POPULATION AND HEALTH : TECHNIQUES OF ANALYSIS AND POLICY PERSPECTIVE

POPULATION PROJECTION
1

MA (SDP), 2nd Semester, 2012

PROJECTION

A projection may be defined as the numerical outcome of a particular set of assumptions regarding the future population Estimates or forecasts are made, in absence of census or related data Some methods projects for total population Other for sub-groups i.e., age, sex, race, other demographic and socio-economic characteristics

POPULATION PROJECTION: FUTURE
ESTIMATES

Raise our understanding of the determinants of population change
y

What impact would a 20% decline in birthrates have on a country·s population size and age structure in 50 years? How many people would move into a local area if a new factory employing 1000 people were opened?

y

Projections often serve as a basis for producing other projections (e.g., households, families, school enrollment, income, and labor force) Provides scopes for informed decision making for policy makers

ASSUMPTIONS IN PROJECTION

Every projections/estimates/forecasts are made on certain assumption It does not attempt to predict whether those assumptions actually will hold true Range of scenarios and made alternative assumption

sites for fire stations.to predict demands for their products.SOME APPLICATIONS: National population projections.determine future water demands.Future Social Security and Medicare obligations State projections. . Business enterprises. welfare expenditures Local projections.new public schools. demand for housing The assumptions are made based on past-trends of population change. since ¶future is intimately tied to the past·.

these are often not ¶mutually-exclusive· . (2001) classified objective methods under(1) Trend extrapolation. can be replicated exactly Smith et al. and (3) Structural models (4) Other socio-economic projections However. techniques. techniques.POPULATION PROJECTION METHODS Subjective methods: data. and assumptions are clearly identified. (2) Cohort-component methods. difficult for replication Objective methods: data. and assumptions are not clearly identified.

INPUT DATA-SOURCE FOR PROJECTION Census Population Registers Vital Registers Other records Surveillance statistics Etc. .

.USEFUL TERMS: Base year: year of the earliest data used Launch year: year of the most recent data used to make a projection Target year: year for which the population is projected Base period: the number of years between the base year and launch year Projection horizon: number of years between the launch year and target year Projection interval: the time increment for which projections are made (e. annually or every 5 years) .g.

TREND EXTRAPOLATION .

logistic curve.TREND EXTRAPOLATION Involves fitting mathematical models to historical data and using these models to project population valuesSimple extrapolation methods: requires data for only two dates linear change. and exponential change. Complex extrapolation methods: requires data for a number of dates to calculate. geometric change.. i. polynomial curve. constant share. linear trend.e. shift share. and share of growth . and ARIMA time series Ratio extrapolation methods: population of a smaller area is expressed as a proportion of the population of its larger.

SIMPLE EXTRAPOLATION .

where t = projection horizon . where.LINEAR PROJECTION Assumption: population will change by the same numerical amount over a given period.Po)/t. Pt=Launch year y Po= Base year y t= Base period y Target year (Y)= Pt+(t * ). as has registered in the past Linear change: Pt=Po (1+r*t) = (Pt.

799 (Official statistics.443 Project population for year 2010 r= 2010 0. 2000= 169.LINEAR PROJECTION.755.119. of Brazil) .033719 194.400.071.190. Govt. EXAMPLE Brazil 1960 = 70.544.786 2010 .

GEOMETRIC PROJECTION Assumption: population will change by the same percentage rate over a given increment of time in the future as during the base period Geometric change: Pt=Po (1+r)t r= (Pt/Po)1/t -1 where. r= annual rate of change y Pt=Launch year y Po= Base year y t= Base period y Target year (Y)= Po(1+r)t where t = projection horizon .

2000= 169. Govt.799 (Official statistics.680 2010 . EXAMPLE Brazil 1960 = 70.022318418 211.071. of Brazil) .443 Project population for year 2010 r 2010 0.419.119.GEOMETRIC PROJECTION.755.544.190.

r= annual rate of change y Pt=Launch year y Po= Base year y t= Base period y Target year (Y)= Po*ert where t = projection horizon . but it views change as occurring continuously rather than at discrete Intervals Exponential change Pt= Po* ert r= {ln(Pt/Po)}/t Where.EXPONENTIAL GROWTH Assumption: closely related to the geometric one.

119. of Brazil) .443 Project population for year 2010 r 2010 0.071.799 (Official statistics.680 2010 .755.419. EXAMPLE Brazil 1960 = 70. 2000= 169.022073007 211.544.EXPONENTIAL PROJECTION.190. Govt.

COMPLEX EXTRAPOLATION METHOD .

COMPLEX EXTRAPOLATION METHOD Uses base-period data for more than two dates. Better deals with nonlinear population change Three basic steps: To assemble historical population data for different dates during the base period To estimate the parameters of the model selected to generate the projection. Fit of a particular model should be understood by studying base year parameter data To generate projections using the model(s) selected .

EXTRAPOLATION METHOD Linear Models: Yi = a + b(Xi ). Where.µ a= is the constant term b= is the slope of the line describing the ´best fitting linearµ relationship between X and Y . Yi is a set of i observations of values of a ´dependent variable. Xi is a set of i observations of an ´independent variable.

years) over the period i i= b to l (b = base year and l = launch year) a and b are the estimated intercept and slope. Pi =is the population for a set of time points (e.g. respectively Ti= is time over the period i = b to l Regression Equation(y) = a + bx Pi = a + b(Ti ) Slope(b)= (N Ti*Pi.( Ti)2) Intercept (a)= ( Pi .b( Ti)) / N N= number of values/ elements .( Ti)( Pi) / (N Ti2 ..LINEAR EXTRAPOLATION Pi = a + b(Ti ) Where.

( Ti)2) ( Pi b( Ti)) / N Ti*Pi Ti*Ti b a N=5.TO CALCULATE Time period (Ti) 1990 1995 2000 2005 Population (Pi) Y1 Y2 Y3 Y4 (N Ti*Pi.( Ti)( Pi) / (N Ti2 . and once all the values are known. projection can be done for Year 2010 Pi = a + b(Ti ) .

POLYNOMIAL MODELS Polynomial models is used for projections in which change is not constrained to be linear Where. Yi =is a set of i observations of values of a dependent variable Xi= is a set of i observations of an independent variable a =represents the constant term bj = the slope of the line describing the ´best-fittingµ relationship between Xji and Y .

LOGISTIC MODELS Logistic approach explicitly allows one to place an upper limit on the ultimate size of the population for a given area The logistic model is consistent with Malthusian and other theories of constrained population growth Where. Y is the population X is the time period a reflects the upper asymptote b and c are parameters that define the shape of the curve e is the base of the natural logarithm .

ARIMA (´AUTOREGRESSIVE INTEGRATED MOVING AVERAGEµ) ARIMA models have occasionally been used in the analysis and projection of populations as a whole and of their demographic attributes. It deals with time-series data .

RATIO EXTRAPOLATION METHOD .

and (3) the share-of growth approaches. Ratio extrapolation methods may be used where an area containing the population to be projected is part of a larger (´parentµ) area for which projections are available The major types include. (1) the constant-share. (2) the shift-share. .

Pit= is the population projection for smaller area (i) in the target year Pil= is the population of the smaller area in the launch year Pjl= is the population of the parent area ( j) in the launch year Pjt is the projection of the parent area in the target year The constant-share method requires historical data for only one date Major drawback.assumes all small areas will grow at same pace .THE CONSTANT SHARE Where.

respectively It can lead to substantial population losses in areas that grew very slowly (or declined) during the base period . y is the number of years in the base period. launch. and b. The smaller area is denoted by i. z is the number of years in the projection horizon. and t refer to the base.SHIFT-SHARE METHOD Where. and target years. the parent area by j. l.

SHARE-OF-GROWTH METHOD Deals with shares of population change rather than population size .

LIMITATION OF EXTRAPOLATION METHOD Do not account for differences in demographic composition or for differences in the components of growth Provide little or no information on the projected demographic characteristics of the population Limited usefulness for analyzing the determinants of population growth or for simulating the effects of changes in particular variables or assumptions .

COHORT COMPONENT .

and later rediscovered independently by Whelpton (1928) Cohort-component method divides the launch-year population into age-sex groups (i. and migration that the cohort goes through in projection horizon Provide in-depth knowledge on population dynamics Cohort-component models typically use either single years or 5-year groups. subsequently used by Bowley (1924).COHORT COMPONENT Most widely used method for producing national-level states/provinces. birth cohorts) and accounts separately for the fertility. and sub-county area population projections Introduced by Cannan (1895). mortality.e.important methodological advancement ..

STEPS IN COHORT COMPONENT METHOD Step 1: Mortality Step 2: Migration (Out or net) Step 3: Fertility Step 4: Population (M/F) .

Simple Assumption Long-horizon needs adjustment for changing rates with changing socio-economic scenario Various extrapolation techniques are used.PROJECTING MORTALITY Age-specific mortality will remain constant (shorthorizon).future will mirror the past in important ways . based on assumption on various socio-economic determinants shaping mortality trends Assumption.

considering younger cohort in the same population. cultural. instead of same cohort in another population Mortality probability gets reduced due to advancement in medical technology Mortality probability can also be taken from life-tables . medical technology and primary causes of death Targeting approach.CONTINUED« Number of techniques ties mortality rate in one population to those in another ¶Targeting approach·.is based on assumption realistic for population to be projected Choice is based on similarities in SE. behavioral characteristics.¶cause-delay·.

and also for countries that completed demographic transition y ¶Targeting approach·.assumed will remain constant through projection horizon.checking for realistic assumption for convergence .PROJECTING FERTILITY Cohort perspective Period perspective (more commonly used) y Current ASFRs. Data based on recent year data or average data for several years Non-linear models used for projecting ASFRs for a wider horizon.

both used widely Gross migration approachy Out-migration rate: Age-sex data for decennial census Age-sex data (5year before the census) y Applying out-migration rate to the launch year of population to provide a projection of the total ¶pool· of inter-state out-migrants for all states.PROJECTING MIGRATION Importance in population growth at national and subnational level Gross & Net migration approaches. This ¶pool· is then allocated to each state based on proportion of inmigrant population received in the base period y .

CONTINUED«. based on contemporary and historical trends Local and national level projections needs separate focus . Net Migration Approach: y Top down: 1) Projection on net-migration 2) Projection on age-sex specific migration Bottom-up: 1) Separate age-sex group projection 2) Projection of net-migration based on age-sex ratio y Projecting migration need careful considerations.

STRUCTURAL MODELS FOR PROJECTION .

.e. nations Urban-Systems model: small geographic areas for land-use. states. . labour-market. i. transportation pattern etc.STRUCTURAL MODELS FOR PROJECTION Adjusting for factors beyond demography. Economic-demographic models: projection of population for larger geographic region and economic activity.

OTHER SOCIO-ECONOMIC PROJECTION .

µ and ´incidence rate methodµ) Cohort progression method .OTHER SOCIO-ECONOMIC PROJECTION Most useful for feeding policy-requirement and directly affected by policy decision Two major approaches: ´Participation ratio methodµ (also known as the ´participation rate methodµ. ´prevalence ratio method.

PARTICIPATION RATIO METHOD Current and historical data are used to construct participation ratios These ratios are projected into the future Projected ratios are then applied to population projections (stratified by age. sex. and other characteristics) for the geographic area(s) under consideration to obtain a set of socioeconomic projections Must have sufficient demographic detail .

.COHORT PROGRESSION RATIO Numbers with the socioeconomic characteristic or the corresponding participation ratios are projected on a cohort basis Using information on changes in the numbers or participation ratios between two previous dates.

Chapter 21 .REFERENCE Methods and Materials of Demography.