Population projections: Use and Interpretation

Population projections:
Uncertainty and the user
perspective
Presentation to
INIsPHO Seminar
Newry, 2 December 2008
Tony Dignan
Overview
•
•
•
•
•
Introduction
Sources of uncertainty
National projections
Sub-national projections
Concluding remarks
Uncertainty
• The future is inherently uncertain
• That is why users look to population projections
to help in planning for the future
– How many school places are required? Hospital
beds? Social and caring facilities?
• Dealing with uncertainty is also the major
challenge in producing forecasts
• How to manage uncertainty in using projections?
• Start by understanding the sources
Sources of uncertainty
• Assumptions
– Judgements made about the future course of events
– Based on what has happened in the (recent) past
– Informed by expert opinion on what the future holds
• But:
– Trends can change direction – turning point problem
– The recent past – distinguishing cyclical events and
longer-term trends
– External and/or unanticipated ‘shocks’ may occur
– Changing economic circumstances
Components of change
• Deaths
– The population ‘at risk’ is already born
• Fertility
– Subject to choices that people make in future years
– Social trends – can change
• Migration
– An array of influences – economic, social, etc
– Relative attractiveness of the country, region, for
which projections are being prepared
– The most volatile component
Past experience
• Changing trends
• Changing economic circumstances
Northern Ireland - Actual and projected population, 1971-2031
2,100
2,000
1,900
Historical
1971-based
'000s
1,800
1977-based
1,700
1985-based
1994-based
1,600
2004-based
2006-based
1,500
1,400
1,300
1951
1961
1971
1981
1991
2001
2011
2021
2031
Why was the 1971 forecast so far off the mark?
Total fertility rate: Northern Ireland, 1971-2006
3.3
3.1
2.9
TFR
2.7
2.5
Actual
1971 Assumed
2.3
2.1
1.9
1.7
1.5
1971
1981
1991
2001
End of the 1960s ‘baby-boom’ – sharp fall in fertility rates
Similar experience across UK and Europe (Shaw, 2007)
Population projections, Ireland, 1996-based
5,000
4,500
Actual
'000s
4,000
M1F1
M1F2
M1F3
3,500
M2F1
M2F2
M2F3
3,000
2,500
1951 1956 1961 1966 1971 1976 1981 1986 1991 1996 2001 2006 2011 2016 2021 2026 2031
The out-turn in 2006 was in excess of each 1996-based scenario – why?
Net migration, Ireland: Actual and assumptions - 1996-based
80
60
40
Actual
M1
M2
0
19
87
19
89
19
91
19
93
19
95
19
97
19
99
20
01
20
03
20
05
20
07
20
09
20
11
20
13
20
15
20
17
20
19
20
21
20
23
20
25
20
27
20
29
20
31
'000s
20
-20
-40
-60
Booming economy - Net inward migration exceeded expectations
Ireland: Actual 2006 compared with 1996-based projected
(M1F2)
All
Age group
65+
45-64
25-44
15-24
0-14
0.0
2.0
4.0
6.0
8.0
10.0
Per cent
Variations in out-turn by age group – older and younger more accurately
predicted – why?
12.0
Ireland: Migration by age group, 2007 - rate
per 1,000 population
60
50
Rate per 1,000
40
30
Immigration
Emigration
Net
20
10
0
-10
-20
0-14
Years
15-24
Years
25-44
Years
45-64
Years
65 Years
and Over
-30
Migration is strongly age-selective
Errors in migration assumption have less effect on older age groups
Over time, will affect younger age groups
National projections – 2006-based
• Variants/scenarios
– Alternative combinations of assumptions
– Typically a ‘high’ and a ‘low’
– NOT upper and lower ‘limits’ for a given confidence
level
– Reflect producers’ uncertainties
– But can also help the user manage uncertainty
• Uncertainties?
– Migration is especially uncertain at this time
Northern Ireland
• Net migration
– Recent inflows driven by EU Accession
countries
– Principal assumption is that this will end
– Revert to balance between in an out
• Variant projections
– High/low, but also young/old population
– Zero migration scenario not greatly different
from principal
Actual and projected net migration, NI, 1966-2031
15
10
5
0
'000s
1966 1971 1976 1981 1986 1991 1996 2001 2006 2011 2016 2021 2026 2031
-5
Principal
High
-10
-15
-20
-25
-30
Low
Principal and selected variant projections, N.
Ireland, 2006-based
2,300
'000s
2,200
2,100
Principal
High Pop
2,000
High Mig
1,900
Zero Mig
Young Pop
1,800
1,700
1,600
1,500
1,400
1971 1976 1981 1986 1991 1996 2001 2006 2011 2016 2021 2026 2031
Old Pop
Low Mig
Low Pop
Ireland
• Net migration
– M1 and M2 assume will stay strong
– Recent data suggests both are too high
– Also a zero net migration assumption (M0)
• Scenarios
– M1 and M2 – compare with M0
– Migration accounts for bulk of growth
– Five years before the next projection!
Net migration, Ireland: Actual and assumptions - 2006-based
80
60
40
0
19
87
19
90
19
93
19
96
19
99
20
02
20
05
20
08
20
11
20
14
20
17
20
20
20
23
20
26
20
29
20
32
20
35
20
38
20
41
'000s
20
-20
-40
-60
Actual
M1
M2
M0
Population projections, Ireland, 2006-based
7,000
6,500
6,000
Actual
'000s
5,500
M1F1
M1F2
5,000
M2F1
4,500
M2F2
M0F1
4,000
M0F2
3,500
3,000
2,500
1971
1981
1991
2001
2011
2021
2031
Sub-national projections
• Top-down – all components by area must
add to national totals
• Major strength – consistent framework
• But uncertainty will be greater
• Internal migration
– Net flows from one area to another
– Sum to zero at national level
– Additional component of change
– Hence more uncertainty
Northern Ireland
• 26 Local Government Districts (LGDs)
• Trends
– Internal migration - Belfast losing, surrounding areas
gaining
– External migration – Distinct geographical patterns
– Major gainers – LGDs in East of NI and West & South
• 2006-based Assumptions
– Both trends are apparent in the sub-national
projections
NI LGDs: Migration 2006, per 1,000 population
Belfast
Carrickfergus
Lisburn
North Down
Newtownabbey
Castlereagh
Craigavon
Antrim
Down
Ards
Banbridge
Ballymena
Larne
Ballymoney
Strabane
Limavady
Coleraine
Moyle
Derry
Dungannon
Cookstown
Newry & Mour
Armagh
Fermanagh
Magherafelt
Omagh
NI
External
Internal
-10
-5
0
5
10
15
Per 1,000 population
20
25
30
NI LGDs: Migration assumptions, 2006-2021,
annual averages
Belfast
Carrickfergus
Lisburn
North Down
Newtownabbey
Castlereagh
Craigavon
Antrim
Down
Ards
Banbridge
Ballymena
Larne
Ballymoney
Strabane
Limavady
Coleraine
Moyle
Derry
Dungannon
Cookstown
Newry & Mour
Armagh
Fermanagh
Magherafelt
Omagh
NI
-10
-5
0
5
10
15
Per 1,000 population
20
25
30
N. Ireland: Population growth by region, actual and
projected
130
Belfast
Rest of Greater
Belfast
East of NI
110
100
North of NI
90
West & South
of NI
80
20
20
17
20
14
20
11
20
08
20
05
20
02
20
99
19
96
19
93
19
90
19
87
19
84
19
81
70
19
Per cent of 2006
120
Projections show greater divergence in growth at sub-national level
than historically the case
N. Ireland: Spatial Development Strategy regions: Population
shares, actual and projected
55
54
53
Per cent
52
51
50
BMA & Hinterland
Rest of NI
49
48
47
46
19
81
19
84
19
87
19
90
19
93
19
96
19
99
20
02
20
05
20
08
20
11
20
14
20
17
20
20
45
Extrapolation of the most recent trends also apparent from shifts in
population shares
What are the uncertainties around that?
Ireland
• Eight Regional Authorities
• Most recent = 2002-based
• Three scenarios for internal migration
– Traditional
– Recent
– Medium
Ireland: Regional population shares
40
35
30
Per cent
25
Dublin
Rest of GDA
20
Rest of S&E
BMW
15
10
5
0
1951
1956
1961
1966
1971
1976
1981
1986
1991
1996
2001
2006
Long-term trend - Stable pattern from early-1970s to mid-1990s
Change from 1996 onwards – Dublin losing share.
Temporary or more permanent?
Internal migration flows - rates per 1,000
West
South-West
South-East
2006
2002
1996
Mid-West
Midland
Mid-East
Dublin
Border
-15.0
-10.0
-5.0
0.0
5.0
10.0
Per 1,000 population
Illustrates change in the pattern of internal migration
Dublin and Mid-East only ‘gainers’ in 1996
Dublin the major ‘loser’ in 2002 and 2006
15.0
Net migration by region: 2002-2006, annual
averages
25
15
10
Internal
External
Net
5
St
at
e
t
W
es
hW
es
t
So
ut
hEa
s
t
t
So
ut
-W
es
M
id
d
la
n
M
id
in
ub
l
-E
as
t
M
id
-10
D
-5
or
de
r
0
B
Per 1,000 population
20
-15
In recent years, Dublin has lost population due to internal migration
But gained most in relative terms from external migration
Dublin's population share: Actual and projected (2002based)
32
Per cent of State
30
28
Actual
Recent
Traditional
Medium
26
24
22
20
1951 1956 1961 1966 1971 1976 1981 1986 1991 1996 2001 2006 2011 2016 2021
Which projection to use?
Concluding Remarks
• Value of projections
– Consistent framework
– Detail of the projections
– Population by age and sex
– Births, deaths, migration effects
– Geographical patterns
– Guide to what will happen in future IF the
assumptions hold good
Concluding remarks
• The turning point problem
– Methodology based on extrapolation
– Change the methodology? Forecasting
model? Whole new set of problems
– Emphasises role of expert judgement
• Risk of over-reliance on the recent past
– Apparent changes may be cyclical rather than
permanent
– Need to have a coherent ‘story’
Concluding Remarks
• Uncertainty is inevitable
– “Any set of projections will inevitably be proved wrong, to a
greater or lesser extent, as a projection of future demographic
events or population structure” (Shaw, 2007)
– Some components more uncertain than others
– How to manage that?
– Understanding the assumptions – qualitative assessment and
indicators
– Scenarios/variants will help to identify the key sources of
uncertainty
– User can undertake sensitivity analysis
– Some scenarios will be more useful than others, depending on
what user is seeking to achieve
Concluding Remarks
• Official projections are ‘trend-based’
• But some policies may have the effect of
changing the trends
• Or be intended to change the trends e.g. spatial
development strategies
• Need for policy based forecasts? But:
– Policy takes time to affect outcomes
– Policy outcomes are themselves uncertain
– Trend-based provides base for comparison – what
needs to be done to achieve a planned outcome?