Constructing composite indices Amie Gaye, Policy Specialist UNDP/HDRO

1
Constructing composite indices Amie Gaye, Policy
SpecialistUNDP/HDRO
HDRO/RBA Regional Technical Workshop on Measuring
Human Development Nairobi September - 2007
2
Structure of presentation

• Why composite indices
• Characteristics of a good composite indicator
• Key steps in constructing composite indices
• Country examples (adapting the HDI new index
• Conclusion

3
What is a composite index/ indicator (CI)?

• a value derived from a combination of various
  indicators based on a theoretical model of a
  multi-dimensional concept that is being measured
• Not an end in themselves

4
Why use composite indices

• When well constructed
• Ability to summarize multi-dimensional issues to
  support policy decisions
• Easier to interpret than finding trend in
  separate indicators
• Ease communication with the general public

5
Characteristics of a good CI

• Conceptually clear
• Policy relevant
• Measurable components
• Methodologically simple and transparent
• Easy to interpret

6
Key steps in constructing a composite index

• Theoretical model- what is poorly defined is
  poorly measured
• Selecting indicators
• A composite index is the sum of its parts
• quality depends largely on quality of underlying
  indicators
• Analysing the component indicators
• Treatment of missing data- these can bias results
• Case deletion
• Imputation

7
Steps in constructing a composite index

• Normalising Avoid adding apples and pears
• Re-scaling (HD approach, 01)
• Setting goalposts national specific context
  should inform decisions
• What is the current situation?
• Where does the country want to go?
• How feasible is the goal based on current trends?

8
Steps in constructing a composite index

• Long healthy life
• Knowledge
• A decent standard of living
• Income maximum
• average annual growth rate of 2 in per capita
  income of the highest group in the country for
  the next 50 years
• Minimum could be set at 30 or 50 of mean income
  per capita (accounting for sub-group difference)
• Use of PPP not necessary

9
Steps in constructing a composite index

• Weighting aggregation
• Weighting is based on value judgment
• Equal weighting implies equal importance (HDI)
• Geometric (power) averaging places greater weight
  on components with higher values (HPI)
• Choice should be consistent with theoretical
  model
• Document weighting procedures

10
Steps in constructing a composite index

• Testing robustness of the CI
• Uncertainty imposes a limit on the confidence of
  an index.
• Sensitivity and uncertainty analysis
• assessment of uncertainties associated with the
  index
• Document and show results of analysis

11
Steps in constructing a composite index

• Links to other variables
• Good practice - examine correlation between other
  variables
• measure explanatory power of the index high
  correlation suggests high quality CI.
• 9. Deconstructing
• Determine contribution of each component
  indicator to the index

12
Some country examples

• The Gambia 1997 School life expectancy
• Costa Rica 2005 security index for each of the
  80 cantons (cantonal security index)
• and adjusts each regions HDI with the security
  dimension
• Thailand 2003 constructs a new index human
  achievement index
• Seven dimensions
• 40 indicators

13
Things to be aware of

• Poorly constructed CI sends misleading policy
  messages
• CI may lead politicians to draw simplistic policy
  conclusions analyse the components making up
  the index bring out which dimension needs
  improvement most
• Scope for disagreement among different groups-
  countries, local government officials, etc.

14
Limitations of the HDI

• Only a basic measure of HD
• Other important dimensions are not measured
• Not useful for monitoring impact of short term
  policy changes
• Mixture of stock and flow variables partially
  measures outcomes of past efforts Adult literacy
  rate is a stock variable, GDPpc is flow

15
Conclusion

• Involve people with expertise
• Keep it simple- too many dimensions, too
  difficult to interpret
• Lack of sensitivity analysis undermines
  confidence in CI so be rigorous
• CHECK for ROBUSTNESS!
• No matter the scientific basis, CI acceptance
  relies on negotiation and peer acceptance.
• Be inclusive
• Subject your work to peer review!

16
OECD Handbook on constructing composite
indicators http//www.olis.oecd.org/olis/ Visit
the EC web site on composite indicators http//co
mposite-indicators.jrc.ec.europa.eu/