The study aims at providing evidence on regional differences in the diffusion of ICT in the public sector in Italy, with a focus on different types of public e-services (eGovernment, eHealth, eEducation and Intelligent Transport Systems). Data are ob-tained by merging four different surveys carried out by Between Co. (2010-11) and Istat - Italy’s National Bureau of Statistics (2009). We pursue a three-fold objective. First, we attempt to overcome the prevailing attitude to consider the various domains of public e-service provision as separate from one another. In other words, measuring the progress of digital government requires a holistic view to capture the wide spectrum of public e-services in different domains (e.g. local and national administrative procedures, transportation, education, etc.) and the different aspects of service provision (not just e-readiness or web interactivity, but also multi-channel availability and take-up). Second, we shall tackle a major drawback of existing statistics and benchmarking studies of public e-services, which are largely based on the count of services provided online, by including more sophisticated indicators both on quality of services offered and back office changes. Third, we develop a sound, open and transparent methodology for constructing a public eServices composite indicator based on OECD/EC-JRC Handbook. This methodology, which incorporates experts opinion into a Data Envelopment Analysis, will allow us to combine data on different e-service categories and on different aspects of their development, and will enable us to define a ranking of Italian regions in terms of ICT adoption and public e-service development.
How advanced are Italian regions in terms of public eServices
1. 1st International EIBURS-TAIPS TAIPS conference on:
“Innovation in the public sector
and the development of e-services”
How advanced are Italian regions in terms
of public e-services?
The construction of a composite indicator to analyze patterns
of innovation in the public sector
Luigi Reggi, Davide Arduini, Marco Biagetti and Antonello Zanfei
EIBURS-TAIPS team, University of Urbino
University of Urbino
April 19-20, 2012
2. Aims and scope
• Providing evidence on regional differences in the
diffusion of public eServices in Italy with a focus on
– different types of public eServices: beyond a
monodimensional analysis based on e-gov
diffusion
– not only front- but also back-end issues
– different channels for service delivery
• Providing a sound, open and transparent
methodology for constructing a public eServices
composite indicator based on OECD/EC-JRC
Handbook
3. Composite indicators (CI)
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 (OECD Glossary of statistical terms)
• Composite indicators are increasingly used by
statistical offices, international organizations (e.g.
OECD, EU, WEF, IMF) and academic researchers to
convey information on the status of countries in
fields such as the environment, economy, society
or technological development: Cox et al., 1992;
Cribari-Neto et al., 1999; Griliches, 1990; Huggins
2003; Grupp and Mogee 2004; Munda 2005;
Wilson and Jones 2002; among others
• The proliferation of these indicators is a clear
symptom of their importance in policy-making,
and operational relevance in macro and micro Searching “Composite indicator” in
Google Scholar => 5x increase in 6 years
economics in general (Granger, 2001)
(Saltelli, 2011)
4. Pros and cons of CI
Pros Cons
Can summarize complex or multi-dimensional May send misleading policy messages if they
issues in view of supporting decision-makers. are poorly constructed or misinterpreted.
Easier to interpret than trying to find a trend in May invite simplistic policy conclusions.
many separate indicators. May be misused, e.g., to support a desired
Facilitate the task of ranking countries on policy, if the construction process is not
complex issues in a benchmarking exercise. transparent and lacks sound statistical or
Can assess progress of countries over time on conceptual principles.
complex issues. The selection of indicators and weights could
Reduce the size of a set of indicators or include be the target of political challenge.
more information within the existing size limit. May disguise serious failings in some
Place issues of country performance and dimensions and increase the difficulty of
progress at the centre of the policy arena. identifying proper remedial action.
Facilitate communication with general public May lead to inappropriate policies if
(i.e. citizens, media, etc.) and promote dimensions of performance that are difficult to
accountability. measure are ignored.
(Saisana and Tarantola, 2002; OECD, 2008)
5. Selected CIs in public e-services field (1/2)
Time Number of
Composite Aggregation
Field/Source coverag countries Sub-indicators
Indicator methodology
e covered
e-Government 191 Web presence
2001- Equal
United Nations Readiness Member Telecommunication infrastructure
2010 weighting
Index States Human capital
eGovernment
198
Brown e-government 2001 - availability of publications, databases and number of on line Equal
Member
University index 2007 services weighting
States
Online sophistication of the 20 basic services (4 stage maturity
European
e-government 2001 - 32 European model: information available on-line, one-way interaction, Equal
Commission /
index 2010 Countries two-way interaction and transaction) weighting
CapGemini
Full online availability of the 20 basic services
Service Maturity Breadth (number of services offered through
33 EU the Internet from the 67 identified services)
Torres et al. Service Equal
2004 municipaliti Service Maturity Depth (3 stage maturity model: simple
(2005) maturity Index weighting
es information dissemination, one way communication, service
and financial transactions)
e-Government
95 Member Equal
Kovačić (2005) Readiness 2003 Based on United Nations data and methodology
States weighting
Index
Baldersheim et 2004 75 Nordic Information features of the web sites (refers to the contents
al. (2008) municipaliti of communication channels between citizens and town hall)
Innovation Equal
es Communication features of the web sites (refers to the extent
score weighting
of interactivity of web sites, or how citizens can actually
communicate via municipal sites)
1,176 Italian Multiple
Arduini et al. Front Office
2006 municipaliti Availability and level of interactiveness of 266 on line services Correspondenc
(2010) Index
es e Analysis
6. Selected CIs in public e-services field (2/2)
Number of
Time Composite Aggregation
Field/Source countries Sub-indicators
coverage Indicator methodology
covered
eProcurement
eProcurement
- eNotification, eSubmisssion and eAwards services availability for
provided by eProcurement platforms in the public the pre -
European 32 European sector award phase Equal
2010
Commission Countries - eOrdering, eInvoicing and ePayment services eProcurement weighting
provided by eProcurement platforms in the public availability for
sector the post -
award phase
European
Commission – 906 acute - Infrastructure Dimension Composite
eHealth
Multivariate
Joint Hospitals in the - Application and Integration Dimension index of
2010 Statistical
Research 27 European - Information flows dimension eHealth
Analysis
Centre Countries - Security and privacy dimension deployment
(Seville)
2 County
eTransportation
Metropolitan
- Real-time network information Advanced
Transportation
Horan et al. - Whether traffic or transit Travel Equal
2006 Authorities
(2007) - Traveler information such as route guidance or Information weighting
(Los Angeles
destination information Systems Index
and
Minneapolis)
7. CIs in public e-services field
Existing CIs in public eServices field
• are specific to a single domain / type of eService
• employ simple equal weighting as standard aggregation
method (with a few exceptions)
• do not assess results with Uncertainty or Sensitivity
Analysis (UA – SA)
Critical remarks have been raised against EC
eGovernment bechmarking index. Criticism is mainly
focused on theoretical framework, indicators chosen,
aggregation scheme adopted (Bannister, 2007; Bretschneider et al,
2005; Fariselli & Bojic 2004; Goldkuhl & Persson, 2006; Jansen, 2005)
8. What is new in our methodology
for a Public eServices CI
1. Expanding the scope of the analysis of eServices diffusion
– A holistic view to capture the wide spectrum of public e-services in different domains
(in our case: eGov, eEducation,eTransportation) and the different aspects of service
provision (e.g. technical and organizational change within PAs and new service
implementation)
2. Improving the quality of the framework
– Using more sophisticated indicators both on quality of services offered and back
office changes
– Robustness check of the framework / classification of indicators
3. Developing a sound, open and transparent methodology
– Asking experts to assess the importance of basic indicators
– Real benchmarking: measuring the distance from the efficiency frontier
– Tracing back the contribution of the different aspects of eService diffusion (e.g. back-
and front-end issues) to intermediate and final indices
– Checking the robustness of results by reiterating the calculation of the CI with 12
other different methods (Uncertainty Analysis)
9. Public e-Services diffusion:
a broad definition
Aims Dimensions of ICT diffusion
Efficiency and effectiveness of public Service provision - front end
service (Fountain, 2001; Codagnone e Undheim, Internal processes / interoperability /
2009) information integration - back end (Millard,
Transparency (Wong & Welch, 2004; Meyer, 2004; Pardo and Tayi, 2007; OECD, 2007)
2009, Dawes 2010) Decision- / policy –making
Participation (Noveck, 2008) (Lampathaki et al., 2010)
Providers Channels
Government: central / local agencies, Institutional websites, public websites
public companies Public kiosks
Third party players - PPPs, apps Digital TV
development (Brito, 2009; Eaves, 2010) Mobile apps
NGOs, citizens - self-help, collaboration (Pieterson et al., 2008)
(Noveck, 2008)
Data sources Domains
Main focus of
eGovernment existing CIs /
Government eEducation benchmarking
Citizens / NGOs / businesses: eTransportation exercises
crowdsourcing (Osimo, 2008; Robinson et al., eHealth Scope of our
2009; Chun et al., 2010) analysis
Smart cities
10. Public eServices CI
- our framework -
• Existing theoretical frameworks are mainly focused on
eGovernment and based on stage models implying linear
progression (Lee, 2010)
[i.e. from stage 1 = input/eReadiness to stage n = outcome]
– academic papers (Andersen & Henriksen, 2006; Hiller & Belanger, 2001; Layne & Lee,
2001; Moon, 2002; Siau & Long, 2005; Scott, 2001; West, 2004)
– institutional reports (Center for Democracy & Technology, 2002; Grant & Chau, 2005;
United Nations, 2001, 2003, 2005, 2008)
– private consulting firms reports (Accenture, 2003; Deloitte Research, 2000)
• Most available frameworks can hardly be applied to the
construction of our CI
“Too often composite indicators include both input and output
measures. […] However, only the latter set of output indicators should
be included if the index is intended to measure innovation
performance”
(OECD/EC-JRC Handbook on Constructing CIs, p.6)
11. Public eServices CI
- our framework -
PILLAR
Public eServices Composite Indicator
₋ Mobility
₋ Intranet monitoring
₋ Certified e-mail systems
₋ Interoperability
₋ eProcurement ₋ Interoperability
& integration
₋ Document & integration
workflow
₋ School website ₋ Travel planner
₋ Restricted areas ₋ Info on traffic
for information and parking
services ₋ Fully interactive ₋ Multi-channel
service provision delivery
₋ Interactive
₋ On line payments ₋ Technology on
whiteboards
didactics ₋ Multi-channel board of public
₋ Repositories of
delivery transport
documents
₋ Wiki platforms ₋ electronic displays
on the street
SUB-PILLAR INDICATORS
12. Data sources
Domain Statistical units Source
eEducation 1,600 schools Between. Survey “Service e-
Platforms”, 2010
eGovernment 5,762 municipalities, 100 Italian Institute of Statistics. Survey
Provincial governments “Information and Communication
and 22 Regional Technologies in Local Public
governments Administrations”, 2009
eTransportation 117 local public transport Between. Survey “Service e-
companies Platforms”, 2011
Valle d’Aosta and Molise (0,7% of total
Italian population) were excluded from
the analysis due to poor data quality in
the eTransportation survey
13. Basic indicators selection &
robustness check of the framework
• An initial set of 30 indicators were assigned to each
“pillar” (e-service domain) and “sub-pillar” (aspect of
innovation activity being considered)
• 8 Principal Component Analyses and KMO tests were
performed (1 for each sub-pillar) to check the
consistency of the framework
• We applied the eigenvalue-one criterion [only one
eigenvalue should exceed the unity (Kaiser, 1960)] to
make sure that indicators in each sub-pillar share no
more than 1 underlying dimension
• 6 indicators that did not pass this test have been
discarded
14. Pillar Sub-pillar code BASIC INDICATORS SELECTED
E1.1 Teachers using interactive whiteboard
ICT in didactics E1.2 Schools extensively using online text and file/document collections
E1.3 Schools extensively using wiki platforms
eEducation
E2.1 Schools with website
E2.2 Schools providing restricted access areas for web-based info services to teachers
Online Services
Schools providing tools to share training aid files on the web (assignments. audio/video of
E2.3
lessons. etc.)
ICT and changes E3.1 School information system integrated with the National Educational Information System
in internal E3.2 Schools information system integrated with the National Library System
organization
E3.3 Schools with Intranet
G1.1 Municipalities with certified e-mail
eGovernment
ICT and changes
in internal G1.2 Municipalities using e-procurement
organization G1.3 Municipalities using document workflow (full case handling)
G2.1 Municipalities providing fully interactive services on the web
Online services G2.2 Municipalities allowing online payments
G2.3 Channels other than the web used to offer public services
T1.1 No. of technological systems on board
ICT during Cities providing information to travelers about traffic or parking by means of electronic
eTransportation
T1.2
transportation displays
T1.3 Buses with on-board computer
ICT and changes T2.1 Cities with data interchange with other entities
in internal T2.2 Cities with a managing authority for local mobility
organization T2.3 CIties with a mobility monitoring system
T3.1 No. of channels used to inform passengers
Online services T3.2 Cities that provide information to travelers about traffic or parking on the web
T3.3 Cities that offer timetables with route planning (travel planner) on the web
15. Steps for computing CI
• What is the relative importance of each Basic
Indicator?
• How to aggregate the Basic Indicators in order
to measure the level of development of each
region in eEducation, eGovernment and
eTransportation?
• How to calculate the final score?
• What is the robustness level of the results we
obtained?
16. Gathering expert opinion through
Budget Allocation (BA)
What is BA?
Experts are given a “budget” of N points, to be distributed over a
number of individual indicators by “paying” more for those indicators
whose importance they want to stress.
(Moldan and Billharz 1997)
(a) Randomly selected from the
corresponding authors of 751 top-journal
Phases: articles reviewed by Arduini and Zanfei
(2011). => 100 papers extracted.
1. Selection of experts for (b) Also included 15 participants at the 1st
International EIBURS-TAIPS Conference
the evaluation that present papers on eServices
diffusion
2. Allocation of budget to
An on-line questionnaire was administered.
indicators
Experts were asked to allocate a 100 points
budget within each sub-pillar, so that the
3. Calculation of weights total number of indicators to evaluate is < 4
(Bottomley et al., 2000)
17. Results of BA
100
80
60
40
20
0
T1.1
T1.2
T1.3
T2.1
T2.2
T2.3
T3.1
T3.2
T3.3
E2.1
E3.2
E1.1
E1.2
E1.3
E2.2
E2.3
E3.1
E3.3
G1.3
G1.1
G1.2
G2.1
G2.2
G2.3
Mean Max Min Median
No expert consensus on the appropriate set of weights
(Mean coef of var among indicators = 0.4426)
– High variation / disagreement
– No single pair of expert suggesting similar weights
We must choose a statistical method to calculate
weights, while trying not to waste the information
provided by the experts
18. Combining Benefit of the Doubt (BoD)
approach with expert opinion
• BoD is a method for data aggregation based on Data
Evelopment Analysis (DEA) (Melyn & Moesen, 1991, Cherchye et al., 2007)
• BoD advantages
– objective statistical/mathematical approach
– it measures “efficiency” => compares a region’s performance
with a benchmark in a multi-dimensional space
– the algorithm tends to use those indicators where the region
shows better performances
• no other weighting scheme yields higher composite indicator value
(political acceptance)
• reveals policy priorities / past choices
• embeds concern for regional diversity
• BoD + Expert constraint (Cherchye et al., 2008)
– We impose that the use of each indicator is limited by expert
opinion. The MIN (MAX) use of an indicator corresponds to
the MIN (MAX) weight it has received from the experts
19. Benefit of the Doubt (BoD) approach
through Data Envelopment Analysis (DEA)
Through DEA we estimate an
efficiency frontier used as a
benchmark to measure the relative
performance of regions
Indicator = ratio of the distance
between the origin and the
actual observed point and that
of the projected point in the
frontier
In our case, CIs of the 3 pillars
are the distance from an ideal Source: rearranged from Mahlberg and Obersteiner (2001)
case with 100% on all basic
indicators
20. Benefit of the Doubt (BoD) approach
through Data Envelopment Analysis (DEA)
Linear programming problem
j indicates the region
s.t. indicators weights i indicates the indicator
bounding constraint
non-negativity constraint
(Charnes et al, 1978)
21. The “pie-share” constraint
• Applying only the bounding and the non-negativity
constraints may allow for extreme scenarios (Cherchye L.,
2008)
– If a region’s value of one sigle indicator dominates those
of other regions, that region will get the max score of 1
even if it has very low values in the other indicators
• We introduce a pie-share constraint that incorporates
expert opinion (Wong and Beasley, 1990)
Li = lower bound = MIN expert weight from BA
Ui = upper bound = MAX expert weight from BA
22. Results
• In the following slides the resulting scores and
ranks from the constrained optimisation are
presented
• The score:
– represents a measure of a region’s efficiency
compared to the benchmark (the “ideal case”)
– is the sum of the pie-shares of each indicators,
that we have grouped toghether at a sub-pillar
level (aspect of innovation activity being
considered)
23. 0,80
eEducation
0,60
0,40
0,20
-
LOM EMR LAZ VEN TOS CAL BOZ SAR PMN PUG ABR MAR LIG CAM UMB BAS FVG SIC TRE
0,50
eGovernment
0,40
0,30
0,20
0,10
0,00
EMR BOZ TOS VEN LOM MAR FVG UMB PMN PUG SIC CAM LAZ LIG SAR CAL ABR BAS TRE
eTransportation
1,00
0,80
0,60
0,40
0,20
-
BOZ EMR TRE LIG FVG TOS MAR UMB CAM VEN LOM CAL PMN BAS SAR ABR LAZ PUG SIC
Online Services
ICT and changes in internal organization
ICT in didactics (eEdu) or during transportation (eTran)
24. Results per pillar (1/4)
scores
• The highest variation in the scores can be found
in eTransportation domain, while eEducation
performances seem not to vary much
• eGov results for Lombardy, Piedmont and
Province of Trento are lower than expected.
– This is probably due to the high proportion of very
small municipalities
25. Results per pillar (2/4)
rankings
• The 3 rankings differ substantially
=> significantly different regional patterns
– Very high variations in the ranking for the Province of
Trento and Lazio. Medium-high variation for
Lombardy, Calabria, Campania
– Other regions show a more homogeneous approach
to public eServices development which is
characterized by different trajectories of diffusion
• High scores for EMR, TOS, BOZ | medium scores for VEN
MAR PIE | low scores SIC, BAS
26. Results per pillar (3/4)
pie shares
Tracing back pillar results through “pie shares”
• eEducation - Pie shares are more or less fixed, i.e. all
regions use the same “mix” of indicators to
maximize their score, under the expert constraint.
This is due to quite similar relative values of each
indicator and to the specific combination of bounds
that experts have imposed
• eTransportation – Pie shares are flexible, so each
region chooses its own set of weights revealing the
areas where investments have been made
• eGovernment – intermediate case
27. Results per pillar (4/4)
pie shares
• Indicators related to ICT diffusion in internal
processes and organizational changes have a major
role in computing the final score of all public e-
services categories (eEdu, eGov and eTra)
• The importance of back office re-organization
through ICTs has emerged in the literature on the
development of organizations, which has
emphasized the essential role of skills that
characterize the different components of an
organizational structure (Fountain and Osorio-Ursua,
2001; Fountain, 2003; West, 2005; Helfat et al.,
2007)
28. Final steps to the CI
1. Normalization: MIN-MAX,
where MAX is the region with
the highest score
2. Final aggregation through
Geometric Mean
– the marginal gain of an increase
in a low score is much higher
than in a high score
– a region has more incentive to
address the dimensions where it
is weak
29. Final scores
and rank
Region CI Rank
EMR 0,94 1
BOZ 0,93 2
TOS 0,80 3
VEN 0,73 4
FVG 0,70 5
MAR 0,69 6
LIG 0,68 7
LOM 0,67 8
UMB 0,65 9
CAM 0,63 10
PMN 0,59 11
CAL 0,57 12
TRE 0,53 13
LAZ 0,52 14
PUG 0,52 15
SAR 0,51 16
ABR 0,45 17
BAS 0,45 18
SIC 0,38 19
30. Uncertainty Analysis (UA)
• UA is a robustness assessment of a CI (Saltelli et al, 2008)
• The uncertainties in the development of a composite
indicator will arise from some or all of the steps in the
construction line (Saisiana et al, 2004)
(a) selection of subindicators
(b) data selection
(c) data editing
(d) data normalization
(e) weighting scheme
(f) weights' values and
(g) composite indicator formula
(e) level of aggregation where the methodology applies
31. 12 alternative scenarios (+ baseline)
weighting scheme level of the method Aggregation Data normalization
DEA Pie shares (min-max
S1 BA) domains Geometric No rescaling
S2 BA mean weight+EW+EW sub-pillars Additive Minmax
S3 BA median weight+EW+EW sub-pillars Additive Minmax
Additive, geometric on
S4 BA mean weight+EW+EW sub-pillars domains Minmax
Additive, geometric on
S5 BA median weight+EW+EW sub-pillars domains Minmax
S6 DEA Pie shares (min-max BA) domains Additive No rescaling
S7 EW - Additive Minmax
Additive on pillars and
S8 PCA+EW+EW sub-pillars domains Minmax
Additive on pillars,
S9 PCA+EW+EW sub-pillars geometric on domains Minmax
S10 PCA+PCA+EW sub-pillars+pillars Additive Minmax
Additive, geometric on
S11 PCA+PCA+EW sub-pillars+pillars domains Minmax
S12 PCA+PCA+PCA sub-pillars+pillars+domains Additive Minmax
Additive, geometric on
S13 PCA+PCA+PCA sub-pillars+pillars+domains domains Minmax
34. Results of UA
• CI final scores based on BoD weights are
among the best possible results a region can
obtain
• Good robustness level, especially for top and
bottom ranked regions
– 13 regions out of 19 show only a 0/1/-1 shift
compared to the median rank
35. Conclusions 1/2
From a methodological point of view
– BoD approach combined with BA is an effective way
incorporate both regional choices and expert judgment
into CI
– Geometric aggregation gives higher scores to regions
showing a more balanced eServices diffusion among the 3
domains
– Uncertainty analysis on rankings shows high robustness
levels for top and bottom ranked regions
36. Conclusions 2/2
Main findings and implications from our analysis:
– ranking reflects hierarchy of regions in terms of per capita
income and industrial development: current development
of public eServices does not seem to correct unbalances
between regions lagging behind and frontrunners
– high heterogeneity in terms of mix of e-service
proficiency: need for a regional differentiation of e-service
promotion policies ;
– there is more cross regional variation in terms of
eEducation and eTransportation than in terms of eGov:
human capital formation and mobility enhancing are
bound to be the real distinctive assets of regions