You are on page 1of 16

Chapter 1

Measuring the digital transition within the PA:


proposal comparison

Susanna Traversa and Enrico Ivaldi

Abstract Although technological innovation has been one of the most important
drivers of economic recovery in the wake of the Covid-19 health emergency, inter-
ventions incorrectly evaluated may have led to an increase in the territorial digital
divide, exacerbating inequalities among the population. In order to promote a fair
and sustainable recovery in accordance with European and national agendas, three
different proposals for public administration digitisation indexes are presented and
compared in this study. Compared to the most commonly used digitisation indices at
European and national level, this study proposes to construct an index inspired by the
specific eGovernment objectives defined within Mission 1 ’Digitisation, Innovation
and Security of PA’ of the Italian National Recovery and Resilience Plan.Through
two aggregative synthesis methodologies (MIN-MAX and AMPI method) and a
non-aggregative one (POSET), the diffusion of eGovernment forms at the regional
NUTS-2 level is reconstructed, with the aim of verifying the presence and extent of a
digital divide along the Italian peninsula. By means of a correlation analysis, a com-
parison of the three indices obtained is finally presented, assessing the consistency
in the presentation of the results for the identification of a performance evaluation
tool for regional eGovernment in Italy.

Susanna Traversa
University of Genoa, Department of Economics, Italian Center of Excellence on Logistics Trans-
ports and Infrastructures
Genoa, Via Francesco Vivaldi 5 16126, e-mail: susanna.traversa@edu.unige.it
Enrico Ivaldi
University of Genoa, Department of International and Political Science,
Genoa, Piazzale E. Brignole, 3a canc. 16125 e-mail: enrico.ivaldi@unige.it

1
2 Susanna Traversa and Enrico Ivaldi

1.1 Introduction

This paper is an extended version of the paper presented at the 51st Scientific
Meeting of the Italian Statistical Society on June, 2022 [57]. The outbreak of the
Covid-19 pandemic in 2020 was one of the most important and impactful events
of recent decades [35, 50]. The unpreparedness of States to deal with an event of
this magnitude has led to the adoption of contingency measures to minimise the
risk of transmission of the virus within the population lancet. Among the effects
resulting from the introduction of restraining measures during the health crisis -
for example, more or less restrictive forms of isolation and social estrangement -
one of the most important is to be found in the strong acceleration of digitisation
in both the social and economic spheres [50, 6, 28]. The strategic and central role
assumed by digitisation policies, can be traced back to the possibility of preserving
alternative forms of socialisation, limiting possible damage to the national economy
of the countries affected, and ensuring the proper provision of basic public services
[50, 17, 20, 10]. The process of digital integration within the public dimension
is defined by the term eGovernment (eGov). It consist in a complex multidimen-
sional construct, of which scientific literature still provides an ambiguous and not
empirically accepted definition [42], mainly due to the rapid technological and dig-
ital evolution, which leads eGov processes to continuous rethinking and innovation
[52]. However, a commonly accepted interpretation of eGov was proposed by the
Organization for Economic Co-operation and Development (OECD). In order to
achieve ”good government” practices, the OECD defines eGovernment as the use of
information and communication technologies (ICTs), particularly the Internet, for
the provision of digital public services by establishing relationships between their
stakeholders (such as citizens, businesses, employees, and the public sphere itself)
[58]. In this way, it is possible to develop new models of public services delivering
in order to improve a greater quality, accountability, transparency and efficiency of
the public sphere [42, 7, 12], as well as responding to the new demand of services
from citizens while containing costs [49]. The effects and performance of Public
Administration (PA) digitisation processes may vary from country to country. In a
2003 study, Heeks stressed that in Least Developed Countries (LDCs) the success
rate of public digital transition policies appears very low [29], with a 60% increase in
eGovernment, which do not satisfy the objectives set out in the political agendas [26].
Among the possible causes, the author dwells on the presence of barriers different in
nature, such as legislative and regulatory barriers, financial barriers, digital divide
[?, 53, 15]. This latter portrays the complexity and inequities in citizens’ access to
internet services, thus limiting population access to public digital services and gen-
erating negative effects that affect state fiscality, credibility and trust in digitisation
policies, and a forecast increase in future operating costs [31, 36]. The presence
of barriers linked to digital illiteracy or disadvantaged socio-economic status also
creates a further type of non-monetary cost to which the state should pay attention,
including the social stigma linked to the lack of own resources to learn, or put into
practice, a new form of dialogue with the PA [36]. Considering the Italian case -
the subject of this project - it follows that it is important to understand the current
1 Measuring the digital transition within the PA: proposal comparison 3

state of the art of eGov at the territorial level in order to limit any distortions by
exacerbating inequalities among the population. With the spread of Sars-Cov-2 in
the early 2020s, there has been an international acceleration in digital integration
processes as a response to pandemic containment policies [50, 27, 5, 9, 45]. The
positive effects stemming from the widespread diffusion of new technologies are
also not limited only to private entrepreneurship and Industry 4.0, for which it is
well-established the possibility to create value thanks to cost containment, an in-
crease in efficiency and productivity in business management processes [55, 47] and
improvement of environmental and economic sustainability. Several studies confirm
how technological development and per-capita GDP are interrelated [?, 51], thus
making digital transition strategies extremely strategic even in public sector, in order
to increase national competitivity, the quality of public services and the efficiency of
e-Government processes [38, 34]. However, the increase in demand for connectivity
has not only produced palliative effects but has also led to the emergence of new
forms of inequality. In fact, several states have had to face digital re-conversion pro-
cesses encountering inadequate digital infrastructures vulnerable to cyber attacks,
a still too widespread digital illiteracy among the population and the presence of
financial barriers [32, 36, 45]. For this reason, policy makers often approach the
digitisation process as an exogenous factor, setting aside the social context when
developing new strategies [47].In order to limit the potential distortions produced
by the acceleration of digitisation processes, the European Union has promoted,
through a series of special investment funds and political agendas, the implementa-
tion of an equitable, sustainable digital strategy inspired by European values [19, 18].
According to the objectives set by the European Institutions, the Member States are
directed towards promoting new reform and investment policies capable of limiting
the spread and strengthening of asymmetries within the national territory EC2020-c.
Among the possible growth opportunities promoted by European institutions the
Italian Government has identified the opportunity of using the European funds to
start a transversal digital implementation programme. In this way, the Italian country
aims at improving its performance compared to the European scenario, where it is
currently extremely disadvantaged compared to the European average (rank 24 out
of 27 in 2020 in European rankings) [20]. For the digital transition 27% of Euro-
pean resources obtained have been allocated in the Italian ”National Recovery and
Resilience Plan” (NRRP), in order to be invested on the basis of the objectives con-
tained in Mission 1 ”Digitisation, innovation, competitiveness, culture and tourism”
[32]. Among the investments included in the recovery plan great importance is given
to the modernisation of the current digital infrastructure, the strengthening of cyber
security - extremely fragile in the Italian public sector, the pursuit of greater inter-
operability between internal databases (thanks also to the development of common
cloud clients between the different administrative levels), and an improvement in
the quality of public digital services currently offered to citizens and businesses.
Therefore, it becomes evident the importance of having tools to assess performance
at both a temporal level-monitoring the progress of the interventions put in place by
the government-and at a territorial one, in order to determine which areas are more
critical compared to the national scenario. In the literature, specific proposals for the
4 Susanna Traversa and Enrico Ivaldi

Italian case are currently limited to the construction of synthetic indices that, tracing
the Digital Economy and Society Index (DESI), propose the development of tools
that assess digitization in its broadest sense [8, 56]. Among recent experiences in
this field, the regional ICT Observatory of Piedmont has carried out a survey in 2019
and 2020 on the technological and digital development of Piedmont, comparing its
performance with that of other Italian regions. Although the proposal is structurally
inspired by the dimensions used for the European DESI, an aggregation phase is not
carried out in order to obtain a synthetic index that provides an overall digitisation
score for the regions. More structured instead is the proposal of the ”Osservatorio
dell’Agenda Digitale del Politecnico di Milano”. The composite index constructed
for the year 2020 was realised using 35 of the 37 variables provided by the European
DESI, although some of the variables used do not have the data available for the
regional territorial level. Lastly, more recent and comprehensive is the proposal put
forward in November 2021 by the Bank of Italy [8]. In an occasional economics and
finance paper, the Bank of Italy presents its own version of rDESI, which maintains
the dimensions and weighting system used for the European DESI, using Principal
Component Analysis as the aggregation technique. All proposals presented note the
actual existence of a digital gap between the regions of Northern and those of South-
ern Italy. The development of an accurate mapping and monitoring quantitative tool
for the NUTS-2 Italian level will represent a useful and strategic analysis of public
policies performances with the aim of promoting investment in the digital transition.
The possibility of constructing an ad-hoc tool for the Italian context, calibrated to
the Public Administration digitization targets envisaged by the NRRP and assess-
ing regional performance in order to emphasise the problem of the digital divide,
therefore deserves further investigation. For this reason, this study introduces three
methodologies for the construction of digitisation indices of the Italian Public Ad-
ministration, based on the intervention dimensions identified by the Government in
the NRRP. By limiting the synthesis techniques to methodologies that allow tempo-
ral comparison, it will also be possible to trace the performance trends of individual
regions over time and provide policy makers useful instruments for the development
and remodulation of interventions.
It becomes therefore necessary to develop quantitative statistical tools able not
only to express the complexity and multidimensional nature of the phenomenon, but
also to allow its study over time with the aim of monitoring its state of progress. In
this paper, a comparison is proposed between three methodologies for the synthesis
of multidimensional indices, in order to identify a possible proposal for the study of
public digitisation within the Italian regional dimension. The choice of implementing
the study of the spread of the eGovernance model in relation to the NUTS-2 territorial
dimension is to be found in the strong heterogeneity that characterises the Italian
nation. Moreover, the monitoring carried out at the national and European levels
confirmed the existence of a marked gap between the northern and southern Italian
regions in terms of digitisation of the peninsula. In this way, the synthetic indices
can be exploited to highlight any criticalities in the Italian context with respect to the
more specific objectives described by the NRRP, thus directing policy interventions.
1 Measuring the digital transition within the PA: proposal comparison 5

1.2 Selection of data

The variables used were selected through I.Stat database by a formative approach
[14], and with the guidelines included by the Italian Government in the M1.C1 Mis-
sion of the NRRP, dedicated to the ”Digitisation, innovation and security within the
PA” (Italian Government, 2020). Therefore, seven variables expressing the develop-
ment of eGovernment were identified: (1) percentage of public institutions adopting
cloud computing services to manage their data (CCS); (2) percentage of people aged
16-74 years who have used the Internet in the last 3 months and have basic digital
skills (BDS); (3) percentage of institutions that have used all IT security measures
(CS); (4) percentage of households with access to a broadband connection (BBC);
(5) percentage of people (>14 years old) who contacted the PA in the last 12 months
to obtain information through the internet (DSI); (6) percentage of people (>14
years old) who interacted with the PA in the last 12months to download pre-filled
forms through the internet (DSPFF); (7) percentage of people (>14 years old) who
have contacted the PA in the last 12months to send filled forms via the internet
(DSSM). The great heterogeneity that characterise the Italian territories, represent
an important and crucial factor in the development of national indices for the eval-
uation of complex phenomena. According to these assumptions, the present thesis
has been elaborated in order to provide a regional scenario of the implementation
of eGovernment. This will allow a greater focus on the different distribution of
digitalization processes on the territory, and provide policy makers some important
tools for the evaluation of territorial investment interventions. The subdivision of
the Italian territory into individual regions was carried out in accordance with the
European Nomenclature of territorial units for statistics (NUTS) at level NUTS-2
(Regulation (EC) No 1059/2003).

1.3 Methodology: comparing synthesis techniques

As emerged from the study of the related literature, the multidimensional character
of digitization even in ”public key” appeared extremely evident and pronounced. In
socio-economic statistics, it is common to approach the study of complex phenomena
through the development of synthetic measures that allow for a simplification of the
phenomenon and greater interpretability of the results for both the researcher and
the public. [40]. Traditionally, the development of synthetic indicators is approached
based on Lazarsfeld’s hierarchical model which theorises that the synthesis pro-
cess must respect a series of requirements and stages aimed at limiting ambiguities
and errors in the interpretation of the results. [37]. For this study, a formative ap-
proach - widely employed for the study of socioeconomic phenomena [16] - has
been adopted to avoid high correlation variables, in order to minimise potential
redundancies within the synthetic indicator [48]. Index synthesis can be handled
by taking either an aggregative or non-aggregative approach [41]. Aalthough the
’synthesis-as-aggregation’ paradigm appears to be the most widely used in the liter-
6 Susanna Traversa and Enrico Ivaldi

ature, the implementation of such techniques is not exempt from potential criticality
from both a conceptual and methodological point of view [1, 21, 3]. The mathe-
matical manipulation of the variables during the aggregation and weighting phases,
entails a high risk related to the loss of representativeness of the scores obtained
from the aggregated index with respect to the initial meaning of the indicators [39].
Moreover, the aggregation process itself intervenes in the multidimensionality of
a phenomenon with the aim of obtaining a one-dimensional measure. This leads
to a ”flattening” effect with respect to the interpretation of the phenomenon itself,
since the possible incomparability between indicators of different natures loses its
meaning [23]. Therefore, considering the complexity of socio-economic phenom-
ena, in recent years some authors have approached the synthesis of indicators by
adopting more sophisticated techniques, capable of enhancing the multidimension-
ality of a phenomenon by avoiding the aggregation and weighting processes of its
individual dimensions, limiting the ”flattering” effect [2, 4, 33, 13]. However, non-
aggregative techniques present, as well, some problems mainly related to the greater
computational and interpretive complexity when compared with techniques that take
advantage of the aggregative approach. Setting as a prerequisite the possibility of
comparing the scores obtained by individual territories over time, a number of syn-
thesis techniques have been identified, which will be further explored and compared
over the course of the research period in order to identify their potential with respect
to the research objectives. Among the aggregative techniques, the first that will be
addressed is the 𝑀𝑖𝑛 − 𝑀𝑎𝑥 aggregation method, which sees its application for
the construction of the European DESI digitization index [20]. It has been applied
for the construction of two DESI index proposals declined in a regional key for
the Italian context [8, 56]. Both proposals in addition to ensuring the possibility of
making a temporal assessment of the phenomenon, detect the existence of a digital
gap between the regions of Northern and Southern Italy.
𝑥 𝑥 𝑗 − 𝑀𝑖𝑛 𝑥 𝑗
𝑟𝑥 𝑗 = (1.1)
𝑀𝑎𝑥 𝑥 𝑗 − 𝑀𝑖𝑛 𝑥 𝑗
𝑛
∑︁
𝐷𝑖𝑔𝑖𝑡𝑎𝑙 𝑀 𝐼 𝑁 − 𝑀 𝐴𝑋 𝑝𝑚 = (𝑥𝑖 ) + .. + (𝑥 𝑛 ) = 𝑥𝑖 (1.2)
𝑖=1

The second aggregate index implemented the partially non-compensatory Ad-


justed Mazziotta and Pareto Index (𝐴𝑀 𝑃𝐼 ± ) method [43]. Thanks to its character-
istics, the Digital 𝐴𝑀 𝑃𝐼± derived from the 𝐴𝑀 𝑃𝐼 ± method represent versatile tools,
easy to interpret and replicate, and suitable for the study of phenomena over time.
Compared with the previous methodology, the 𝐴𝑀 𝑃𝐼 ± index exploits an enhanced
re-scaling process as its normalisation technique, which allows it to overcome some
of the critical issues typical of 𝑀𝑖𝑛 − 𝑀𝑎𝑥, such as the strong sensitivity to outliers
and the inability to center the average. By balancing these problems through the
integration of some typical properties of the indexing method, the indicators will
therefore be normalised to a range of variation and center to a reference (e.g. the
average) [44]. In addition, the possibility of providing for the identification of goal-
1 Measuring the digital transition within the PA: proposal comparison 7

posts calculated on the starting year of the period under consideration allows for the
carrying out of a time analysis in an absolute sense whereby the index values will be
related to the performance recorded for the reference year [43].
𝑥𝑖 𝑗 − 𝑀𝑖𝑛 𝑥 𝑗
𝑟𝑖 𝑗 = 60+70 (1.3)
𝑀𝑎𝑥 𝑥 𝑗 − 𝑀𝑖𝑛 𝑥 𝑗

𝑀𝑎𝑥 𝑥𝑖 𝑗𝑡 − 𝑀𝑖𝑛 𝑥𝑖 𝑗𝑡
𝑅𝑒 𝑓𝑖 𝑗 ± △ 𝑤𝑖𝑡ℎ △ = (1.4)
2

𝐷𝑖𝑔𝑖𝑡𝑎𝑙 𝐴𝑀 𝑃𝐼± = 𝜇𝑟𝑖 ± 𝜎𝑟𝑖 (1.5)


With regard to the hypothesis of a non-aggregative synthetic index, among the
techniques that have observed increased popularity in recent years is Partially Or-
dered Sets (POSET) [54, 21]. Among the strengths of the POSET methodology is
the possibility of employing them to construct synthetic indexes that exploit both
qualitative and quantitative variables. Moreover, by not requiring a process of ag-
gregation or weighting of variables, it allows limiting some of the critical issues
typical of composite indices (e.g., flattering effect) due to the presence of incompa-
rability within the multidimensional system of indicators. In fact, the existence of
incomparability represents an element that in the case of POSET allows for greater
contextualisation of the results that will be obtained from the indicator synthesis,
highlighting profiles of digitization that differ from each other, making it possible
to identify areas of the country that have - or do not have - similar eGov processes.
[30] POSET is represented by means of the use of a Hasse Diagram (Fig.??), which
allows us to highlight the comparability or non-comparability relationships between
two profiles, based on the following properties [11, 22]: reflexivity, asymmetry,
transitivity. As pointed out by [25], the synthesis of the POSET into a vector must
necessarily be done by exploiting order-preserving techniques. For this reason, the
average height (avh) approach represents the most common method for the synthesis
of multi-indicator systems through the study of the POSETs. By assuming that the
average rank of an element 𝑥 ∈ 𝑋 of P is the average over Ω 𝑝 of the number of
elements 𝑦 ∈ 𝑋 : 𝑦 ⊴𝑙 𝑥 in linear extensions 𝑃𝑙 = (𝑋, ⊴𝑙 ) ∈ Ω 𝑝 . From a practical
point of view, the Digital 𝑃𝑂𝑆𝐸𝑇 is obtained by following a step-wise procedure:
• Extract all linear extensions of P creating 𝑂𝑚𝑒𝑔𝑎 𝑝 .
• For each element 𝑝 ∈ 𝑃 · and for each 𝑙 ∈ Ω(𝑃·) assign a rank 𝑟 𝑙 ( 𝑝) of p in l,
representing 1+ the number of angular coverings joining p to the maximum of l.
• The third step is to calculate the average of 𝑟 ( 𝑝) of 𝑟 𝑙 ( 𝑝) over Ω(𝑃) for each
𝑝 ∈ 𝑃· .
However, as anticipated, these are techniques that require the use of advanced com-
putational tools, making their development complex. Computation of the results was
conducted by means of the statistical software Rstudio and the package “parsec”
[24].
8 Susanna Traversa and Enrico Ivaldi

1.4 Discussion

Focusing firstly on the rankings obtained from the application of the aggregate
indices - Digital 𝑀 𝐼 𝑁 −𝑀 𝐴𝑋 and Digital 𝐴𝑀 𝑃𝐼 − - a high degree of agreement with
respect to the rankings attributed to the individual exceptions emerges (Tab.1.1).
This is further confirmed by the results of Kendall’s correlation coefficient, which
reports a high positive correlation between the two proposals. The test that reports a
greater affinity is the one carried out on the two composite indices (𝜏𝑏 = 0.91). On the
contrary, the index values are lower when compared to the non-aggregative variant.
In fact, the Digital 𝑃𝑂𝑆𝐸𝑇 shows a lower - though still strong - correlation with the
indices Digital 𝑀 𝐼 𝑁 − 𝑀 𝐴𝑋 e Digital 𝐴𝑀 𝑃𝐼± . The divergence of the Digital 𝐴𝑀 𝑃𝐼 − ranks
from Digital 𝑀 𝐼 𝑁 −𝑀 𝐴𝑋 appears generally contained, with the sole exception of TRE.
While the North-Eastern region ranks second in Digital 𝑀 𝐼 𝑁 − 𝑀 𝐴𝑋 , Digital 𝐴𝑀 𝑃𝐼 −
assigns it only rank 7.
The underestimation of performance with respect to the Italian 2020 context is
also confirmed by the maps resulting from cluster analysis of Ward [59, 46], in which,
on the basis of the partitions calculated for Digital 𝐴𝑀 𝑃𝐼 − , TRE is placed within the
second interval for the level of digitalisation of the public sector (Fig.??).
However, when considering the non-aggregative proposal, the Digital 𝑃𝑂𝑆𝐸𝑇 rank-
ing once again rewards the performance of TRE, by attributing first rank instead.
The analysis of the Hasse Diagram (Fig.??) in fact revealed the dominant position of
TRE with respect to the set of profiles, as well as a contained internal variability with
regard to the average rank which, as shown graphically and during the construction
of the index by means of the average high, oscillates in a range [1:6].
On the other hand, the two aggregate indices agree on the recognition of the best
performance of VEN, a region that compared to the Digital 𝑃𝑂𝑆𝐸𝑇 ranking falls to
the fifth rank. TRE and VEN also represent non comparable profiles (TRE ∥ VEN).
In fact, as can be seen from the study of the performances obtained in the basic
indicators - although the two regions score higher than the Italian average - VEN
shows in 2020 better performances in terms of implementation of cloud computing
services and citizens’ use of digital public services (Fig.??).
On the other hand, TRE shows greater percentages in the variables BBC, CS and
BDS, pointing to a better implementation from the digital infrastructure viewpoint,
as well as to a greater adoption - in absolute terms - of cyber security protocols and to
a more widespread diffusion of basic digital skills among the population. Focusing
now on the maps obtained from the cluster analysis, although a digital gap between
northern and southern regions of Italy is evident, a counter tendency emerges in
some northern regions - VDA, LOM, LIG - in the TOS and for the SAR.
For these regions, there is a different partition than the average of the macro-area to
which they belong.
In particular, it is interesting to note that LIG, TOS, and SAR (Fig.??) represent
also the only three regions whose profiles are not comparable within POSET Hasse
diagram. The variability in performance is also confirmed by the average ranks graph
(Fig.??) and the different trend in percentage terms with respect to the development
of the individual dimensions of digitisation. The interval between the minimum and
1 Measuring the digital transition within the PA: proposal comparison 9

maximum value of the range is extremely wide and the same for all three regions
(range = 19).
With regard to LOM and VDA, despite being in the northern macro-area of
the peninsula, both aggregate indices place VDA and LOM in the intermediate
class (third cluster). From the point of view of rank positioning, there are no major
fluctuations for LOM, which is rewarded more by the Digital 𝑃𝑂𝑆𝐸𝑇 (rank 7𝑡 ℎ ),
although for VDA the three indices report different ratings but still within a contained
range [9,11].
However, if we consider the Digital 𝑃𝑂𝑆𝐸𝑇 , we can see that, in reality, even though
the two regions belong to the same level, they are not comparable at all. LOM, in
particular, is dominated by LAZ (4𝑡 ℎ POSET rank) and, by transitive property, by
TRE (1𝑠𝑡 Digital 𝑃𝑂𝑆𝐸𝑇 rank). On the other hand, it performs better than MAR,
ABR and CAL, to which it is linked by a descending path and which occupy,
respectively, in the average height classification, the 18𝑠𝑡 , 16𝑠𝑡 and 20𝑠𝑡 positions.
Even though there is no minimum node among the Digital 𝑃𝑂𝑆𝐸𝑇 components - but a
set of minimum ones - the existing criticalities for the CAL region are confirmed. In
fact, as reported by the Digital 𝑃𝑂𝑆𝐸𝑇 index, also in the case of the two proposals of
aggregative nature, CAL represent the region with the lowest level of implementation
of eGovernment processes in 2020. Furthermore, by investigating the relationships
between the different profiles thanks to the dominance matrix, it is observed that the
CAL region corresponds to the profile with the most cover relations above. As in the
case of CAL, all three indices confirm SIC’s position at 19𝑠𝑡 , referring to it as the
second most critical region for digital transition in Public Administration.
By comparing the results obtained in each individual dimension of eGovernment,
the two regions’ performance actually shows minimal difference, with the only dis-
cordance observed in the CCS dimension. A further similarity between the results
obtained from the three indices is related to the presence of a marked digital gap
between the northern and southern Italian regions. Although no cluster analysis was
carried out for the Digital 𝑃𝑂𝑆𝐸𝑇 index, by analysing the positioning of the nodes on
the Hasse diagram and comparing it with those of the southern regions in the three
rankings provided, the existence of a territorial gap can be confirmed.
If from one side the cluster analysis offers the possibility to identify a clear territorial
separation between the northern and central regions - facing the Tyrrhenian Sea - and
the southern and central regions - facing the Adriatic Sea -, the information about
the performance within the POSET system can be easily deduced from the analysis
of the Hasse diagram.
In addition to the non-comparable profiles and CAL and SIC, the bottom Hasse
line is composed exclusively of regions belonging to the macro area of Southern
Italy. The positioning at the bottom of the Hasse diagram, as previously stated, indi-
cates a condition of total dominance - if a cover relation between pairs of nodes is
present - such as to identify among these profiles a lower level of digital development.
The profiles expressing all subordination to the higher ones are MOL (Digital 𝑃𝑂𝑆𝐸𝑇 =15𝑡 ℎ
rank), ABR (Digital 𝑃𝑂𝑆𝐸𝑇 =16𝑡 ℎ rank), BAS (Digital 𝑃𝑂𝑆𝐸𝑇 =17𝑡 ℎ rank), MAR
(Digital 𝑃𝑂𝑆𝐸𝑇 =18𝑡 ℎ rank) - profiles already treated and not compatible are ex-
10 Susanna Traversa and Enrico Ivaldi

cluded. Although these four profiles are incompatible with each other, they share -
either directly or by transitive property - a subordinate relationship to the FRI profile.
Considering Fig.??, it is evident that FRI has a better position in terms of Italian
ranking, thanks to percentages that are above the national average in all the surveyed
dimensions and holding the second rank in both the Digital 𝐴𝑀 𝑃𝐼 − and Digital 𝑃𝑂𝑆𝐸𝑇
rankings, and the third in the case of the Digital 𝑀 𝐼 𝑁 −𝑀 𝐴𝑋 . On the other hand, the
ranks attributed by the three indices to the BAS region are discordant: 15𝑡 ℎ posi-
tion (Digital 𝑀 𝐼 𝑁 − 𝑀 𝐴𝑋 rank), 14𝑡 ℎ position (Digital 𝐴𝑀 𝑃𝐼 − rank) and 17𝑡 ℎ position
(Digital 𝑃𝑂𝑆𝐸𝑇 rank). Looking for motivations through the study of the average high
graph, a wide downward variability is observed. This indicates a greater propensity
for the index to take values that would place it at the bottom of the rankings - range
of rank variation between [6,20].

1.5 Conclusion

The digitalisation of the public sector represents nowadays a topical and strategic
issue. During the Covid-19 emergency, weaknesses in the current digital transition
process were evident within the Italian system. In spite of the progress made since
the promotion of the European Digital Agenda, Italy still struggles to meet European
standards, ranking in the global classification drawn up by the 2021 report on the
Digital Economy and Society Index (DESI) in 21 place over 28 (considering EU
average), and in 18𝑡 ℎ place in the sole dimension dedicated to the diffusion of
digital public services. In order to better understand, from a territorial perspective,
which areas of the country show greater difficulties from the point of view of the
implementation of eGovernment models, three proposals of synthetic indices have
been presented in the course of this study, in order to express the multidimensionality
of the phenomenon and carrying out a temporal analysis. The methodological choice
fell on the elaboration of two composite indices - i.e. aggregative approach - and a
non-aggregative alternative by using the technique of Partially Ordered Set. First of
all, thanks to the cluster analysis carried out on the composite indices, the existence
of a digital gap in terms of eGov adoption is evident.
All three indices indicate that SIC and CAL are the most critical regions compared
to the Italian context, with the 19𝑡 ℎ and 20𝑡 ℎ rankings respectively. From the POSET
index perspective, however, a discrepancy emerges between the two profiles, which
are actually incomparable, as observed by studying the original data. SAR, LIG and
TOS present distinctive features. While on the one hand the composite indexes place
these three regions - two of which are in the centre-north - among those with medium
digitalisation performances, POSET provides more information. From the study of
the non-aggregative index, in addition to the lack of comparability with any other
profile, it’s emerged also a great internal variability from the point of view of the
range between which the ranks are assigned to the regions by the simulation.
As a final operation on the indicators, a correlation analysis between the three in-
dices was carried out, using Kendall’s aggregation coefficient (𝜏𝑏 ) as a technique.
1 Measuring the digital transition within the PA: proposal comparison 11

After verifying the correct existence of the necessary requirements for the use of
Kendall’s method, the results for the three indices were obtained. Through the func-
tion ”corr.test()” implemented in R, an extremely positive correlation was obtained in
all three cases. Between Digital 𝐴𝑀 𝑃𝐼 − and Digital 𝑀 𝐼 𝑁 − 𝑀 𝐴𝑋 for example, 𝜏𝑏 =0.91
- indicating a near-superposition between the two rankings. Lower is instead, al-
though very positive, 𝜏𝑏 =0.87 calculated between Digital 𝑀 𝐼 𝑁 −𝑀 𝐴𝑋 and POSET.
Finally, between Digital 𝐴𝑀 𝑃𝐼 − and POSET there is a correlation 𝜏𝑏 =0.78.
Although the values reported by the correlation index may suggest the choice be-
tween Digital 𝑀 𝐼 𝑁 − 𝑀 𝐴𝑋 and POSET, it is necessary to consider the higher robustness
obtained by the influence analysis test for the Digital 𝐴𝑀 𝑃𝐼 − index. Therefore, it is
suggested that the final comparison between the latter aggregative technique and the
POSET index should be the non-aggregative alternative.
In the light of what emerged during the analysis of the technical literature and
from the considerations during the discussion of the results, we confirm the limita-
tions linked to the use of synthetic indicators to the advantage of non-aggregative
measures, which make it possible to understand the real variability that exists for a
statistical unit, the existing relations of order between profiles without thinking about
incompatibilities incurring a ”flattening” risk in the interpretation of the results.
Over the coming years, it will be necessary through an update of the data-sets to
continue monitoring the evolution of eGovernment systems at the NUTS-2 level, in
order to understand the real state of art of government policies in terms of digital
transition.
12 Susanna Traversa and Enrico Ivaldi

COD. Digital 𝑀𝐼 𝑁 − 𝑀 𝐴𝑋 Digital 𝐴𝑀 𝑃𝐼− Digital 𝑃𝑂𝑆𝐸𝑇


ABR 17 18 16
BAS 15 14 17
CAL 20 20 20
CAM 13 13 12
EMI 6 5 3
FRI 3 2 2
LAZ 4 3 4
LIG 10 11 9
LOM 9 9 7
MAR 18 17 18
MOL 16 15 15
PIE 5 4 6
PUG 14 16 14
SAR 8 8 9
SIC 19 19 19
TOS 12 12 9
TRE 2 7 1
UMB 7 6 7
VDA 11 10 13
VEN 1 1 5
Table 1.1 Ranking comparison obtained for Digital 𝑀 𝐼 𝑁 −𝑀 𝐴𝑋 , Digital 𝐴𝑀 𝑃𝐼 − , Digital 𝑃𝑂𝑆𝐸𝑇 .
References

[1] Alaimo, L. S., Ciacci, A., and Ivaldi, E. Measuring sustainable development
by non-aggregative approach. Social Indicators Research (2020), 1–22.
[2] Alaimo, L. S., et al. Sustainable development and national differences: An
european cross-national analysis of economic sustainability. RIEDS-Rivista
Italiana di Economia, Demografia e Statistica-Italian Review of Economics,
Demography and Statistics 72, 3 (2018), 101–123.
[3] Alaimo, L. S., Ivaldi, E., Landi, S., and Maggino, F. Measuring and eval-
uating socio-economic inequality in small areas: An application to the urban
units of the municipality of genoa. Socio-Economic Planning Sciences (2021),
101170.
[4] Alaimo, L. S., and Maggino, F. Sustainable development goals indicators at
territorial level: Conceptual and methodological issues—the italian perspec-
tive. Social Indicators Research 147, 2 (2020), 383–419.
[5] Alfano, V., and Ercolano, S. The efficacy of lockdown against covid-19: a
cross-country panel analysis. Applied health economics and health policy 18
(2020), 509–517.
[6] Almeida, F., Santos, J. D., and Monteiro, J. A. The challenges and op-
portunities in the digitalization of companies in a post-covid-19 world. IEEE
Engineering Management Review 48, 3 (2020), 97–103.
[7] Alshehri, M., Drew, S., Alhussain, T., and Alghamdi, R., Eds. The Effects
of Website Quality on Adoption of E-Government Service: AnEmpirical Study
Applying UTAUT Model Using SEM (December 2012).
[8] Benecchi, A., Ciapanna, E., Bottoni, C., Frigo, A., Milan, A., and Scar-
inzi, E. Digitalisation in italy: Evidence from a new regional index. Bank of
Italy Occasional Paper, 662 (2021).
[9] Bratta, B., Romano, L., Acciari, P., and Mazzolari, F. The impact of
digitalization policies. evidence from italy’s hyper-depreciation of industry 4.0
investments. DF Working Papers.
[10] Casquilho-Martins, I., and Belchior-Rocha, H. Responses to covid-19
social and economic impacts: A comparative analysis in southern european
countries. Social Sciences 11, 2 (2022), 36.

13
14 Susanna Traversa and Enrico Ivaldi

[11] Davey, B. A., and Priestley, H. A. Introduction to lattices and order. Cam-
bridge university press, 2002.
[12] Dawes, S. S. Governance in the digital age: A research and action framework for
an uncertain future. Government Information Quarterly 26, 2 (2009), 257–264.
[13] Di Bella, E., Leporatti, L., Maggino, F., and Gandullia, L. A poset
based indicator of gender equality at sub-national level. Advances in Statistical
Modelling of Ordinal Data (2018), 109.
[14] Diamantopoulos, A., and Siguaw, J. A. Formative versus reflective indi-
cators in organizational measure development: A comparison and empirical
illustration. British journal of management 17, 4 (2006), 263–282.
[15] Dierx, A., Ilzkovitz, F., Pataracchia, B., Ratto, M., Thum-Thysen, A.,
and Varga, J. Does eu competition policy support inclusive growth? Journal
of Competition Law & Economics 13, 2 (2017), 225–260.
[16] Edwards, J. R., and Bagozzi, R. P. On the nature and direction of relationships
between constructs and measures. Psychological methods 5, 2 (2000), 155.
[17] European Commission. 2030 digital compass: the european way for the digital
decade, 2020. Available on Eurlex.
[18] European Commission. Europe’s moment: Repair and prepare for the next
generation, 2020. Available on line.
[19] European commission. Digital economy and society index (DESI) 2020
thematic chapters, 2020c. Available on line.
[20] European commission. Digital economy and society index (desi) 2021 the-
matic chapters, 2021. Available on line.
[21] Fattore, M. La valutazione statistica del benessere: ordinamenti parziali e
nuovi paradigmi metodologici. Lavoro, Economia, Finanza 10 (2013).
[22] Fattore, M. Partially ordered sets and the measurement of multidimensional
ordinal deprivation. Social Indicators Research 128, 2 (2016), 835–858.
[23] Fattore, M. Synthesis of indicators: The non-aggregative approach. In Com-
plexity in society: From indicators construction to their synthesis. Springer,
2017, pp. 193–212.
[24] Fattore, M., and Arcagni, A. Parsec: An r package for poset-based evaluation
of multidimensional poverty. In Multi-indicator systems and modelling in
partial order. Springer, 2014, pp. 317–330.
[25] Fattore, M., Arcagni, A., and Maggino, F. Optimal scoring of partially
ordered data, with an application to the ranking of smart cities. Smart Statistics
for Smart Applications Book of Short Papers SIS (2019).
[26] Furuholt, B., and Wahid, F. E-government challenges and the role of political
leadership in indonesia: the case of sragen. In Proceedings of the 41st Annual
Hawaii International Conference on System Sciences (HICSS 2008) (2008),
IEEE, pp. 411–411.
[27] Gabryelczyk, R. Has covid-19 accelerated digital transformation? initial
lessons learned for public administrations. Information Systems Management
37, 4 (2020), 303–309.
[28] Grasso, M., Klicperová-Baker, M., Koos, S., Kosyakova, Y., Petrillo,
A., and Vlase, I. The impact of the coronavirus crisis on european societies.
1 Measuring the digital transition within the PA: proposal comparison 15

what have we learnt and where do we go from here?–introduction to the covid


volume, 2021.
[29] Heeks, R. Most egovernment-for-development projects fail: how can risks be
reduced?
[30] Heeks, R. e-government as a carrier of context. Journal of public policy 25, 1
(2005), 51–74.
[31] Hwang, M.-S., Li, C.-T., Shen, J.-J., and Chu, Y.-P. Challenges in e-
government and security of information. Information & Security 15, 1 (2004),
9–20.
[32] Italian Government. Piano Nazionale di Ripresa e Resilienza, 2021.
[33] Ivaldi, E., Penco, L., Isola, G., and Musso, E. Smart Sustainable Cities
and the Urban Knowledge-Based Economy: A NUTS3 Level Analysis. Social
Indicators Research 150, 1 (July 2020), 45–72.
[34] Kotarba, M. Measuring digitalization–key metrics. Foundations of Manage-
ment 9, 1 (2017), 123–138.
[35] Kuc-Czarnecka, M. Covid-19 and digital deprivation in poland. Oeconomia
Copernicana 11, 3 (2020), 415–431.
[36] Larsson, K. K. Digitization or equality: When government automation covers
some, but not all citizens. Government Information Quarterly 38, 1 (2021),
101547.
[37] Lazarsfeld, P. F. Evidence and inference in social research. Daedalus 87, 4
(1958), 99–130.
[38] Lindgren, I., and Jansson, G. Electronic services in the public sector: A
conceptual framework. Government Information Quarterly 30, 2 (2013), 163–
172.
[39] Madden, D. Ordinal and cardinal measures of health inequality: an empirical
comparison. Health Economics 19, 2 (2010), 243–250.
[40] Maggino, F., Ed. Complexity in Society: From Indicators Construction to their
Synthesis, vol. 70 of Social Indicators Research Series. Springer International
Publishing, Cham, 2017.
[41] Maggino, F. Dealing with Syntheses in a System of Indicators. In Complexity
in Society: From Indicators Construction to their Synthesis, F. Maggino, Ed.,
vol. 70. Springer International Publishing, Cham, 2017, pp. 115–137. Series
Title: Social Indicators Research Series.
[42] Malodia, S., Dhir, A., Mishra, M., and Bhatti, Z. A. Future of e-
government: An integrated conceptual framework. Technological Forecasting
and Social Change 173, 121102 (2021).
[43] Mazziotta, M., and Pareto, A. Measuring well-being over time: The ad-
justed Mazziotta–Pareto index versus other non-compensatory indices. Social
Indicators Research 136, 3 (2018), 967–976. ISBN: 1573-0921 Publisher:
Springer.
[44] Mazziotta, M., and Pareto, A. Composite Indices Construction: The Per-
formance Interval Approach. Social Indicators Research (Apr. 2020).
[45] Meliciani, V., and Pini, M. Digitalizzazione e produttività in italia: Oppor-
tunità e rischi del pnrr.
16 Susanna Traversa and Enrico Ivaldi

[46] Murtagh, F., and Legendre, P. Ward’s Hierarchical Agglomerative Clus-


tering Method: Which Algorithms Implement Ward’s Criterion? Journal of
Classification 31, 3 (Oct. 2014), 274–295.
[47] Musik, C., and Bogner, A. Book title: Digitalization & society, 2019.
[48] Nardo, M., Saisana, M., Saltelli, A., Tarantola, S., Hoffman, H., and
Giovannini, E. Handbook on constructing composite indicators: methodology
and user guide. organisation for economic cooperation and development (oecd).
Statistics Working Paper JT00188147, OECD, France (2005).
[49] OECD. e-government for better government, 2005.
[50] OECD. OECD digital economy outlook 2020. OECD Publishing (2020).
[51] Parra, J., Pérez-Pons, M.-E., and González, J. Study based on the inci-
dence of the index of economy and digital society (desi) in the gdp of the
eurozone economies. In International Symposium on Distributed Computing
and Artificial Intelligence (2020), Springer, pp. 164–168.
[52] Relyea, H. C. E-gov: Introduction and overview. Government information
quarterly 1, 19 (2002), 9–35.
[53] Rückert, D., Weiss, C., and Veugelers, R. EIB Working Paper 2020/07-The
growing digital divide in Europe and the United States (Volume 2020/7). 2020.
[54] Schröder, B. S. Ordered sets. Springer 29 (2003), 30.
[55] Stock, T., and Seliger, G. Opportunities of sustainable manufacturing in
industry 4.0. procedia CIRP 40 (2016), 536–541.
[56] Traversa, S., Ciacci, A., Ivaldi, E., and González-Relaño, R. Measuring
the digital gap in italy: A nuts-2 level index. In Globalization, Income Dis-
tribution and Sustainable Development. Emerald Publishing Limited, 2022,
pp. 265–281.
[57] Traversa, S., and Ivaldi, E. Measuring the digital transition within the pa:
proposals comparison. In Book of the Short Papers (2022), A. Banzanella,
M. Bini, C. Cavicchia, and R. Verde, Eds., vol. SIS-2022. 51st Scientific Meet-
ing of the Italian Statistical Society, Springer, pp. 827–831.
[58] Twizeyimana, J. D., and Andersson, A. The public value of e-government–a
literature review. Government information quarterly 36, 2 (2019), 167–178.
[59] Ward, J. H. Hierarchical grouping to optimize an objective function. Journal
of the American statistical association 58, 301 (1963), 236–244. ISBN: 0162-
1459 Publisher: Taylor & Francis Group.

You might also like