Professional Documents
Culture Documents
of composite indicators
Dr.ssa Mariateresa Ciommi
Dipartimento di Scienze Economiche e Sociali - UNIVPM
Main
references
Part 1:
Introduction
Definition:
A composite indicator is formed when individual indicators are
compiled into a single index, on the basis of an underlying model of the
multi-dimensional concept that is being measured.
Source: OECD, 2004, “The OECD-JRC Handbook on Practices for Developing Composite Indicators”, paper presented
at the OECD Committee on Statistics, 7-8 June 2004, OECD, Paris.
Definition
Definition:
A composite indicator is formed when individual indicators are
compiled into a single index, on the basis of an underlying model of the
multi-dimensional concept that is being measured.
Source: OECD, 2004, “The OECD-JRC Handbook on Practices for Developing Composite Indicators”, paper presented
at the OECD Committee on Statistics, 7-8 June 2004, OECD, Paris.
Definition
Definition:
A composite indicator is formed when individual indicators are
compiled into a single index, on the basis of an underlying model of the
multi-dimensional concept that is being measured.
Source: OECD, 2004, “The OECD-JRC Handbook on Practices for Developing Composite Indicators”, paper presented
at the OECD Committee on Statistics, 7-8 June 2004, OECD, Paris.
Definition
CI are mathematical combinations (or aggregations)
of a set of indicators.
#
𝑥!" : value of the sub-indicator j for
country i at time t, with i=1,…,n and
Definition j=1,…,k
(More # #
𝐼!" : normalised value of 𝑥!"
technical)
𝑤!" : weight associated to sub-indicator j,
j=1,…,k and ∑%"$! 𝑤!" = 1
𝐶𝐼!# : value of the composite indicator for
country i at time t.
Composite indicators provide a single measure for a
not directly observable multidimensional
phenomenon, based on indicators or variables.
Before They are very useful because they can effectively rank
starting: countries, and regions, providing information to both
policymakers and the public since the results are easily
Variable, understood.
Indicator and However, the idea of summarizing complex
Index phenomena into single numbers is not straightforward.
It involves both theoretical and methodological
assumptions that need to be assessed carefully to
avoid producing results of dubious analytic rigour.
Before The construction of a composite indicator can be seen
starting: as an obstacle course, from the availability of data to
the choice of the individual indicators to their
Variable, treatment in order to compare and aggregate them.
Indicator and Therefore, criticisms could grow along with each step
Index of the proces
The term “variable” derives from the late Latin
variabilis, meaning “that varies, which tends to vary”.
In statistics, we mean a characteristic found in one or
Before more statistical units belonging to a population or a
sample as a result of a direct survey.
starting: Nowadays this definition seems almost reductive since
Variable, we live in the era of the deluge in which we are
Indicator and surrounded by big data or administrative sources used
for statistical purposes and, therefore, not only by
Index direct surveys.
In more general terms, the variable is defined as a
characteristic associated with a statistical unit such
as income, age or beds in a hospital.
The term “indicator” comes from the late Latin
indicator, meaning “who indicates” or “element that
indicates or signals something”.
In the economic field, we indicate some
Before macroeconomic quantities (employment rate, inflation
starting: rate, GDP per capita, etc.) that are considered
significant for the purpose of evaluating the
Variable, performance of the economy in a given country and in
Indicator and a given time.
While any normalized variable can be called an “indicator”,
Index we usually mean that the indicator is a ratio and that,
therefore, it is composed of a numerator, which can be
considered the variable that provides the meaning, and a
denominator, which is the variable that allows its
comparability in space and/or over time.
Let’s take the example of wanting to compare the
health-care infrastructure of two European countries,
Germany and the Netherlands:
if we considered the variable “number of hospital beds” it
Before would make no sense to make the comparison because,
obviously, Germany will have more sleeping places than
starting: Holland.
If, on the other hand, we calculate the indicator “number of
Variable, beds per 1,000 inhabitants”, we obtain a normalized value
according to the population variable and then we can make a
Indicator and comparison between the two countries.
Step 3.
Imputation of
missing data
Hot deck imputation. Filling in blanks cells with
individual data, drawn from “similar” responding units.
For example, missing values for individual income may be
replaced with the income of another respondent with similar
Imputation: characteristics, e.g.age, sex, race, place of residence, family
relationships, job, etc.
Implicit
Substitution. Replacing non-responding units with
modelling. unselected units in the sample.
For example, if a household cannot be contacted, then a
previously non-selected household in the same housing block is
selected.
Cold deck imputation. Replacing the missing value with
a value from an external source,
e.g. from a previous realisation of the same survey.
Unconditional mean/median/mode imputation. The sample mean (median,
mode) of the recorded values for the given individual indicator replaces the
missing values.
Regression imputation. Missing values are substituted by the
predicted values obtained from regression.
The dependent variable of the regression is the individual indicator
hosting the missing value, and the regressor(s) is (are) the individual
indicator(s), showing a strong relationship with the dependent variable,
Imputation: i.e. usually a high degree of correlation.
Step 4.
Multivariate
Analysis
Analysing the underlying structure of the data is still an art
Step 4.
Multivariate
Analysis
Avoid adding up apples and oranges
Step 5.
Normalization
“by equations”
Avoid adding up apples and oranges
Step 5.
Normalization
“by equations”
Step 5.
Normalization
“by equations”
Step 5.
Normalization
“by equations”
Step 5.
Normalization
“by equations”
Step 5.
Normalization
“by equations”
Step 5.
Normalization
“by equations”
Comparison
between
normalization
methods
Comparison
between
normalization
methods
Comparison
between
normalization
methods
The polarity of an indicator is the sign of the
relationship between the indicator and the
phenomenon to be measured.
For example, in the calculation of a development index,
GDP has positive polarity (+), while the infant mortality
rate has negative (-).
Similarly, in the calculation of a poverty index, GDP
negative (-), while the infant mortality rate has positive
Polarity polarity (+).
In order to aggregate correctly a set of indicators,
it is necessary that all have positive polarity.
The main methods of reversing the polarity is
divided into:
Linear transformation
Non linear transformation
It's the easiest way to reverse the polarity of an
indicator and it is based on subtracting from the
maximum value the value of each unit:
#
Methods to 𝑥!" = max{𝑥!" } − 𝑥!"
!
WEIGHTING METHODS
1. Weights based on principal components analysis or factor
analysis
2. Data envelopment analysis (DEA)
DEA employs linear frontier that would be used as a
benchmark to measure the relative performance of a
Step 6. country. This requires construction of a benchmark (the
frontier) and the measurement of the distance between
Weighting and countries in a multi-dimensional framework.
Aggregation 3. Benefit of the doubt approach (BOD)
The application of DEA to the field of composite
indicators is known as the “benefit of doubt approach“
(BOD) and was originally proposed to evaluate
macroenomic performance. In the BOD approach, CI is
defined as the ratio of a country’s actual performance to
its benchmark performance:
The relative importance of the indicators is a source of contention
Some
Aggregation
Methods
The relative importance of the indicators is a source of contention
Some
Aggregation
Methods
The relative importance of the indicators is a source of contention
Some
Aggregation
Methods
The relative importance of the indicators is a source of contention
Some
Aggregation
Methods:
Example
In general, we have: M0 £ M1 £ M2
The relative importance of the indicators is a source of contention
!
Power means (of r ordine): 𝑀!# = (∑& #
"$% 𝑦!" 𝑤" )
" where 𝑤" is the weight
(0 < 𝑤" < 1) e ∑&
"$% 𝑤" = 1
PROS:
Semplicity of calculation;
Easy interpretation.
The relative importance of the indicators is a source of contention
A) Normalization
Each individual indicators is transformed in a z-score with
mean equal to 100 and standard error equal to 10 (mean=100
Some and s.e.=10); the normalized values will be in the range 70-
130.
Aggregation
This procedure allows to depurate the indicators both from
Methods: the unit of measure and from the variability and it does not
require the deifnition of a target values (ideal unit), because it
The Mazziotta- replaces the vector with the of average values.
Pareto Method In this way, it is easy to identify the units with a level of the
phenomenon above average (values greater than 100) and
the units with a level below the average (values less than
100).
Some
Aggregation
Methods:
The Mazziotta-
Pareto Method
The relative importance of the indicators is a source of contention
B) Penalization
The aggregation function (arithmetic mean of standardized
Some values) is 'corrected' by a penalty coefficient that depends, for
Aggregation each unit, on the variability of the indicators compared to the
average value ('horizontal variability')
Methods: This variability, measured by the coefficient of variation (CV),
allows to penalize the score of the units that, with the same
The Mazziotta- arithmetic average, have a higher imbalance between the
Pareto Method values of the indicators.
The penalty can be added or subtracted, depending on the
type of the studied phenomenon (poverty, development, etc.).
The relative importance of the indicators is a source of contention
Some
Aggregation
Methods:
The Mazziotta-
Pareto Method
The relative importance of the indicators is a source of contention
Some
Aggregation
Methods:
The Mazziotta-
Pareto Method
Some
Aggregation
Methods:
The Mazziotta-
Pareto Method
• It is variant of MPI, based on a transformation Min-Max instead
in waste z-score
• The transformation Min-Max is based on two goalposts: a
minimum and a maximum that represent the possible range of
Some variation of each indicator throughout the period considered and
all units
Aggregation
Methods: AMPI vs MPI
The Adjusted - MPI the value 100 represents the mean of the values of all units
- AMPI the value 100 represents the mean of the goalposts.
Mazziotta- You can 'fix' the goalposts in order to put equal to a 100 = reference
Pareto Method value (eg., the national average in a given year).
Step 6.
Weighting and
Aggregation
Sensitivity analysis can be used to assess the robustness of composite indicators
Step 7.
Uncertainty
and Sensibility
analysis:
De-constructing composite indicators can help extend the analysis
Step 8.
Back to the
data
Step 9.
Link to
other
indicators
A well-designed graph can speak louder than words
Step 10.
Visualization
of the results
The European
Statistics Code of
Practice
(Principles 11-15)
focuses on
Part 3: statistical outputs
as viewed by
more on CI users. Six quality
dimensions are
considered:
1. Relevance
2. Accurancy (credibility, objectivity)
3. Timeliness
4. Accessibility” and Clarity”
Six quality 5. Comparability
dimensions: 6. Coherence
1. Relevance In the context of composite indicators, relevance
refers to the has to be evaluated considering the overall
purpose of the indicator.
degree to Careful evaluation and selection of basic data have
which statistics to be carried out to ensure that the right range of
meet current domains is covered in a balanced way.
and potential Given the actual availability of data, ”proxy” series
are often used, but in this case some evidence of
needs of the their relationships with “target” series should be
users; produced whenever possible.
In the context of composite indicators, accuracy of basic data is extremely
important.
2. Accuracy
refers to the The credibility of data products refers to confidence that users place in
those products based simply on their image of the data producer, i.e.,
closeness of the brand image.
computations One important aspect is trust in the objectivity of the data. This implies
the exact or with appropriate statistical standards and policies and that practices are
transparent (for example, data are not manipulated, nor their release
true values; timed in response to political pressure).
Other things being equal, data produced by “official sources” (e.g. national
statistical offices or other public bodies working under national statistical
regulations or codes of conduct) should be preferred to other sources.
3. Timeliness In the context of composite indicators, timeliness is especially
refers to the length of important to minimise the need for the estimation of missing data or
time between the for revisions of previously published data.
availability of the
information and the As individual basic data sources establish their optimal trade-off
event or phenomenon between accuracy and timeliness, taking into account institutional,
it describes. organisational and resource constraints, data covering different
domains are often released at different points of time. Therefore
“Punctuality”
refers to the time lag special attention must be paid to the overall coherence of the
between the target vintages of data used to build composite indicators (see also
delivery date and the coherence).
actual date of the
release of the data;
In the context of composite indicators, accessibility of basic
4.“Accessibility” refers to the data can affect the overall cost of production and updating of
physical conditions in which
users can access statistics: the indicator over time.
distribution channels, ordering
procedures, time required for It can also influence the credibility of the composite indicator
delivery, pricing policy,
marketing conditions (copyright, if poor accessibility of basic data makes it difficult for third
etc.), availability of micro or
macro data, media (paper, CD-
parties to replicate the results of the composite indicators.
ROM, Internet, etc). In this respect, given improvements in electronic access to
“Clarity” refers to the statistics’
information environment: databases released by various sources, the issue of coherence
appropriate metadata provided across data sets can become relevant.
with the statistics (textual
information, explanations,
documentation, etc); graphs, Therefore, the selection of the source should not always give
maps, and other illustrations;
availability of information on preference to the most accessible source, but should also
the statistics’ quality (possible take other quality dimensions into account.
limitation in use);
In the context of composite indicators, the wide range of data used
5. Comparability to build them and the difficulties due to the aggregation procedure
refers to the require the full interpretability of basic data.
measurement of the
impact of differences in The availability of definitions and classifications used to produce
applied statistical basic data is essential to assess the comparability of data over time
concepts and
and across countries (see coherence):
measurement tools and
procedures when for example, series breaks need to be assessed when composite
statistics are compared indicators are built to compare performances over time.
between geographical
areas, non-geographical Therefore the availability of adequate metadata is an important
domains or over time; element in the assessment of the overall quality of basic data.
In the context of composite indicators, two aspects of
6. Coherence coherence are especially important: coherence over
refers to the time and across countries.
adequacy of the Coherence over time implies that the data are based
data to be on common concepts, definitions and methodology
reliably over time, or that any differences are explained and
can be allowed for. Incoherence over time refers to
combined in breaks in a series resulting from changes in concepts,
different ways definitions, or methodology.
and for various Coherence across countries implies that from country
uses. to country the data are based on common concepts,
definitions, classifications and methodology, or that
any differences are explained and can be allowed for.
The main
characteristics
of the
measurement
scales.
Part 4:
Example of CI
The BES Project & SDG
Environmental Sustainability Index (WEF)
Air Quality Index (WEF)
Environment Index (World Travel and Tourism Council)
Environmental Performance Index (WEF, Yale &
Columbia Universities)
Environment Living Planet Index (UNEP & WCMC)
National Biodiversity Index (Secretariat of the
Convention on Biological Diversity)
Natural Capital Index (RIVM, The Netherlands)
Human Development Index (United Nations)
Health System Achievement Index (WHO)
Corruption Perceptions Index (Transparency
International)
World Income Inequality Database:Gini Index (United
Society Nations)
Wellbeing Index (Prescott-Allen)
Genuine Progress Indicator (Redefining Progress)
Economic Sentiment Indicator (EC)
Composite Leading Indicators (OECD)
Internal Market Index (EC)
Doing Business Indicators (World Bank)
Index of Economic Freedom (Heritage Foundation)
Economy
Economic Competitiveness Index (Institute for
Management Development)
Human Tourism Index (World Travel and Tourism
Council)
Summary Innovation Index (EC)
Innovative Capacity Index (Porter and Stern)
Investment/Performance in the knowledge based
Innovation economy (EC)
Technology Technology Achievement Index (United Nations)
Information The Networked Readiness Index (Harvard University -
Centre for International Development)
E-Government Rankings (World Markets Research
Centre)
Globalization Index (Foreign Policy Magazine)
World Competitiveness Index (IMD)
Growth Competitiveness Index (WEF)
Globalization Current Competitiveness Index (WEF)
The Globalisation Index (G-Index) (World Markets Research
Centre)
Beyond GDP
“The gross national product includes air pollution
and advertising for cigarettes and ambulances to
clear our highways of carnage. It counts special
locks for our doors and jails for the people who
break them. GNP includes the destruction of the
redwoods and the death of Lake Superior. It grows
with the production of napalm, and missiles and
nuclear warheads... it does not allow for the health
Beyond GDP of our families, the quality of their education, or the
joy of their play. It is indifferent to the decency of our
factories and the safety of our streets alike. It does
not include the beauty of our poetry or the strength
of our marriages, or the intelligence of our public
debate or the integrity of our public officials. It
measures everything, in short, except that which
makes life worthwhile.”
Corrective to GDP Initiative aiming at meauring the well-being
Measure of Sustainable Welfare - MEW Nordhaus e Tobin, 1971
2004 - OECD World Forum on Key Indicators.
Index of Sustainable Economic Well-being – ISEW Daly e Cobb, 1991
Statistics, Knowledge and Policy (Palermo).
2006 - Istanbul Declaretion+ Global Project on
Synthetic indices and rankings
measuring the progress of societies.
Relative National Standards of Living Bennett, 1937
2007 - WWF, Club of Rome, EP – Beyond GDP
Level of Living Index UNRISD, 1966
Human Development Index UNDP, 1991 2008 - Sen-Stiglitz-Fitoussi Commission
Indice di Qualità della Vita delle Provincie Sole 24 Ore, 1993 2009 - European Commission – “GDP and
Indice di Qualità Regionale dello Sviluppo - QUARS Sbilanciamoci!, 2006 Beyond: Measuring progress in a changing world”.
Happy Planet Index NEF, 2006 2011 - OECD Better Life Initiative. “How’s Life”
Better Life Index OECD, 2011 2011 - Eurostat –Sponsorship Group on
Measuring Progress, Well-being and Sustainable
Set of indicators Development
Report of the President's Research Committee on Social Trends USA, 1933
Millennium Development Goals ONU, 2000
Sustainable Development Indicators EC “What we measure affects
National Accounts of Wellbeing NEF, 2009 what we do”
The 12
dimensions of
well-being
The context
Benessere Equo e Sostenibile
Equitable and Sustainable Well-being
For Italy:
BES
Multidimensional aspects of
quality of life, including health Guarantee for next
generations
penalization
A possible
solution:
vertical
weights
In this paper, we have introduced a new class of composite indicators for
measuring the well-being at the local level, which takes into account the
variability between and within the local units.
A new class of This class has two extreme special cases: on the one hand, the Adjusted
Mazziotta Pareto Index (AMPI), which takes into account the unbalanced
CI for well- distribution among the indicators belonging to the same well-being
domain, and, on the other hand, the Gini based weighted average (GW
being and RGW), which depends on the distribution of each indicator across the
local units. Combinations of them are also possible (GAMPI and RGAMPI).
M1 - (EW)
1a) AMPI
'2
1 )
𝐼!$ = ' · 5 𝑟!"$ · 𝐺."$ − 𝛽 · 𝑆*342 · 𝑐𝑣*342
∑"%&
2
(𝐺 ."$ ))
"%&
A class of CI
Method α 𝛃 Equation
6%
1
The equally-weighted average 0 0 𝐸𝑊!5 = < 𝑟!"5 >
𝑛5
"$%