1. Population Studies & Research Institute
PROJECTIONS
Ben Jarabi
Population Studies & Research Institute
University of Nairobi

2. Projections - Objectives
At the end of the session, participants will be able to:
Outline the relevance of pop projections
Detect & adjust errors in age reporting
Prepare necessary inputs for a pop projection
Outline projection methodologies
Generate a pop projection at national level
Generate a pop projection at sub-national level
Generate sectoral pop projections 2

3. Presentation Outline A. Introduction
B. Evaluation and adjustment of census data
C. Preparation of inputs for pop projections
D. Projection methodologies
E. National and provincial projections
F. Sub-national population projections
G. Sectoral population projections 3

4. A. Introduction 4

5. Projections - Definition
A population projection is:
An extrapolation of historical data into the future
An attempt to describe what is likely to happen under certain explicit assumptions about the future as related to the immediate past
A set of calculations, which show the future course of fertility, mortality and migration depending on the assumptions used 5

6. Projections - Rationale
Primary needs of people in Kenya cannot be gauged rationally without information on expected size, composition and distribution of the population for different geographic units and at various points in time 6

7. Projections - Rationale
Specific areas in which projections are a necessary input:
Allocation of resources (planning)
Advocacy
Research
For M&E (basis for target setting; source of denominators esp. non-routine data)
To determine drivers of consumption – corporate sector 7

8. Projections - Inputs Base population
Assumptions about the course of events (fertility, mortality, migration)
A method by which the assumptions are applied

9. B. Evaluation & Adjustment9

10. Types of Errors Content Coverage Age mis-reporting Digit preference
Omitting a unit that should have been included
Including a unit more than once
Including a unit that should not have been included 10

11. Evaluation - Rationale
Age structure is very important with respect to social and economic characteristics - hence the need for accuracy in age and sex structure
Knowledge of age structure is essential to the analysis of fertility, mortality, and migration
Errors by age & sex are replicated and repeated in population projections 11

12. Detecting errors – age reporting
Age misreporting may be suggested by irregularities evident in indices or graphs
Population pyramid
Age and sex ratios
Cohort comparison
Summary indices of “irregularities” in
age structure or in age-sex structure 12

13. Evaluation - Digit Preference
Frequently used indices for detecting digit preference:
Myers
Whipple’s
Bachi
Ramachandran
They provide not only an overall idea of the extent of age misreporting but also indicate the preference for certain ending age digits 13

14. Evaluation - Age Ratios
Age ratios for 5-year age groups are used as indices for detecting possible age misreporting
Normally age ratios are expected to be similar throughout the age distribution, and all of them should be close to a value of 100 14

15. Evaluation - Age Ratios
An age ratio is defined as:
  Px
5ARx = 100
1/2 (5Px-5 + 5Px+5)
 Where: 5ARx = age ratio for ages x to x+4 
5Px = population at ages x to x+4
The larger the departure of this ratio from 100, the larger the error 15

16. Evaluation - Sex Ratios
The level of the sex ratios depends on the number of male and female births and on the mortality of the population
All populations have more male than female births, and so the sex ratio at the early ages is expected to be slightly over 100
Since mortality is usually higher for males than females, the sex ratio is reduced continuously up to the oldest ages 16

17. Evaluation - Sex Ratios
A sex ratio is defined as:
  MPx
5SRx = 100
5FPx
 Where: 5SRx = sex ratio for ages x to x+4 
5MPx & 5FPx = male & female populations, respectively, at ages x to x+4
The larger the departure of this ratio from 100, the larger the error 17

18. The Age-Sex Accuracy Index
The UN suggested a joint accuracy index to summarize the age & sex ratios
The index of sex-ratio score (SRS) is defined as: The mean difference between sex ratios for the successive age groups, averaged irrespective of sign
The index of age-ratio score (ARS) is defined as: The mean deviation of the age ratios from 100 percent, also irrespective of sign 18

19. The Age-Sex Accuracy Index
Based on empirical relationships between the sex-ratio scores and the age-ratio scores, the following index is defined as the joint score (JS) or age-sex accuracy index
JS = 3xSRS + ARSM + ARSF 19

20. The Age-Sex Accuracy Index
The age and sex structure of a population will be:
accurate if the joint score index is under 20
inaccurate if the joint score index is between 20 and 40
highly inaccurate if the index value is over 40 20

21. Correcting for Age Misreporting
Smoothing techniques have frequently been used for correcting data for age misreporting
These techniques involve the application of a formula to the original data 21

22. Correcting for Age Misreporting
Smoothing techniques may be classified into 2 categories:
Those which accept the population in each 10‑year age group & separate it into two 5‑year age groups without modifying the total population size
Those which smooth the 5‑year age groups and modify slightly (either up or down) the population being smoothed 22

23. Smoothing Methods Methods that preserve the original total:
The Carrier‑Farrag and Karup‑King‑Newton
The Arriaga formula
Arriaga’s “strong smoothing”
Methods that alter the total slightly:
The United Nations method 23

24. Smoothing Methods There is no generalized solution for all populations
The technique to be used will depend on the errors in the age and sex distributions
While, as Arriaga and Associates (1994) note, differences in results across procedures are small, a decision to use strong smoothing should not be taken lightly
The whole age distribution need not be smoothed if only part is considered problematic 24

25. Evaluation and adjustment
Exercise 1
Evaluation and adjustment 25

26. C. Preparation of inputs for population projections26

27. Smoothing Methods There is no generalized solution for all populations
The technique to be used will depend on the errors in the age and sex distributions
While, as Arriaga and Associates (1994) note, differences in results across procedures are small, a decision to use strong smoothing should not be taken lightly
The whole age distribution need not be smoothed if only part is considered problematic 27

28. Projections - Inputs
Accurate baseline data are critical to producing accurate population projections
Population size and age structure
Fertility
Mortality
Net migration 28

29. Projections - Inputs Base population Projection Fertility Mortality
Population adjusted for coverage, consistency with fertility & mortality, smoothed
Projection
Fertility and projected fertility
Fertility
Mortality & projected mortality
Mortality
Migration & projected migration
Migration 29

30. Inputs - Base Population
A population by sex & age is required to serve as the base population for the starting date of the projection
Usually, the base population is taken from the latest available census
Census enumerations are not always perfect - the reported data on age and sex may be affected by errors – hence the need for adjustment
Where necessary, move the adjusted population from a given date (e.g. a census date) to another date (e.g. midyear) 30

31. Projections - Inputs
Current levels of the demographic processes (fertility, mortality, and migration) are determined for the base year and then projected to future years

32. Projecting future fertility
Level of fertility has the greatest effect on population growth due to its multiplier effect: additional children born today will have additional children in the future
Fertility projections are made by projecting the course of TFR over time, and translating this into ASFRs
Normally, long-term projections assume that fertility will eventually stabilize at replacement level, leading to a stationary population 32

33. Projecting future mortality
Mortality projections are based on projecting future life expectancy at birth for males and females
Life expectancy (like TFR), is a period measure, and does not reflect the actual experience of a particular individual
Mortality projections must also specify how mortality is distributed over different age groups for both sexes 33

34. Projecting future mortality
Future migration levels are more difficult to project than fertility or mortality
Although fertility generally has a larger impact on long-term population growth, volatile migrations can exert a strong influence as well
In addition, since no single, compelling theory of migration exists, projections are generally based on past trends and current policies; however, data on historical migration are sparse 34

35. Exercise 2 Preparation of inputs35

36. D. Projection methodologies36

37. Projections - Methodology
Projection methodologies can be divided into two main categories:
procedures for projecting the population considering fertility, mortality, and migration, by age and sex (component method)
procedures for projecting the population using mathematical functions applied to population figures but not to each of the components (ratio method) 37

38. Projections - Methodology
Cohort component - projects separately, the components of population change (fertility, mortality & net migration)
Ratio method – generates total size of the population but not the age-sex structure
With HIV prevalence > 1%, mortality is projected to include HIV/AIDS impact 38

39. Cohort component method
This method simulates how a population changes according to its components of growth: fertility, mortality, and migration
Based on past information, assumptions are made about future trends in these components of change
Then, the projected rates are applied to the age and sex structure of the population, in a simulation taking into account that people die according to their sex and age, that women have children, and that some people change their residence 39

40. Cohort component method
Base population is grouped into cohorts defined by age and sex
The projection proceeds by updating the population of each age- and sex-specific group according to assumptions about three components of population change 40

41. Cohort component method
Each cohort survives forward to the next age group according to assumed ASMRs
Migration is accounted for by applying age- and sex-specific net migration rates to each cohort as well
Projected ASFRs rates are applied to the female population in childbearing ages to estimate the number of births 41

42. Cohort component method
A sex ratio at birth is used to divide total births into males and females
These births are exposed to the appropriate mortality schedule and then the survivors fed into the projection model 42

43. Cohort component method
Time span
No standard time span over which a projection should be made
Select a span that is equal to the maximum length of time required for completion of the planned activities
NB: the longer the time span, the greater the potential deviation of the projected from the actual population
It is usually most convenient to project population by time intervals equal to the age intervals 43 43 43

44. Ratio method
Ratio method is applied mainly for projecting the population of small areas within a country for which all inputs required by the component method are not always readily available
The method is also useful in the projection of urban and rural populations 44

45. Ratio method
The method is based on the ratio of the sub-national unit, say district, population to that of the national population
After the ratio of the district to national population is obtained, assumptions are made on the future values of ratios
Once the future values of ratios are fixed, the population of the district can be obtained by applying that ratio to the projected national population in that year 45

46. Ratio method Year 2001 2006 2011 2016 National population 640054
708185
(P2)
773854
(P3)
840603
(P4)
District population
78855 -
Ratio of district to ntl pop, 2001
0.1232
(X1)
Projected district pop
87248
(P2 x X1)
95338
(P3 x X1)
103562
(P4 x X1) 46

47. Ratio method
Once the projection for each small area has been made, ensure the sum of the population of all small areas tallies with the national total
Using the national total as a control, adjust proportionally the projections of the small areas 47

48. Population projection
There is no single method or technique that can improve accuracy
Accuracy depends on the quality of the input data and the assumptions made about the course of future change 48

49. Population projection
Identifying which projection method is optimal for a specific type of projection depends on several factors
Of crucial importance is whether the projections are to be carried out for larger geographical areas (e.g. nations and groups of countries) where uncertainty is lower, or smaller areas (e.g. sub-national, urban) where migration makes future population changes more volatile and projections more difficult 49

50. E. National and provincial projections50 50 50

51. Generating national and provincial projections
Exercise 3
Generating national and provincial projections 51 51 51

52. F. Sub-national population projections52 52 52

53. Sub-national projections
Generating sub-national projections that are both internally consistent and consistent with a national projection is usually more challenging than preparing a national projection
Each region presents the same data problems as the national projection but, in addition, preserving consistency across regions and dealing with data problems that are often more severe than those at the national level adds to the challenge 53 53 53

54. Generating sub-national projections
Exercise 4
Generating sub-national projections 54 54 54

55. G. Sectoral population projections55 55 55

56. Generating sectoral projections
Exercise 5
Generating sectoral projections 56 56 56