Teaching projections and

challenges regarding the

theory and practice

Presented at the International Seminar on Population Estimates and Projections:

Methodologies, Innovations and Estimation of Target

Population applied to Public Policies

9-11 November 2011,Rio de Janeiro

Moema Gonçalves Bueno Fígoli – CEDEPLAR/UFMG

context:

The idea that permeates this presentation is that while several methods for projecting population and its components are available, in practice the projections require analysis of the results under the light of knowledge of the determinants of demographic phenomena and demand subjective decisions / adjustments. In my opinion, this is the hardest part in teaching projection.

“Although the theoretical models available to the forecaster improve over time this does not lead to substantially more accurate forecast. (…) However, to effectively utilize the improved theoretical model, one must be able to accurately identify and forecast the determinants of change, and that has proved challenging. (Alho, Statistical Demography and Forecasting ).”

Goals:

In this context I wish to discuss • Estimates and projections of mortality and fertility

• Uncertainties in projections

• Errors in the population base

• Effect of errors in the estimates and projections of the components in the projection of population in time

• Courses offered by CEDEPLAR

Estimates fertility and mortality

Problem: lack of reliable data and

discrepancy between the correction methods

BR e Ufs (2001/2009): TFT estimadas

com dados das PNADs (sem correção)

0,5

2,0

3,5

2001 2002 2003 2004 2005 2006 2007 2008 2009

AC

RO

RR

AP

TO

MA

CE

RN

PB

PE

AL

SE

BA

MG

ES

RJ

SP

PN

SC

RS

MT

MS

GO

DF

PA

PI

AM

BR

Estimates fertility

What the previous data indicate is that the

TFT (calculated here directly) unexpectedly

fluctuate from year to year, this can be by

error from data, reference period, sample

size, etc..

What is the procedure for setting them?

The recipe would be to apply the method of

Brass? The traditional P2 /F2?

BR e Ufs (2001/2009):

P2/F2 - dados das PNADs

0,8

1,3

1,8

2001 2002 2003 2004 2005 2006 2007 2008 2009

AC

RO

RR

AP

TO

MA

CE

RN

PB

PE

AL

SE

BA

MG

ES

RJ

SP

PN

SC

RS

MT

MS

GO

DF

PA

PI

AM

BR

Estimates fertility

Obviously, due to the same reasons stated above, the oscillations in the ratios P2/F2 are evidence that the "recipe" can not apply.

Apply the "recipe" leads to even less plausible results than the results "not correct".

BR e Ufs (2001/2009): TFT estimadas

com dados das PNADs (com correção)

0,5

2,0

3,5

2001 2002 2003 2004 2005 2006 2007 2008 2009

AC

RO

RR

AP

TO

MA

CE

RN

PB

PE

AL

SE

BA

MG

ES

RJ

SP

PN

SC

RS

MT

MS

GO

DF

PA

PI

AM

BR

TFT* =s/correçao

TFT_= Corrigidas com P2/F2 do ano indicado

1,0

2,0

3,0

4,0

2001 2002 2003 2004 2005 2006 2007 2008 2009

AP_

AP*

AM_

AM*

Estimates of mortality

Esperança de vida ao nascer conforme dados do

registro civil -mulheres 2000

0,00

20,00

40,00

60,00

80,00

100,00

Ron

Amaz

Pará

TcoPia

uí

RG

NPE

RSE

MG R

J

Paran

áRG

SM

G DF

UF

℮o

Estimates of mortality

EEB Preston Média(1) EGB EGBM GE GEC Média(2) Média(3)Mulher Rondônia 1,466 1,477 1,471 0,853 0,822 1,129 1,035 0,943 1,072Mulher Acre 1,271 1,406 1,335 0,736 0,982 0,812 0,910 0,850 0,967Mulher Distrito Federal 1,105 1,136 1,120 0,962 0,838 0,917 0,969 0,918 0,978

Sexo UF

Fatores de correção de sub-registro por método

Projections of fertility

Literature suggests several ways: 1 - In countries where the transition is under way we analyze: • When did the Beginning, • What is the rate of decline - it is reasonable in a pace equivalent to the past (using mathematical models?) • How will be the structure (time quantum, cohort analysis) 2 - In countries where fertility is already below the replacement level • The fertility rate will continue to fall? • It will follow the pace of the past? • Will return to replacement level? • Will Stabilize above or below replacement level?

TFT: Suécia, Inglaterra e vários países sub - desenvolvidos

0,5

2,0

3,5

5,0

6,5

8,0

1870 1885 1900 1915 1930 1945 1960 1975 1990 2005

TFT

Sweden England Brazil China

Iran Thailand Mexico Chile

Projections of fertility

Looking at the graph we can indicate the beginning of fall and the pace past. Models of extrapolation would be indicated, especially for countries with the transition in progress?

How will be the structure?

All these questions can not be answered accurately – uncertainty

That is, the challenge is to teach how to incorporate formal models and substantive knowledge

Projections of mortality

Mortality, although greatly vary in level and

pattern, as your age pattern varies in a

predictable way according to the level of life

expectancy, the projection is usually based

on extrapolation of the historical trend or

model interpolation between boards.

Uncertainty in population projection

We do not know at what level fertility will reach We do not know that new diseases and / or new treatments will emerge We do not know the ways that migration will follow Must be taught how to take into account the uncertainties - scenarios, probabilistic projections?

Errors in the population base Brasil (2010) População por sexo e idade Brasil (2010) População por sexo e idade (Valores absolutos)(Valores absolutos) segundo:segundo:

-- Censo de 2010Censo de 2010

-- Projeções do IBGE revisadas em 2008. Projeções do IBGE revisadas em 2008.

8600000,0 6450000,0 4300000,0 2150000,0 0,0 2150000,0 4300000,0 6450000,0 8600000,0

H 2010 (Censo)

M 2010 (Censo)

H (projeção)

M (projeção)

Errors in the projections

Errors in the projections

Most projection errors occur in the population younger

and older.

The error in the population base contributes greatly to

the error in the first year of the projection.

Error in the estimation of components grows with time.

Main cause of error in the components is the initial

estimate, this can be improved with better data and

methodologies for components estimation and

projection

The courses offered by CEDEPLAR

Required - Application of demographic analysis

Discussion of sources of error in projection

• Adjust data to Projections

• Adjust the available estimates of fertility, and

methodologies for projection.

• Ditto for mortality.

• Ditto for Migration.

• Method Demographic Components

The courses offered by CEDEPLAR

2 - Options

a) projection multirreginal

b) Projection of small areas

c) Projections Advanced

Moema Gonçalves Bueno Fígoli

moema@cedeplar.ufmg.br