Presentation is loading. Please wait.

Presentation is loading. Please wait.

Indices of Deprivation Measuring change between ID2004 and ID2007 Kate Wilkinson University of Oxford.

Similar presentations


Presentation on theme: "Indices of Deprivation Measuring change between ID2004 and ID2007 Kate Wilkinson University of Oxford."— Presentation transcript:

1 Indices of Deprivation Measuring change between ID2004 and ID2007 Kate Wilkinson University of Oxford

2 Outline of presentation  How are the Indices constructed?  What changes have there been in indicators / methodology between 2004 and 2007?  What impact do the changes have on the comparisons that can be made?  How should the data be used?  What other consistent measures of change are available?

3 Making a Domain (1) Indicator 1Indicator 2Indicator 1 Shrinkage Rank indicator scores and normalise Indicator 2 Sub-domain1Sub-domain2 Combine indicators (with weights) Rank sub-domain scores and convert to exponential distribution Combine sub-domains (with weights) = DOMAIN SCORE

4 What does shrinkage do?  In some small areas, indicators may be ‘unreliable’, particularly where populations at risk are small.  Level of ‘unreliability’ measured by calculating standard error.  Shrinkage adjusts small area indicator scores by ‘borrowing strength’ from a more robust score.  District level average score used as ‘more robust’ figure.  All small area scores move, but only those with a large standard error move substantially.

5 Example of shrinkage from the employment domain

6 Making a domain (2)  Methodology differs slightly for each domain  For example:  HEALTH – No sub-domains, indicators shrunk, ranked and normalised then combined with weights to produce domain score  EMPLOYMENT – No sub-domains, indicators represent a single measure (rate of working age individuals involuntarily excluded from the labour market), single indicator is shrunk to create a domain score  EDUCATION – Two sub-domains, indicators shrunk, ranked and normalised, indicators combined to produce sub-domain scores, sub-domain scores ranked and exponentially transformed then combined to produce the domain score

7 Making an Index of Multiple Deprivation  Each domain constructed with slightly differing methodology, so domain scores not comparable between domains  To make an index domain scores must be standardised  Exponential distribution transforms domain scores to the same scale and ‘spreads out’ the 10% most deprived so cancellation effects are minimised where areas experience severe deprivation on any domain Domain 1Domain 2Domain 3 Rank domain scores and convert to exponential distribution Combine with theoretically derived weights = IMD SCORE

8 What has changed – 2004-2007  Underlying methodology identical  Weights for combining domains unchanged  Weights for combining indicators may change (slightly) if derived using factor analysis  Data used to create underlying indicators and / or methods used to create underlying indicators has changed in some cases

9 Summary of changes by domain (1) INCOME – Definition of income deprivation identical “comprehensive, non-overlapping counts of both in- work and out-of-work means-tested benefits”. Adjustments made due to changes in structure of benefit and tax credits (PC, WTC and CTC). EMPLOYMENT – Method of creating unemployment claimant counts improved to completely remove possibility of double counting. Very little effect on actual numbers.

10 Summary of changes by domain (2) HEALTH – No changes. Factor analysis is used to create weight to combine indicators so these will differ slightly between 2004 and 2007. EDUCATION – Methodology for creating Key Stage indicators improved as better data now available. Reduces the number of tied ranks. Skills sub-domain identical in 2004 and 2007 (not possible to update). Factor analysis used to create weights to combine indicators so these will differ slightly.

11 Summary of changes by domain (3) BARRIERS – No changes. CRIME – No changes. LIVING ENVIRONMENT – No changes. Central heating indicator identical. OTHER FACTORS – 2001 population estimates have been revised since 2004. Impact likely to be very small.

12 Measuring change  IMD scores and domain scores are not stable across time and should not be compared BUT…  Ranks of scores can be compared for domains and IMD  Changes in underlying indicators between 2004 and 2007 are small – unlikely that these would result in significant change in the rank of a domain score or final IMD score  All change is relative – an area may increase or decrease rank without any actual change occurring  The smaller the area of analysis the more likely that change will be observed i.e. comparing national rankings with regional rankings

13 Average district score 2004 and 2007

14 IMD score for the South East 2004 and 2007

15 Unpicking reasons for change  Change in the rank of an IMD score for an area may occur for many reasons:  Methodological changes between indices (unlikely but possible)  Actual change in level of deprivation in that area compared to other areas  No change in level of deprivation in the area but relative position changes  IMD ranks and domain ranks do not tell us about absolute change only relative change

16 Unpicking reasons for change  Uncovering the reasons behind the change is not always straight forward  Look at change of ranks on individual domains  Look at change in underlying indicators – shrunk indicator data published on CLG website (and for previous years some indicators available on http://www.neighbourhood.statistics.gov.uk/).  Again best to look at change in ranks as indicator scores may not be comparable if methodology has changed.  Change in individual indicators may point to absolute change IF the indicators are identical in methodology and underlying data, as we have seen this is not always the case.  Local knowledge is key to understanding change

17 Other useful measures  Economic Deprivation Index (EDI) and Children’s Economic Deprivation Index (CIDI) soon to be published  EDI contains an Income and Employment domain:  Income Deprivation Domain – proportion of people under 60 living in households receiving one of two out-of-work means-tested benefits: Income Support or income-based Job Seeker’s Allowance  Employment Deprivation Domain – proportion of people of working age claiming one of three out-of-work benefits: Job Seeker’s Allowance (income or contributions based), Severe Disablement Allowance or Incapacity Benefit  CIDI measures children living in households in receipt of the benefits described above  Domains shrunk, exponentially transformed and combined with equal weights

18 EDI and CIDI  EDI and CIDI use consistent series of population estimates  Methodology identical in each year so changes in absolute and relative deprivation can be accurately measured  EDI / CIDI produced for every year 1999-2005 and will continue to be updated each year  Correlates 0.95 with IMD – good proxy for multiple deprivation

19 Example: Change in absolute rates of income deprivation

20 Example: Change in relative rates of income deprivation

21 Questions and Discussion


Download ppt "Indices of Deprivation Measuring change between ID2004 and ID2007 Kate Wilkinson University of Oxford."

Similar presentations


Ads by Google