Creating Socio-Economic Household Data at the Small

Area Level: An Introduction to Spatial Microsimulation

Ann Harding

Presentation to Department of Geography Seminar Series, University of California Santa Barbara, USA, 12

May 2008

National Centre for Social and Economic Modelling (NATSEM), University of Canberra

2

What are microdata and microsimulation models?

Focus on individuals or households Start with large microdata sets (admin or sample

survey) Primarily used to estimate impact of government

policy change on these individuals or households• Impact on small sub-groups

• Aggregate impact

• Impact on government revenue or expenditure

Static models of taxes and transfers

4

Income tax and social security

STINMOD model is now maintained by NATSEM for Australian government departments (Family and Community Services, Education Science and Training, Treasury, Employment and Workplace Relations)

13 years old STINMOD simulates all the major income tax and cash transfer

programs (age pension, family payments etc) Used regularly in research – distributional impact of welfare

state, impact of minimum wage rises, EMTRs Constructed on top of Income Surveys

and Expenditure Surveys (2 versions)

5

6

7

8

Disposable income of sole parents with one child aged 8+, 2006-07 Impact of 2005 ‘welfare to work’ budget changes

300

400

500

600

700

800

0 100 200 300 400 500 600 700 800Private Income ($ pw)

Dis

po

sa

ble

Inc

om

e (

$ p

w)

Current policy - disposable income

New policy - disposable income

9

EMTRs of sole parents with one child aged 8+, 2006-07 *

* EMTR of 65% means that person keeps 35 cents from an additional dollar of earnings

0

10

20

30

40

50

60

70

80

0 100 200 300 400 500 600 700 800Private Income ($ pw)

ET

R (

%)

Current policy - EMTRs New policy - EMTRs

Dynamic models

11

Dynamic microsimulation modelling

Simulates the events that happen to ordinary Australians over their lifetime

Starts in 2001 with 180,000 people (1% of the Australian population – the Census sample)

Models individuals (the micro level) Uses regression equations to model human

behaviour over time (dynamic) APPSIM (Australian Population and Policy Simulation

Model) currently under devt – won ARC grant last year with 13 govt depts as research partners

Earlier version was DYNAMOD3

12

DYNAMOD3’s Simulation Cycle

Immigration

Macro-economics Emigration

Annual EarningsAnd Savings

Labour Force

EducationCouple Formation

Y

YNew Quarter?

Pregnancy and Births

Couple Dissolution

DeathsDisability & Recovery

Simulation Clock

Australia 2001

Australia 2050

New Year?

Spatial models

14

Characteristics of available datasets

National sample surveys

Census of Popn & Housing

?

Population detail

High Medium High

Geographic detail

Low High High

15

Synthetic Spatial Microdata

Solution:

Combine the information-rich survey data with the geographically disaggregated Census data

Using ‘spatial microsimulation’ (synthetic estimation) to:

create detailed unit record data for small areas(synthetic spatial microdata)

16

Constructing small-area estimates SMALL AREA DATA

2001 Census data at SLA

level: - XCP data for SLAs

UNIT RECORD DATA (SOURCE)

1998-99 Household Expenditure Survey

UNIT RECORD DATA (AMENDED)

- Updated to 2001

-

Enhanced income

REWEIGHTINGUSING

LINKING VARIABLES

SMALL-AREA ESTIMATES

1) Unit record dataset

2) Set of weights for each SLA

Iterative process to identify a set of variables suitable for reweighting

Major data preparation task

17

What is ‘reweighting’?

turning the national household weights in the HES survey file into …

… household weightsof small-areas

Unit record

Household ID

Weekly income

Weekly rent

Other variables

Household weight

1 1 7 3 . 10292 2 11 4 . 1573 2 11 4 . 1574 2 11 4 . 1575 3 11 0 . 10036 3 11 0 . 10037 4 10 4 . 708 4 12 4 . 709 6 12 0 . 703

10 6 12 0 . 703. . . . . .. . . . . .

13964 6892 . . . .

7,122,000Num of households in Aust

NSW SLA1

NSW SLA2

NSW SLA3

Other SLAs

0 0 0 .0 0 0 .0 0 0 .0 0 0 .

2.45 13.54 16.38 .2.45 13.54 16.38 .

0 0 0 .0 0 0 .

3.27 0 0 .3.27 0 0 .

. . . .

. . . .

. . . .

12465 25853 27940 .

Num of households in small areas

18

Linkage variables available in the 2001 Census and 1998-99 HES

Family level variablesFamily typeFamily income

Household level variablesDwelling structureTenure typeHousehold incomeHousehold typeHousehold sizeNumber of dependentsNumber of carsRent paidMortgage repayments

Person level variablesAgeSex

Social marital statusCountry of birthLevel of schoolingNon-school qualificationsEducational institution attendingStudy statusHours workedIndividual incomeOccupationLabour force statusYear of arrivalRelationship in household

19

Application 1: Analysis of Specific Population Sub-Groups

Allows – for small areas:

• identification and analysis of specificsocio-demographic groups and characteristics

• analysis at various population levels:e.g. persons, income units, households

Examples – children in low income families; children in jobless families; unskilled youth, those in housing stress

20

Ipswich (C) - East

Ipswich (C) - Central

Wacol

Archerfield

Chermside

Nundah

Clayfield

Nathan

Windsor

Annerley

Fairfield

Inala

Richlands

New Farm

Kelvin Grove

Toowong

Indooroopilly

East Brisbane

Dutton ParkTaringa

Bowen Hills

Fortitude Valley - Remainder

Milton

St Lucia

South Brisbane

Spring Hill

Highgate HillWest End (Brisbane)

Kangaroo PointCity - Remainder

Percentage of householdsin unaffordable housing

0.88% - 6.05%

6.06% - 9.48%

9.49% - 14.00%

14.01% - 29.27%

21

Hervey Bay (C) - Pt B

Boulia (S)

Belyando (S)

Mount Isa (C)

Longreach (S)

Duaringa (S)

Broadsound (S)

Nebo (S)

Emerald (S)

Peak Downs (S)

Pine Rivers (S) Bal

Calliope (S) - Pt A

Guanaba-Currumbin Valley

Beaudesert (S) - Pt A

Cairns (C) - Trinity

Thuringowa (C) - Pt A Bal

Cairns (C) - Northern Suburbs

Tara (S)

Cook (S) (excl. Weipa)

Etheridge (S)

Herberton (S)

Wondai (S) Kilkivan (S)

Kolan (S)

Miriam Vale (S)

Tiaro (S)

Nanango (S)

Mount Morgan (S)

Households in poverty (%)

0 - 8

9 - 12

13 - 17

18 - 30

22

Age profile of those in poverty in postcode in metro Sydney

35-44 years38%

45-54 years16%

55-64 years15%

65+ years2%

25-34 years27%

< 25 years2%

(Aust average 4.1%)

(Aust average 20.3%)

(Aust average 35.1%

(Aust average 17.7%

(Aust average 12.3%)

(Aust average 10.5%)

23

Application 2: Predict spatial impact of a policy change

Spatial microdata now linked with NATSEM’s existing microsimulation models to model the immediate distributional/revenue impact of a policy change

• link synthetic spatial output to STINMOD and model changes to the tax and transfer system for small geographic areas

• Currently modelling changes in Commonwealth Rent Assistance, income tax, social security and family payments

24

Where did the $5bn of 2005-06 tax cuts go?

Updated 2001 population numbers to 2005-06 using ABS estimates pop’n growth by SLA

Updated household incomes and rules of govt programs to 2005-06

Tax threshold Tax rate

Tax threshold Tax rate

$6,000 0.17 $6,000 0.15

$21,600 0.3 $21,600 0.3

$58,000 0.42 $63,000 0.42$70,000 0.47 $95,000 0.47

2004-05 2005-06

25

Legend

CUT_HH_TTA3.5 - 9.59.6 - 13.713.8 - 19.319.4 - 34.1

±20

Km

Estimated average tax cut per household per week, Sydney SLAs, 2005-06

$3.50 - $9.50 (lightest)

$9.51 - $13.70

$13.71 - $19.30

$19.31 - $34.10 (darkest)

26

Estimated average dollar tax cut per household per week, by regions, 2005-06

State/ Territory

Capital city Major urban Regional town Rural town Rural area All regions

NSW $14.6 $9.1 $8.8 $9.0 $8.0 $12.2 (1.3) (0.8) (0.8) (0.8) (0.7) (1.1)

VIC $12.6 $9.7 $8.3 $8.4 $8.5 $11.5 (1.1) (0.8) (0.7) (0.7) (0.7) (1.0)

QLD $11.0 $9.0 $8.8 $7.1 $9.4 $9.9 (1.0) (0.8) (0.8) (0.6) (0.8) (0.9)

ACT $16.0 na na na $11.7 $16.0 (1.4) (1.0) (1.4)

All four $13.3 $9.1 $8.6 $8.5 $8.6 $11.5 (1.2) (0.8) (0.7) (0.7) (0.7) (1.00)

Note: Regions are aggregates of SLAs. Figures in bracket are multiples of $11.5. Convergent SLAs only.

Example of aggregating the microdata

27

HOUSEMOD

Spatial model (SLA) Models receipt of Commonwealth Rent Assistance

• Means-tested assistance to low income private renters

Can change rules of CRA and predict spatial impact Has been extended to add:

• public renters as well

• plus projections for 20 years

28

% in unaffordable housing

29

Application 3: Where to put govt offices? (access channel planning)

The Centrelink CuSP Model (Customer Service Projection Model)

Centrelink needed an evidence based methodology to help:• match services available to customers’ needs and preferences

• deliver the service via the most suitable channel and

• in most efficient way.

CuSP model assists Centrelink strategic decision-making by:

• producing projections of Centrelink customers and channel use

• over the next 5 years

• for small areas

• and under alternative scenarios about the future

30

Projected % changes in customer numbers – 2002-07

31

Application 4: Forecasting current & future need for services

CAREMOD model simulates current characteristics of older Australians at a detailed regional level (SLA)• Imputing functional status and thus likely need for different types of

care

• Industry partners: NSW Dept of Ageing, Disability and Home Care and Fed Dept of Health and Ageing (2003-2005)

Also new ARC grant 2007-2009 for examining spatial implications of population ageing over next 20 years (esp. for needs-based planning of govt services) – with four states and territories

32

Man ly

Curt in

Eppi ng

Mos man

Burban k

Toow ong

Kenmore

Ipsw i ch

Cronul l a

Rose Ba yVaucl use

Red Hi llAshg ro ve

Lane Cov e

Bel levu e H il l

±

PERCENTAGE OF POPULATION AGED 55

S Y D N E YI N S E T

Quintiles (Percent)

2.1 - 7.9

7.9 - 9.0

9.0 - 10.0

10.0 - 11.2

11.2 - 38.6

WITH CARE MODALITY 5

100 0 10050

Kilometres

YEARS AND OVER

% needing high level institutional care

33

Where do self-funded retirees live

34

Other forthcoming ARC funded & ARCRNSISS initiatives By end 2008, estimates of poverty, housing stress & smoking

expenditure to be available via web to ARCRNSISS members• At labour market area level

• At SLA level

• Based on new 03-04 Income and Household Expenditure Surveys

Have also developed small area index of social exclusion for children

Would like to link the SLA estimates to administrative data about usage of government services• Eg do poor children use public health services more or less than children

from affluent families?

Continue to refine the technology

35

Evidence based policy making

Growing demand for decision support tools• Reduce risk to policy makers making billion dollar decisions

• Assess distributional implications of policy change before implemented

• Improve predictive capacity & strategic planning

NATSEM has now constructed dozens of microsimulation models, based on ABS or admin microdata

Exciting new developments are• Spatial microsimulation models

• Health and housing models

• Next generation of dynamic models

For free copies of all publications as released, email

hotline@natsem.canberra.edu.au

36

Selected referencesSpatial Microsimulation (spatial estimates of poverty, disadvantage etc)Lloyd, R, Harding, A and Greenwell, H, 2001, 'Worlds apart: postcodes with the highest and lowest poverty rates in today's Australia', Eardley, A, and

Bradbury, B (eds), Refereed Proceedings of the National Social Policy Conference 2001 , SPRC Report 1/02, pp. 279–97 (www.sprc.unsw.edu.au)

Harding, A., Lloyd, R., Bill, A., and King, A., ‘Assessing Poverty and Inequality at a Detailed Regional Level – New Advances in Microsimulation ’, in M McGillivray (ed), Perspectives on Human Wellbeing, UN University Press, Helsinki.

Taylor, E, Harding, A, Lloyd, R, Blake, M, “Housing Unaffordability at the Statistical Local Area Level: New Estimates Using Spatial Microsimulation ”, Australasian Journal of Regional Studies, 2004, Volume 10, Number 3, pp 279-300

S.F., Chin, A., Harding, R., Lloyd, J., McNamara, B,.Phillips and Q., Vu, 2006, “Spatial Microsimulation Using Synthetic Small Area Estimates of Income, Tax and Social Security Benefits” , Australasian Journal of Regional Studies, vol. 11, no. 3, pp. 303-336

Chin, S.F. and Harding, A. 2006, Regional Dimensions: Creating Synthetic Small-area Microdata and Spatial Microsimulation Models. Technical Paper no. 33, April*

Child Social Exclusion Index (small area index of social exclusion specifically developed for children)Harding, A., McNamara, J., Tanton, R., Daly, A., and Yap, M., “Poverty and disadvantage among Australian children: a spatial perspective” Paper for presentation at 29th

General Conference of the International Association for Research in Income and Wealth , Joensuu, Finland, 20 – 26 August 2006

CuSP Model (spatial model for service delivery and access issues)King, A. , 2007, ‘Providing Income Support Services to a Changing Aged Population in Australia: Centrelink’s Regional Microsimulation Model’, in Gupta,

A and Harding, A , 2007, Modelling Our Future: Population Ageing, Health and Aged Care, North Holland, Amsterdam.CAREMOD (spatial model of care needs)

Brown, L and Harding, A. 2005, ‘The New Frontier of Health And Aged Care: Using Microsimulation to Assess Policy Options’, Quantitative Tools for Microeconomic Policy Analysis, Productivity Commission, Canberra (www.pc.gov.au/research/confproc/qtmpa/qtmpa.pdf )

L, Brown, S, Lymer, M,Yap, M,Singh and A, Harding “Where are Aged Care Services Needed in NSW – Small Area Projections of Care Needs and Capacity for Self Provision of Older Australians” , Aged Care Association of Victoria State Conferences, May 2005 *

Lymer, S., Brown, L. Harding, A. Yap, M. Chin, S.F. and Leicester, S. Development of CareMod/05, NATSEM Technical Paper no. 32, March 2006*

HOUSEMOD (spatial model of housing assistance and housing issues)King, A and Melhuish, T, 2004, The regional impact of Commonwealth Rent Assistance, Final report, Australian Housing and Urban Research Institute,

Melbourne, November (www.ahuri.edu.au)

Kelly, S., Phillips, B. and Taylor, E., “Baseline Small Area Projections of the Demand for Housing Assistance”. Final report, The Australian Housing and Urban Research Institute RMIT-NATSEM AHURI Research Centre. May 2006 (ahuri.edu.au)

DYNAMOD (dynamic microsimulation model - now being replaced by APPSIM)Kelly, S, Percival, R and Harding, A., 2001, ‘Women and superannuation in the 21st century: poverty or plenty?’, Eardley, A, and Bradbury, B (eds),

Refereed Proceedings of the National Social Policy Conference 2001 , SPRC Report 1/02, pp. 223–47. (www.sprc.unsw.edu.au)

Kelly, S, 2007, ‘ Self Provision In Retirement? Forecasting Future Household Wealth in Australia’ , in Harding, A and Gupta, A., Modelling Our Future: Population Ageing, Social Security and Taxation (eds), North Holland, Amsterdam.

Cassells, R., Harding, A. and Kelly, S., 2006, Problems and Prospects for Dynamic Microsimulation: A Review and Lessons for APPSIM . NATSEM Discussion Paper no. 63, December *.

* Means available on NATSEM website at www.natsem.canberra.edu.au

37

Selected referencesSTINMOD applications (static tax-benefit model)Toohey, M and Beer, G, 2004, Financial incentives to work for married mothers under A New Tax System’, Australian Journal of Labour Economics, vol. 7,

no. 1, p. 53–69, January Harding, A., Warren, N., Robinson, M. and Lambert, S., 2000, ‘The Distributional Impact of the Year 2000 Tax Reforms in Australia’, Agenda, Volume 7,

No 1, pp 17-31. McNamara, J, Lloyd, R, Toohey, M and Harding, A, 2004, Prosperity for all? How low income families have fared in the boom times, Report commissioned

by the Australian Council of Social Service, the Brotherhood of St Laurence, Anglicare NSW, Family Services Australia, Canberra, October.*

A. Harding, R. Lloyd & N. Warren, 2006, "The Distribution of Taxes and Government Benefits in Australia", in Dimitri Papadimitriou. (ed), The Distributional Effects of Government Spending and Taxation, Chapter 7, Palgrave Macmillan, New York., pp. 176-201.

Harding, A, Vu, Q.N, Percival, R & Beer, G, “ Welfare-to-Work Reforms: Impact on Sole Parents” Agenda, Volume 12, Number 3, 2005, pages 195-210 (www.agenda.anu.edu)

Harding, A., Payne, A, Vu Q N and Percival, P., 2006, ‘Trends in Effective Marginal Tax Rates, 1996-97 to 2006-07’, ,AMP NATSEM Income and Wealth Report Issue 14, September (available from www.amp.com.au/ampnatsemreports)

Lloyd, R, 2007, ‘STINMOD: Use of a static microsimulation model in the policy process in Australia’, in Harding, A and Gupta, A., Modelling Our Future: Population Ageing, Social Security and Taxation (eds), North Holland, Amsterdam.

CHILDMOD (static child support model)Ministerial Taskforce on Child Support, 2004, In the Best Interests of Children – Reforming the Child Support Scheme, Report of the Ministerial Taskforce

on Child Support, May (see Chap 16 for output from CHILDMOD) (http://www.facsia.gov.au/internet/facsinternet.nsf/family/childsupportreport.htm)

NSW Hospitals Model (spatial model of socio-economic status and hospital usage and costs)Walker, A., Thurect, L and Harding, A. 2006, Changes in hospitalisation rates and costs – New South Wales, 1996-97 and 2000-01. The Australian

Economic Review, vol. 39, no. 4, pp. 391-408. (Dec)Walker, A. Pearse, J, Thurect, L and Harding, A. 2006, Hospital admissions by socioeconomic status: does the ‘inverse care law’ apply to older

Australians? Australian and New Zealand Journal of Public Health, vol. 30, no. 5, pp 467-73. (October)Thurecht, L, Bennett, D, Gibbs, A, Walker, A, Pearse, J and Harding, A., 2003, A Microsimulation Model of Hospital Patients: New South Wales, Technical

Paper No. 29, National Centre for Social and Economic Modelling, University of Canberra.*Thurecht, L, Walker, A, Harding, A, Pearse, J, 2005, ‘The “Inverse Care Law”, Population Ageing and the Hospital System: A Distributional Analysis’,

Economic Papers, Vol 24, No 1, March A, Walker, R, Percival, L, Thurecht, J, Pearce, 2005 “ Distributional Impact of Recent Changes in Private Health Insurance Policies” Australian Health

Review, 29(2),167-177, May

MediSim (static model of the Pharmaceutical Benefits Scheme)Brown. L., Abello, A., Phillips, B. and Harding A., 2004, "Moving towards an improved microsimulation model of the Australian PBS' Australian Economic

Review., 1st quarterAbello, A., Brown, L., Walker, A. and Thurecht, T., 2003, An Economic Forecasting Microsimulation Model of the Australian Pharmaceutical Benefits

Scheme, Technical Paper No. 30, National Centre for Social and Economic Modelling, University of Canberra.*Harding, A., Abello, A., Brown, L., and Phillips, B. 2004 The Distributional Impact of Government Outlays on the Australian Pharmaceutical Benefits

Scheme in 2001-02, Economic Record, Vol 80, Special Issue, September Brown, L., Abello, A. and Harding, A.2006. Pharmaceuticals Benefit Scheme: Effects of the Safety Net. Agenda, vol. 13, no. 3, pp211-224

HEALTHMOD (under development: model of hospital, medical and pharmaceutical sectors)Lymer, s, Brown, Payne, and Harding, 2006, ‘Development of HEALTHMOD: a model of the use and costs of medical services in Australia, 8th Nordic

Seminar on Microsimulation Models, Oslo, June (http://www.ssb.no/english/research_and_analysis/conferences/misi/)