, is the number of individuals in the
T0
population aged x or more and in particular is the total size of the population. The
growth rate is zero.
Remarks
Real populations are not stationary because:
i. the number of births differ from year to year
ii. the mortality experience differs from year to year
iii. the births and deaths are not uniformly distributed over each year
STABLE POPULATION

A restricted amount of variation for a stationary population can be introduced by

supposing that the number of births is increasing at a constant rate, r, per year.
dP
=rP
dt
This gives rise to the differential equation

P= Ae rt
Which yields the solution where A is a constant

P0 = Aer ( 0 ) =A P1 =Ae r P2 =Ae 2 r Pn = Aenr Pn+1 =Ae r (n+1 ) =Pn e r

, , … ,

er
The total number of births in one year is times the total number of births in the
preceding year.
It is still supposed that the age specific death rate remain unaltered. As a consequence,
er
the total size of the population will be increasing at a proportional rate per year and
x+1
L x α ∫ e−ry l y dy
x
the proportionate composition by age will be stable with . Such a
population is called a STABLE POPULATION.
In a stable population:
 the crude birth and death rates are constant
 the age structure is fixed
 the total population size increases by a constant growth known as the intrinsic
growth rate
 the population is closed
Quasi stable population has similar characteristics as the stable population model
except that fertility is approximately constant and mortality is steadily declining.
Although natural populations are never stable but the idea of stable population is used
widely to:
 Analyze the implications of vital rates
 Show the manner in which animal will grow over time under specified
conditions
 Evaluate or correct age distributions
In the study of stable populations, the following assumptions are made:
i. The population is closed, that is, incrementation is only through births and
decrementation is only through deaths
ii. C(x,t) is the proportion of the population between x and x+d x at time t and C is
independent of t, i.e. C(x,t)=C(x)
iii. The survival function is also independent of time t, that is, l(0,t)=l0
As a consequence of these assumptions, the following properties hold for a stable
population
i. The intrinsic death rate d, is constant over time
ii. The intrinsic birth rate b, is constant over time
iii. The intrinsic growth rate r, where r=b-d is constant over time
iv. The proportion of age distribution is constant over time
The above assumptions and their consequences may be used to derive the relationship
between age structure, fertility, mortality and growth rates for a stable population.
The following derivations shown below shall apply only to female population but may
be extended to male or mixed population.
Let P(t) be the total number of females living at time t
C(x,t)=proportion of persons out of the total population of P(t) living at age interval [x,
x+dx]
 B (t )
b ( t )= =birth rate
P (t ) at time
t

r is the intrinsic rate of population increase, S(x)=probability of surviving from birth to

l
S ( x )= x
l0
age x, so that, f(x)=probability that a female child will be born

P ( t ) C ( x ,t ) dx
The number of persons between ages x and x+dx is and these are the
survivors of birth x years ago so that

P ( t ) C ( x ,t ) dx=B ( t−x ) S ( x ) dx

Multiplying both sides by f(x), yields the number of female births per woman among
the population

f ( x ) P (t ) C ( x ,t ) dx=f ( x ) B ( t−x ) S ( x ) dx

Integrating both sides of the above equation gives the total number of births per annum
at time t
∞
B ( t )=∫ B ( t−x ) S ( x ) f ( x ) dx
0

Thus, given S(x), f(x) which are always non-negative, the tast will be to determine the
form of B(t), the birth function.

B ( t )= Ae rt
Assume that Replacing this in the earlier equation gives:
.
∞
Aert =∫ Ae
(t− x ) r
S ( x ) f ( x ) dx
0
∞
rt
Ae = Ae rt
∫ e−rx S ( x ) f ( x ) dx
0
∞
⇒∫ e−rx S ( x ) f ( x ) dx=1
0

This equation is a basic equation and can be used to examine other features and
properties of the stable population.
Property 1: the basic equation above has one real solution r=r 0 for a defined interval
[a,b]
b
k ( r ) =∫ e−rx S ( x ) f ( x ) dx
0
b
d ( )
k r =−∫ xe−rx S ( x ) f ( x ) dx <0
dr 0
Define

k ( r ) →∞ as r →−∞
k ( r ) →0 as r →∞
Clearly
Since k(r) is a continuous function of r it follows that there is at least one real solution
for the equation. However, the functions x, e-rx, S(x), f(x) are all non-negative over the
interval [a,b] and simultaneously positive over part of the interval, hence k’®<0 for all
real x in [a,b], so k(r) is a monotonic strictly decreasing so there exists only one solution
of the equation k(r)=1

C ( x )=be−rx S ( x )
Property 2: Show that the age structure is stable and that
Let the proportion of females in the age groups [x,x+dx] be C(x,t), then,
r (t− x )
B ( t−x ) S ( x ) Ae S(x)
C ( x , t )= =
P (t ) P (t )
Ae e S ( x ) B ( t ) e−rx S ( x )
rt −rx
C ( x , t )= = =b ( t ) e−rx S ( x )
P (t ) P (t )

In a stable population, the mortality and fertility rate are constants, thus
b(t)=b, C(x,t)=C(x) and l(x,t)=lx

C ( x )=be−rx S ( x )

The relative age structure is independent of t and therefore it is stable.

Property 3: Relationship between birth rate, death rate and intrinsic rate of growth
S ( x )=l x ⇒ l 0=1
Set
∞ x+1
b ∫ e l x dx=1
−rx
L x α ∫ e−ry l y dy
0 x
if
 ∞ ∞
∫ C ( x) dx=∫ be−rx S ( x ) dx=1
0 0
Since
1 1
b= ∞ =∞
∫ e−rx S ( x ) dx ∫ e−rx l x dx
0 0
Hence
1
d= ∞ −r
∫e −rx
S ( x ) dx
Also r=b-d thus d=b-r , thus, 0