THE POBAL HP DEPRIVATION INDEX AN INTERTEMPORAL AND
THE POBAL HP DEPRIVATION INDEX AN INTER-TEMPORAL AND MULTIJURISDICTIONAL ANALYSIS Trutz Haase & Jonathan Pratschke Stormont, 13 th November 2013
THE 2011 POBAL HP DEPRIVATION INDEX The purpose of the presentation is • to provide an overview of the conceptual components which underlie the HP Deprivation Index • to discuss the treatment of rural deprivation • to provide a practical demonstration of the HP Index • to draw out the Index’ advantages when modelling the social gradient of health outcomes and developing resource allocation models
Conceptual Underpinnings of the Pobal HP Deprivation Index
A COMPREHENSIVE DEFINITION OF POVERTY q Relative Poverty “People are living in poverty if their income and resources (material, cultural and social) are so inadequate as to preclude them from having a standard of living which is regarded as acceptable by Irish society generally. ” (Government of Ireland, NAPS, 1997) q Relative Deprivation “The fundamental implication of the term deprivation is of an absence – of essential or desirable attributes, possessions and opportunities which are considered no more than the minimum by that society. ” (Coombes et al. , Do. E – UK, 1995)
TRADITIONAL APPROACH: EXPLORATORY FACTOR ANALYSIS (EFA) q Ordinary Factor Analysis (EFA) reduces variables to a smaller number of underlying Dimensions or Factors V 1 V 2 F 1 V 3 V 4 V 5 F 2 V 6 q EFA is essentially an exploratory technique; . i. e. data-driven q all variables load on all factors q the structure matrix is the (accidental) outcome of the variables available q EFA cannot be used to compare outcomes over time
NEW APPROACH: CONFIRMATORY FACTOR ANALYSIS (CFA) q Confirmatory Factor Analysis also reduces observations to the underlying Factors, however d 1 V 1 d 2 V 2 d 3 V 3 d 4 V 4 d 5 V 5 d 6 V 6 L 1 L 2 q CFA requires a strong theoretical justification before the model is specified q the researcher decides which of the observed variables are to be associated with which of the latent constructs q variables are conceptualised as the imperfect manifestations of the latent concepts q CFA model allows the comparison of outcomes over time q CFA facilitates the objective evaluation of the quality of the model through fit statistics
THE UNDERLYING DIMENSIONS OF SOCIAL DISADVANTAGE q Demographic Decline (predominantly rural) § population loss and the social and demographic effects of emigration (age dependency, low education of adult population) q Social Class Deprivation (applying in rural and urban areas) § social class composition, education, housing quality q Labour Market Deprivation (predominantly urban) § unemployment, lone parents, low skills base
THE BASIC MODEL OF THE SA-LEVEL POBAL HP DEPRIVATION INDEX d 1 Age Dependency Rate d 2 Population Change d 3 Primary Education only d 4 Third Level Education d 5 Persons per Room d 6 Professional Classes d 7 Semi- and Unskilled Classes d 8 Lone Parents d 9 Male Unemployment Rate d 10 Female Unemployment Rate Demographic Growth Social Class Composition Labour Market Situation
A LONGITUDINAL SA-LEVEL SEM MODEL, 2006 -2011 2006 2011
A LONGITUDINAL ED-LEVEL SEM MODEL, 1991 -2006 Initial Growth Rapid Growth Slow-Down There is only a small correlation between the urban and rural components of the index. This confirms theoretical underpinning of the model which stipulates that urban and rural disadvantage are conceptually different and that the unemployment rate, for example, is not a useful indicator of rural deprivation.
A MULTIPLE GROUP MODEL SPANNING FIVE CENSUS WAVES, 1991 -2011 Multiple Group Model fitted simultaneously across five census waves • imposing identical structure matrix • and identical path coefficients 1991 1996 2002 2006 2011
THE POBAL HP DEPRIVATION INDEX SPANNING 5 CENSUS WAVES, BASED AN ON ED-LEVEL ANALYSIS 06 SA n=18, 488 ED n = 3, 409 91 96 86 91 96 02 06 06 NUTS 4 n = 34 91 96 86 91 96 02 06 06 NUTS 3 n = 8 91 96 86 91 96 02 06 06 NUTS 2 n = 2 91 96 86 91 96 02 06 06 NUTS 1 n = 1 91 96 86 91 96 02 06 06 01 NI 01 NI 06 11 06 11 91 96 02 06 11 91 96 02 06 11 Haase et al. , 1996 Haase, 1999 Pratschke & Haase, 2001 Level at which model is estimated Level to which data is aggregated Pratschke & Haase, 2004 Haase & Pratschke, 2005 Haase & Pratschke, 2008 Haase & Pratschke, 2010 Haase & Pratschke, 2011 Haase & Pratschke, 2012
MAPPING DEPRIVATION most disadvantaged most affluent marginally below the average disadvantaged very disadvantaged extremely disadvantaged marginally above the average affluent very affluent extremely affluent
ED-LEVEL ABSOLUTE INDEX SCORES 1991
ED-LEVEL ABSOLUTE INDEX SCORES 1996
ED-LEVEL ABSOLUTE INDEX SCORES 2002
ED-LEVEL ABSOLUTE INDEX SCORES 2006
ED-LEVEL ABSOLUTE INDEX SCORES 2011
ED-LEVEL RELATIVE INDEX SCORES 1991
ED-LEVEL RELATIVE INDEX SCORES 1996
ED-LEVEL RELATIVE INDEX SCORES 2002
ED-LEVEL RELATIVE INDEX SCORES 2006
ED-LEVEL RELATIVE INDEX SCORES 2011
HP DEPRIVATION SCORES IN COMPARISON, 1991 -2011 HP Deprivation Index N Minimum Maximum Mean Std. Deviation HP 1991 ED absolute 3, 409 -28. 0 73. 3 0. 0 10. 0 HP 1996 ED absolute 3, 409 -27. 4 45. 7 4. 3 9. 2 HP 2002 ED absolute 3, 409 -30. 6 42. 1 8. 4 9. 9 HP 2006 ED absolute 3, 409 -35. 0 39. 9 9. 2 9. 3 HP 2011 ED absolute 3, 409 -43. 7 41. 6 -1. 4 10. 1 HP 1991 ED relative 3, 409 -28. 0 73. 3 0. 0 10. 0 HP 1996 ED relative 3, 409 -34. 4 45. 1 0. 0 10. 0 HP 2002 ED relative 3, 409 -39. 4 34. 0 0. 0 10. 0 HP 2006 ED relative 3, 409 -47. 4 32. 9 0. 0 10. 0 HP 2011 ED relative 3, 409 -41. 9 42. 7 0. 0 10. 0
OVERLAY OF PAIRED RELATIVE INDEX SCORES, 1991 -2006
The Pobal HP Deprivation Measures Small Area (SA) Level Analysis, 2006 - 2011
THE POBAL HP DEPRIVATION INDEX - DUBLIN INNER CITY (ED LEVEL), 2006
THE POBAL HP DEPRIVATION INDEX - DUBLIN INNER CITY (SA LEVEL), 2006
SA-LEVEL ABSOLUTE INDEX SCORES 2006
SA-LEVEL ABSOLUTE INDEX SCORES 2011
SA-LEVEL RELATIVE INDEX SCORES 2006
SA-LEVEL RELATIVE INDEX SCORES 2011
Towards a Deprivation Index Covering Multiple Jurisdictions
METHODOLOGICAL CHALLENGES (OVERVIEW) q Comparability of Spatial Units (COA, SA) q Comparability of Indicator Variables q Temporal Synchronicity (2011 Census) q Common Dimensionality of Deprivation q Common Statistical Model q Standardisation of Index Scores across Multiple Jurisdictions
COMPARABILITY OF INDICATOR VARIABLES q Comparable Indicator Variables q Population Change q Age Dependency q Lone Parents Ratio q Male Unemployment Rate q Female Unemployment Rate q Average Number of Persons per Room q Significantly Varying Indicator Variables q Proportion of Adult Population with Primary Education Only q Proportion of Adult Population with Third Level Education q Proportion of Population in Higher and Lower Professional Classes q Proportion of Population in Semi- and Unskilled Manual Classes
COMMON DIMENSIONALITY OF DEPRIVATION d 3 v 3 Age Dependency Rate d 2 v 2 Population Change d 5 v 5 Primary Education Only d 6 v 6 Third Level Education d 11 v 11 Persons per Room d 7 v 7 Professional Classes d 8 v 8 Semi/Unskilled Classes d 4 v 4 Lone Parents d 9 v 9 Male Unemployment d 10 v 10 Female Unemployment Demographic Decline Social Class Disadvantage Labour Market Deprivation
STANDARDISATION OF INDEX SCORES ACROSS MULTIPLE JURISDICTIONS q Joint standardisation of all factor scores q Simple additive approach to combining factors scores q Resulting in comparable deprivation scores North and South, based on an identical factor structure.
HP DEPRIVATION INDEX FOR NORTHERN IRELAND REPUBLIC OF IRELAND 2001/2006
HP DEPRIVATION INDEX 2001 / 2006
HP DEPRIVATION INDEX 2001 / 2006
STRENGTHS OF CFA-BASED DEPRIVATION INDICES q true multidimensionality, based on theoretical considerations q provides for an appropriate treatment of both urban and rural deprivation q no double-counting q rational approach to indicator selection q uses variety of alternative fit indices to test model adequacy q identical structure matrix and measurement scale across multiple waves q true distances to means are maintained (i. e. measurement, not ranking) q distinguishes between measurement of absolute and relative deprivation q allows for true inter-temporal comparisons q can be developed for multiple jurisdictions
Applications of the Pobal HP Deprivation Index
APPLICATIONS OF THE POBAL HP DEPRIVATION INDEX q Local development § Local Community Development Programme (LCDP), RAPID § Childcare Initiatives, Family Resource Centres, County Development Plans q Health § Mortality Studies, Epidemiological Studies, Primary Health Care, Health Inequality q Education § Educational Disadvantage, Higher Education Access Route q Environment § National Transport Planning, National Spatial Strategy q Statistical Methods and Research Design § Optimising the Sampling Strategy for CSO Household Surveys § Social Equality / Inequality (EU-SILC, QNHS, GUI, TILDA, SLAN, NDS)
HEALTH RISKS AND RELATIVE AFFLUENCE / DEPRIVATION Deprived Affluent Health Risks SD -3 -2 -1 0 1 2 3 0. 1% 2. 1% 13. 6% 34. 1% 13. 6% 2. 1% 0. 1% High Moderate Low
MODELLING POPULATION SHARES ACCORDING TO RELATIVE DEPRIVATION T – TOTAL POPULATION L – LOW (48. 3%) M – MEDIUM (22. 4%) H – HIGH ( 7. 4%) M: -1 STD 22. 4% H: -2 STD 7. 4% L: 0 STD 48. 3% Population T : >5 STD (Total Population)
THE HSE RESOURCE ANALYSER Data Sources Reference Models Model Choices 2011 Census of Population 2011 Pobal HP Deprivation Index Total Population 100% 60% Reference Database for 18, 488 Small Areas Administrative data on current allocations Low Medium High Deprivation 48. 2% 22. 4% 7. 4% 5% 15% 20% Data aggregation to spatial area of interest (Region, ISA, PCT etc. ) Combined Target Allocation
OPTIMISING SAMPLING METHODOLOGIES FOR CSO HOUSEHOLD SURVEYS Comparison of Sampling Designs in the Estimation of Employment (E), Unemployment (UE), Long-term Limiting Illness (LLI) and Education (ED) Model Sample Design Relative Standard Error E UE LLI 3 3 ED Mean Square Error E UE LLI 3 3 3 95% Confidence Interval ED E UE LLI 3 3 ED EU - SILC 2 SCS 1, 300 x 4 3 2 SSCS NUTS 4 x Area 8 1, 300 x 4 3 2 SSCS NUTS 3 x Area 5 x HP Ind 5 1, 300 x 4 2 2 2 SSCS NUTS 3 x HP Index 10 1, 300 x 4 1 1 2 1 3 2 1 2 2 2 1 1 1 3 3 2 1 2 2 2 1 1 3 1 1 2 3 3 QNHS 2 SCS 1, 300 x 20 2 SSCS NUTS 4 x Area 8 1, 300 x 20 3 2 SSCS NUTS 3 x Area 5 x HP Ind 5 1, 300 x 20 2 2 SSCS NUTS 3 x HP Index 10 1 1, 300 x 20 3 3 3 1 1 2 2 1 1 3 3 1 1 2 2 2 1 1 2 2 1 Haase, T. and Pratschke, J. Optimising the Sampling Methodology for CSO Household Surveys, CSO, 2012
SMALL AREA ESTIMATION The BIAS project Imperial College London Small area estimation Nicky Best, Sylvia Richardson, Virgilio Gómez Rubio This work is being carried out in collaboration with ONS. The basic methodological problem is to estimate the value of a given indicator (e, g. income, crime rate, unemployment) for every small area, using data on the indicator from individual-level surveys in a partial sample of areas, plus relevant area-level covariates available for all areas from e. g. census and administrative sources. http: //www. bias-project. org. uk/resdesc. htm#SAE
EVALUATING THE RESOURCE DISTRIBUTION FOR ELDERLY CARE: SMALL AREA ESTIMATION (SAE) HSE Administrative data on current Resource Distribution Use CFA to create Multidimensional Needs Index Survey data: TILDA (n = 8, 000) Combine data using spatial covariates for Small Area Estimation (SAE) SAPS (SA): 2011 Census (n = 18, 499) Use Pobal HP Deprivation Index Combine to Area Level (Region, ISA, PCT) Undertake Gap and Equality Analysis
- Slides: 50