Wrapping up the Europe 2020 Strategy: a Multidimensional Indicator Analysis

Abstract

The Europe 2020 Strategy was launched by the European Commission in 2010 to promote smart, sustainable, and inclusive growth across EU member states. As the strategy draws to a close in 2020 and is superseded by the Sustainable Development Goals and the Green Deal, this work aims to assess the progress made over the last decade, and to carry forward lessons for future endeavours. A composite indicator approach is adopted, which aggregates the distance of each country or region to politically-agreed targets. This allows a high-level summary of progress, but also examines detailed trends at national and regional levels, as well as by degree of urbanisation and by development. The results show that although the EU has moved forward as whole, some regions have lagged behind or even moved backwards, and within some countries their regions moving further away from one another. Progress has been particularly strong in education, but more work is needed in the environmental dimensions.

Full text

Journal Pre-proof
Wrapping up the Europe 2020 Strategy: a Multidimensional Indicator Analysis
William Becker, Hedvig Norlén, Lewis Dijkstra, Stergios Athanasoglou
PII: S2665-9727(20)30059-3
DOI: https://doi.org/10.1016/j.indic.2020.100075
Reference: INDIC 100075
To appear in: Environmental and Sustainability Indicators
Received Date: 17 April 2020
Revised Date: 29 September 2020
Accepted Date: 2 October 2020
Please cite this article as: Becker, W., Norlén, H., Dijkstra, L., Athanasoglou, S., Wrapping up the
Europe 2020 Strategy: a Multidimensional Indicator Analysis, Environmental and Sustainability
Indicators, https://doi.org/10.1016/j.indic.2020.100075.
This is a PDF file of an article that has undergone enhancements after acceptance, such as the addition
of a cover page and metadata, and formatting for readability, but it is not yet the definitive version of
record. This version will undergo additional copyediting, typesetting and review before it is published
in its final form, but we are providing this version to give early visibility of the article. Please note that,
during the production process, errors may be discovered which could affect the content, and all legal
disclaimers that apply to the journal pertain.
© 2020 The Author(s). Published by Elsevier Inc.

Wrapping up the Europe 2020 Strategy:
a Multidimensional Indicator Analysis
William Becker
1*
, Hedvig Norlén
1
, Lewis Dijkstra
2
and Stergios Athanasoglou
3
1
European Commission, Joint Research Centre, Via E Fermi 2749, Ispra (VA), Italy
2
European Commission, Directorate General for Regional Development and Urban Policy, Brussels, Italy
3
University of Milan - Bicocca, Economics Department, Milan, Italy
*
Corresponding author. Email: william.becker@bluefoxdata.eu
Abstract
The Europe 2020 Strategy was launched by the European Commission in 2010 to promote smart,
sustainable, and inclusive growth across EU member states. As the strategy draws to a close in 2020 and is
superseded by the Sustainable Development Goals and the Green Deal, this work aims to assess the
progress made over the last decade, and to carry forward lessons for future endeavours. A composite
indicator approach is adopted, which aggregates the distance of each country or region to politically-agreed
targets. This allows a high-level summary of progress, but also examines detailed trends at national and
regional levels, as well as by degree of urbanisation and by development. The results show that although
the EU has moved forward as whole, some regions have lagged behind or even moved backwards, and
within some countries their regions moving further away from one another. Progress has been particularly
strong in education, but more work is needed in the environmental dimensions.
Keywords: Europe 2020; composite indicator; sustainable development; regional development
Journal Pre-proof

Wrapping up the Europe 2020 Strategy:
1
a Multidimensional Indicator Analysis
2
Abstract 3
The Europe 2020 Strategy was launched by the European Commission in 2010 to promote smart, 4
sustainable, and inclusive growth across EU member states. As the strategy draws to a close in 2020 and 5
is superseded by the Sustainable Development Goals and the Green Deal, this work aims to assess the 6
progress made over the last decade, and to carry forward lessons for future endeavours. A composite 7
indicator approach is adopted, which aggregates the distance of each country or region to politically-8
agreed targets. This allows a high-level summary of progress, but also examines detailed trends at 9
national and regional levels, as well as by degree of urbanisation and by development. The results show 10
that although the EU has moved forward as whole, some regions have lagged behind or even moved 11
backwards, and within some countries their regions moving further away from one another. Progress 12
has been particularly strong in education, but more work is needed in the environmental dimensions. 13
14
Keywords: Europe 2020; composite indicator; sustainable development; regional analysis 15
1 Introduction 16
1.1 The EU2020 Strategy
17
In 2010, José Manuel Barroso launched the European Commission’s Europe 2020 Strategy in the wake of 18
the 2008 financial crisis, and as a successor to the Lisbon Strategy. The strategy, consisting of 19
measurable targets and underlying proposals, aimed to promote “smart, sustainable, and inclusive” 20
growth across the 28 EU member states
1
[1]. 21
The Europe 2020 Strategy (hereafter ‘EU2020’) identified eight headline targets, accompanied by 22
measurable indicators, to be attained by the end of 2020, involving employment, research and 23
development, climate and energy, education; and social inclusion and poverty reduction—see Table 1. 24
EU Cohesion Policy is also linked to the Europe 2020 strategy, as the former provides the investment 25
framework to meet the defined goals. 26
Policy Area
No.
Objective
Acronym
SDG in EU
monitoring
system
Employment
1
75% of 20
-
64 year
-
olds to be employed
EMP
8
R&D
2
3% of EU GDP to be invested in R&D
R&D
9
1
At the time of writing, the UK was a member of the EU. Since the EU2020 strategy included the UK for almost its
entire duration, and this is a retrospective study, it will be included as a member state in this paper,
Journal Pre-proof

Climate
change and
energy
sustainability
3
Greenhouse gas emissions 20% (or even
30%, if the conditions are right) lower than
1990
GHG
13
4
20% of energy from renewables
REN
7
5
20% increase in energy
efficiency compared
to 2005
EFF
7
Education
6
Reducing the rates of early school leaving
below 10%
ESL
4
7
At least 40% of 30
-
34
–
year
-
olds completing
tertiary education
TERT
4
Fighting
poverty and
social
exclusion
8
At least 20 million fewer people in
or at risk
of poverty and social exclusion
AROPE
1
Table 1: EU2020 Objectives, corresponding policy areas and acronyms. SDG numbers refer to the use of these indicators in
27
the official EU Sustainable Development Goals monitoring system.
28
The EU-level targets in each indicator were also complimented by national targets, which are adaptions 29
of EU-level targets that are realistically attainable given each member state’s particular circumstances. 30
In general, countries that are further from EU targets have lower national targets, and vice versa. 31
Indeed, some countries have actually set targets higher than the EU values (see Appendix for more 32
details). 33
Since 2010, the EU2020 strategy has been one of the guiding stars behind European policy-making. In 34
recent years, it has been complemented and perhaps overshadowed by the Sustainable Development 35
Goals (SDGs). The SDGs are a set of 17 global goals designed to achieve a ‘better and more sustainable 36
future for everybody’, which were adopted in a UN General Assembly in 2015 and to be achieved by the 37
year 2030 [2]. However, the SDGs are based on many of the same principles as the EU2020 strategy, 38
including reducing poverty, improving education, and tackling climate change. 39
The SDGs were accompanied by an official global set of 231 indicators which were agreed on in a UN 40
resolution [3]; since then the number has increased to 247, including some repetitions. These indicators 41
were intended to be a basis for regional and national systems of monitoring. However, many have poor 42
data coverage, and in practice, national and regional statistical offices have adopted a streamlined 43
subset, with modifications appropriate to the national/regional context. The EU has its own set of SDG 44
indicators which are reported on an annual basis [4], and these can be directly mapped onto the EU2020 45
indicators (see again Table 1). The EU SDG indicators include seven of the eight EU2020 indicators. 46
More recently, the EU announced its “Green Deal”, which aims for a carbon-neutral society by 2050 [5]. 47
While this focuses on the environmental dimension of sustainable development, it also aims for a “just 48
and inclusive transition”, and to generate jobs by supporting industry and the green economy. 49
Therefore, both the Green Deal and the SDGs have considerable overlap with the preceding EU2020 50
strategy. 51
The EU2020 strategy, the SDGs and the Green Deal all have a notion of inclusiveness at their core, 52
although the meaning differs to some extent. The EU2020 strategy aimed for “inclusive growth, 53
Journal Pre-proof

fostering a high-employment economy delivering social and territorial cohesion” [1]. This implies growth 54
for all regions of the EU, as well as pursuing gender equality and equality for older citizens, although 55
none of the indicators explicitly measure these concepts (they can however be disaggregated in some 56
cases). The SDGs, on the other hand, put the concept of “leaving no one behind” at the heart of its 57
agenda, and two of the SDGs explicitly target these concepts—SDG5 (gender equality) and SDG 10 58
(reduced inequalities). The inclusive element of the Green Deal is a little harder to pinpoint, but includes 59
a “Just Transition Mechanism” which aims to support transitions in carbon-intensive regions with 60
greater socioeconomic challenges [5]. 61
With the ten-year period of the EU2020 strategy drawing to a close (at the time of writing, in early 62
2020), it is interesting to take stock of the progress made towards the targets set in 2010. Despite being 63
superseded by the SDGs and the Green Deal, the strategy is still highly relevant since it is used for 64
setting the targets in assessing sustainable development within the European Union [6]. But the picture 65
is complex: the situation varies considerably between one country and the next, even more so at the 66
regional level. Has growth really been “inclusive”, as the strategy set out to achieve? Where does more 67
work need to be done, and what lessons can be learned, looking forward to the SDG targets of 2030? 68
Tools already exist to monitor progress towards EU2020 targets at the national level - see for example 69
[7], [8] covering EU countries and [9] which additionally includes candidate member states. Extensions 70
to these indexes have investigated progress relative to country groups, targets and best performers in 71
[10], and in [11] the effects of institutions on progress towards EU2020 targets was analysed. Notably, 72
however, no studies can be found in the literature which examine progress towards EU2020 targets at 73
the sub-national level. As this study demonstrates, there can be a huge heterogeneity between regions 74
inside the same country, and this information is crucial in helping direct investment and building 75
effective cohesion policies. 76
This paper aims to examine the progress made towards EU2020 targets using composite indicator 77
approach. The “EU2020 Index” summarises progress towards EU2020 targets at the national and 78
regional (NUTS2) levels, as well as at examining trends with respect to degrees of development and 79
urbanisation. It is calculated covering the whole period of the EU2020 strategy, from 2010-2018, with 80
later years still unavailable due to data limitations. 81
1.2 Composite Indicators
82
Composite indicators are aggregations of measurable indicators which aim to quantify concepts that are 83
not directly measurable, such as financial secrecy [12], human development [13], or climate hazards 84
[14]. They allow the identification of overall trends, and have the advantage of being easily conveyed to 85
non-specialists, such as policy-makers and the general public, in the form of rankings and maps which 86
facilitate easy comparisons of countries, regions or institutions. 87
Due to their simplicity, composite indicators feature in the media and can have a strong impact on 88
public perception and policy [15]. The number of composite indicators has increased rapidly in recent 89
years—partial inventories can be found in [16] and [17]. However, composite indicators are mainly 90
Journal Pre-proof

communication tools, or can be considered as “access points” to a hierarchical set of indicator data, 91
rather than decision-making tools in their own right. 92
The construction of composite indicators is a delicate process because it involves a number of subjective 93
decisions—this means that scores and rankings are not objective but are dependent on the assumptions 94
made in the construction of the index. Composite indicator guidelines stress that the assumptions and 95
uncertainties used in their construction should always be made clear, with ideally all data and 96
processing being made publicly available [18]. 97
Since the EU2020 strategy is built around headline indicators, a composite indicator approach is a 98
natural way to summarise overall progress. As this work shows, it allows a relatively simple starting 99
point for making sense of a complex spatiotemporal data set. The present EU2020 Index is built to 100
follow best practices is composite indicator development, and includes a discussion and quantification 101
of uncertainty. 102
The approach here is to use a straightforward methodology that can be easily applied and understood. 103
The methodology is in a similar vein to the Regional Lisbon Index by DG REGIO [19], which was designed 104
to measure regional performance in meeting the goals set forth by the 2000 Lisbon Treaty. In their 105
index, Regional performance in a particular indicator was measured via the ratio of its distance to the 106
target over the maximum such distance across all regions. The Lisbon Index was then calculated as the 107
simple average of performance across indicators and particular attention was paid to intuitiveness and 108
consistency. 109
The EU2020 Index follows a similar approach: the progress of each geographical entity towards meeting 110
an individual goal is measured via the (appropriately normalised) distance between the value of the 111
respective indicator and its target. Subsequently, the Europe 2020 Index score is calculated by 112
considering a weighted arithmetic average of these percentage shortfalls over the set of all indicators. 113
This paper is structured as follows. In Section 2 the framework that is common to all four versions of the 114
EU2020 Index (national, regional, and by degrees of development and urbanisation) is described, 115
including indicators, data, targets and construction. Sections 3 to 6 then present the results, along with 116
specific methodological notes for each version of the index. Finally, Section 7 gives an overall discussion 117
and conclusions of the work. 118
2 A General Framework for the EU2020 Index 119
This section presents the underlying general framework of the EU2020 Index. In the following, acronyms 120
are used to denote each indicator as shown in Table 2. The four versions of the EU2020 index will also 121
hereafter be referred to as the “national”, “regional” (at NUTS 2
2
level), “DDev” (by degree of 122
development) and “DUrb” (by degree of urbanisation) indexes, for the sake of conciseness. 123
2
The current NUTS 2016 classification valid from 2018 has been used.
Journal Pre-proof

1.
Percentage of 20
-
64 year
-
olds in employment (EMP)
2. Percentage of GDP invested in R&D (R&D)
3. Greenhouse gas emissions, percentage of 2005 value (GHG)
4. Percentage of energy from renewable sources (REN)
5. Percentage
primary
energy
consumption
compared to 2005 (EFF)
6. Percentage 18
-
24 year olds with at most lower secondary education (ESL)
7. Percentage of 30
-
34 year
-
olds with completed tertiary education (TERT)
8. Percentage population at risk of poverty or social exclusion (AROPE)
Table 2: EU2020 indicators and abbreviations
124
2.1 Indicators and Data Availability
125
The availability of data differs from one indicator to the next: this is summarised in Table 3, as well as 126
the data sources used. The table uses a simple rating system to rate the completeness of the data set. 127
“Complete” means that there are no missing data entries from 2010 up to the year stated. “Good” 128
means the same, but that there are some few exceptions. “Fair” means that there are a more 129
substantial number of exceptions. This is meant as an overview to outline the completeness of the data 130
underpinning each version of the index; greater details are given in the following sections. 131
132
Data Availability
Indicator Source National Regional DUrb DDev
EMP Eurostat To 2018, complete To 2018, good
To 2018, good To 2018, good
R&D Eurostat To 2017, complete To 2016, fair No data To 2016, fair
GHG Eurostat, EEA
3
To 2018,2017, good
No data No data No data
REN Eurostat To 2017, complete No data No data No data
EFF Eurostat To 2018, complete No data No data No data
ESL Eurostat To 2018, complete To 2018, good
To 2018, good To 2018, good
TERT Eurostat To 2018, complete To 2018, good
To 2018, good To 2018, good
AROPE Eurostat To 2018, complete To 2018, fair To 2018, good To 2018, fair
Table 3: Data availability and sources of EU2020 indicators
133
From the table it is evident that the national index has very good data coverage in all eight indicators, 134
whereas the other indexes have poorer data availability. The regional and DDev indices both have five 135
indicators available, but lack all three of the environmental indicators (arguably, these are anyway less 136
meaningful at the sub-national level). The DUrb index additionally has no data on the R&D indicator. In 137
terms of data quality, the DUrb index has quite good coverage in its available indicators, while the 138
regional and by DDev indexes have good to fair coverage. This should be kept in mind when drawing 139
conclusions from the results. 140
The GHG data is a special case in that the measurement of emissions is divided into two categories: 141
Emissions Trading Scheme (ETS) emissions, and non-ETS “Emissions Sharing Directive” (ESD) emissions. 142
3
European Environment Agency
Journal Pre-proof

ETS emissions data is taken from the European Environment Agency [20] and excludes emissions from 143
aviation since they are only available from 2012 onwards, however they only account for around 3% of 144
the total (considering direct emissions only). ESD emissions are taken from Eurostat. The ETS and ESD 145
indicators are combined by taking the weighted distance of each to its respective target, with weightings 146
of 45% and 55% respectively, representing the proportional contributions of each to the EU total 147
emissions. 148
Where data was missing for a given region, a sensible imputation procedure was adopted. If data was 149
not available for a particular year, the latest available year was used. In some particular regions, data 150
was not available at the NUTS2 level, and this was proxied by NUTS1 data. 151
Some indicators were also adjusted to account for time effects. Table 4 summarises these adjustments: 152
ESL and TERT both use a three-year average centred on the index year, whereas AROPE uses one year 153
ahead of the index year. For instance, the EU2020 Index for year 2016 would use 2015-2017 averages 154
for education (ESL and TERT) data, 2017 poverty and social exclusion data, and all other indicator data 155
from 2016. The consideration of a three year moving average for ESL and TERT was pursued in light of 156
many regions’ small sample sizes for these indicators. The one-year look-ahead convention for AROPE 157
was adopted to accommodate the temporal structure of the EU-SILC survey from which these data are 158
drawn. 159
Year
EU 2020 Index X
EMP X
R&D X
GHG X
REN X
EFF X
ESL Average of {X-1, X, X+1}
TERT Average of {X-1, X, X+1}
AROPE X+1
Table 4: Construction of Europe 2020 index for a year X
160
2.2 Targets
161
Each of the EU2020 indicators comes with an overall EU-level target, as given in Table 1. However, there 162
also exist national targets which were created to accommodate the heterogeneity of EU28 countries 163
[21]. This was done for a majority of country-indicator pairs (by the member countries themselves), but 164
not for all. Notably, the UK lacks employment, R&D, education, and poverty targets, while a handful of 165
other countries lack targets in R&D and poverty reduction. In general, Member States selected lower 166
national targets when the distance to the EU target was great. Only the Nordic Member States, Austria 167
and the Netherlands set most targets higher. Nevertheless, the distance to national targets remained 168
higher for the member states far removed from the EU targets, than for the ones close to them. Note 169
that no targets are available at the regional level. 170
The energy efficiency indicator is a special case with a more complex target. Whereas other indicators 171
have percentage targets which can be easily applied to individual member states, the energy efficiency 172
Journal Pre-proof

target is relative to a projection made in 2007: it requires a 20% reduction in primary energy 173
consumption (PEC) from 2005 values relative to a “business as usual” projection with no targeted policy 174
measures in place [22]. At the EU level, this translates to an actual reduction of 13.4% compared to 2005 175
values. However, this percentage cannot be applied to all member states since each has a different 176
projection associated with it. As a result, individual target values are used for each country, obtained 177
from Eurostat data tables. This is therefore the only indicator with individual targets for each country, 178
even in the index constructed with EU targets. 179
In the large majority of this paper, the EU-level targets are used, to allow a clear comparison between 180
countries. Arguably, even though countries started at very different points in 2010, Europeans should be 181
entitled to the same standards in these key issues. However, some comparison is also given with 182
national targets with the national-level index. For more details on national targets, including imputation 183
issues, please see the Appendix. 184
2.3 Constructing the Index
185
In this section, we describe the methodology we used to calculate the index. Specifically, we provide a 186
brief description of its mathematical structure and discuss issues related to outlier treatment. 187
2.3.1 Normalisation 188
Consider a region (which may also denote a country in the national index) and a set of  indicators 189 





. For each indicator, define the constant 

to equal 1 if higher values correspond to better 190
performance, and -1 if they correspond to worse performance. The variables 

and 

denote region 191 ’s target and performance with respect to the ith indicator. The set of targets for a region  is denoted 192
by the -dimensional vector 







, and its actual performance in all indicators by the 193
vector 





 194
Focusing on indicator , the variable 

denotes the distance between a region ’s performance relative 195
to its target (with no extra “points” awarded if the target is met): 196



















(
1
)
The distances 

  are now normalised by the 95
th
percentile of distances over all R regions 197
in indicator , denoted as 



. The 95
th
percentile is used rather than the maximum distance to 198
adjust for the effect of outliers. This normalised value

is therefore defined as: 199




!
















"

(2)
where the subtraction from one is used to ensure that smaller distances to the target score higher. The 200
above quantity ranges between a minimum of 0, if region  has the greatest distance-to-target with 201
respect to indicator , and a maximum of 1, if it meets or exceeds the target. Clearly, higher values imply 202
better performance. 203
Importantly, the normalisation of distances to target (i.e., the denominator in the expression of

) was 204
done by considering the all distances in that indicator over all available years (2010-2017) and with 205
Journal Pre-proof

respect to both national and EU-level targets. This ensures comparability over time and over both types 206
of target. 207
2.3.2 Weights and aggregation 208
Suppose now that each indicator  is assigned a weight of #

$, such that %#



&
. This is the 209
standard approach to aggregating variables in a composite indicator [18]. Taking a weighted arithmetic 210
average over the set of all indicators yields the total performance

#, of region : 211
'(
))




#


*
#




&


(3)
This quantity is bounded below by 0 and above by 1, with higher values implying better performance. 212
Given a set of weights, the quantity

# is therefore the EU2020 Index value for a region r. This 213
approach is reminiscent of (though distinct from) scholarly contributions in the measurement of 214
different multidimensional phenomena involving thresholds and cut-off points, such as poverty [23]. 215
The weights used in all versions of the EU2020 index are derived from the idea of equal weighting over 216
policy areas. Looking back at Table 1, one can see that some policy areas have one associated indicator, 217
whereas others have two or three. Each policy area is weighted equally in the index, so that if a policy 218
area contains more than one indicator, the weight of that policy area is equally divided between its 219
indicators. To clarify via an example, for the NUTS2 regional index, the available indicators were EMP, 220
R&D, ESL, TERT, and AROPE. To reflect balance across policy areas, the component scores of indicators 221
EMP, R&D and AROPE were assigned weights of 0.25 each (given that each uniquely represents a policy 222
area), while a weight of 0.125 was assigned each to ESL and TERT (which both contribute to the same 223
policy area). An alternative weighing scheme using equal weights for all indicators is investigated in 224
Appendix B, and the impacts of this methodological choice are noted in the conclusions of the study. 225
2.4 Uncertainty
226
Composite scores and rankings are by their nature uncertain, because they involve necessarily 227
subjective choices regarding the methodology of their construction. That said, the EU2020 index is in 228
many ways less uncertain than most composite indicators, because the starting point is a set of 229
politically-agreed indicators and targets. Thus, there is no ambiguity in which indicators to include or 230
exclude, or how to structure the index. 231
Nevertheless, a number of methodological uncertainties remain in the construction and weighting of the 232
index. To explore this uncertainty, a simple sensitivity analysis is performed in Appendix B, investigating 233
two key choices: an alternative weighting scenario using equal weights across indicators, and the effect 234
of excluding EFF (which is based on somewhat unclear targets). The results show that the weighting 235
does have a significant impact on the results, although top and bottom-ranked countries remain fairly 236
stable. Excluding EFF, on the other hand, has only a very minor impact on the results. 237
Additionally, in a separate study a full global sensitivity analysis was performed, along the lines 238
discussed in [24]. A Monte Carlo analysis was applied which randomly explored the effect of changing 239
Journal Pre-proof

weights within plausible bounds (+/- 20% in this case), and assigning different five different possible 240
aggregation methods. This analysis generated 1500 alternative rankings. 241
At the regional level, it was found that nominal EU2020 scores are always within the 95% confidence 242
intervals established by the 1500 Monte Carlo rankings. Of the 250+ NUTS2 regions, most have fairly 243
narrow confidence intervals, within 10 places (of 281), which implies a relatively robust ranking. The full 244
details of this latter uncertainty and sensitivity analysis can be found in a European Commission 245
technical report on the EU2020 Index in [25]. 246
3 National-level analysis 247
The national index is the most complete in terms of representing EU2020 objectives, consisting of all 248
eight indicators, all of which have good data coverage. Exceptions are R&D and REN, where data are 249
missing for 2018, Croatia has no AROPE data for 2009 and France has no R&D data for 2017. 250
Furthermore, some GHG data is missing for some years for Bulgaria, Croatia, and Romania. Imputed 251
values were taken from the nearest known year. 252
The national index is the only version of the EU2020 index which includes the GHG and EFF indicators. 253
The former is actually a weighted distance to targets of ETS and ESD emissions, as described in Section 254
2.1. Notably, individual ETS and ESD distances to targets are not allowed to be negative in the weighted 255
average, so surpassing the target in one will not compensate for a shortfall in the other. 256
257
Figure 1: National EU2020 Index, showing changes from 2010 to 2018, using EU targets and sorted by 2018 values
258
The results at the national level are shown in Figure 1. Scores are taken with respect to EU targets, as 259
opposed to national targets: this allows a clearer comparison between countries. In these figures, a 260
value of 100 means that a country has reached or surpassed all of the EU2020 targets. 261
The overall picture is familiar: while no single country has met all EU-level targets, Scandinavian 262
countries such as Denmark, Finland and Sweden have the highest scores, while southern and eastern 263
0
10
20
30
40
50
60
70
80
90
100
EU28 DK FI SE AT DE SI CZ NL BE UK PL FR EE HU SK PT LU LT IE LV CY MT HR ES BG EL IT RO
2018 2010
Journal Pre-proof

countries, such as Greece, Italy and Romania, have the lowest. Both Greece and Italy have low 264
employment rates (59.5% and 63% respectively in 2018), but Greece also has high poverty levels (31.8% 265
in 2018), whereas Italy has particularly low scores on tertiary graduates (27.8% in 2018) and a relatively 266
large number of early school leavers (14.5% in 2018). 267
Of particular interest is how countries have progressed over 2010-2018: while all countries have made 268
overall progress, some have progressed relatively little. One such case is Sweden, but this is because it 269
had anyway reached most EU targets in 2010 (except REN, GHG and EFF) and in 2018 was only short on 270
GHG and EFF. Luxembourg also stands out in this respect, with very little progress over the 9-year 271
period. While it had already achieved ESL and TERT targets in 2010, its R&D expenditure has dropped 272
(from 1.62% in 2010 to 1.26% in 2017), and the percentage of people at risk of poverty or social 273
exclusion has actually increased from 17.8% to 21.5% over the same period. This fact is perhaps 274
particularly notable given that Luxembourg has the highest GDP per capita of any EU member state. 275
In contrast, Hungary, Latvia and Poland have shown dramatic progress (46, 38 and 37 points 276
respectively), with notable improvements in employment rate and poverty reduction for all three of 277
these countries. The EU as a whole has also made a 28 point increase since 2010. 278
How is the distance to EU2020 targets linked to national wealth? Figure 2 shows plots of EU2020 index 279
scores in 2018 against GDP per capita (GDPpc). This shows that EU2020 scores are indeed closely linked 280
to wealth (correlation = 0.66, +,), with the exception of two outliers: Ireland and Luxembourg. 281
These are the top two countries in GDPpc, but only have middling ranks on the EU2020 Index. This 282
indicates that these two countries could get more for their money. On the other hand, countries such as 283
Poland, Czechia and Slovenia do relatively well, being well above the average score given their GDPpc. 284
285
EU28
AT
BE
BG
CY
CZ
DE
DK
EE
EL
ES
FI
FR
HR
HU
IT
LT
LV
MT
NL
PL
PT
RO
SE
SI
SK
UK
IE
LU
30
40
50
60
70
80
90
100
0 10000 20000 30000 40000 50000 60000 70000 80000 90000
EU 2020 Score (wrt EU targets)
Real GDP per capita (EUR/capita)
Journal Pre-proof

Figure 2: EU2020 Index scores with respect to EU targets, against real GDP per capita in 2018. Trend line excludes Ireland and
286
Luxembourg.
287
288
Figure 3: National EU2020 Index using national targets, sorted by 2018 values
289
290
Finally, although the focus of this work is on EU-level targets, it is helpful to compare results with those 291
obtained using national targets: this is shown in Figure 3. Here the picture is somewhat different - 292
Czechia has the highest score of 96.5, having achieved its national targets in five of eight indicators 293
(except early school leavers, energy efficiency and greenhouse gas emissions). At the other end, 294
Luxembourg again stands out, this time because it is the furthest from its national targets of any EU 295
country, and has only achieved one of its seven EU2020 national targets. 296
Particularly significant differences between the scores with respect to national and EU targets are 297
Croatia, and Latvia (which are respectively 20 and 15 places lower in the EU-targets index than the 298
national-targets index), and France and Finland (which are respectively 12 and 14 places higher). In the 299
case of Croatia, this difference is due to the fact that it has a particularly low national target for 300
employment rate (62.9%, the lowest in the EU), as well as lower (or higher in the case of AROPE and 301
GHG) national targets in all other indicators except for early school leavers. On the other hand, France 302
and Finland have equal or more ambitious targets in all indicators. 303
4 Regional differences 304
National scores represent the average state of progress in each country, but this can obscure significant 305
and important differences at the sub-national level. The regional EU2020 index examines progress at the 306
NUTS2 level (being the most granular data available), in all 281 regions in member states of the EU. At 307
this level of resolution, the data availability is poorer and GHG, EFF and REN indicators are not present in 308
the index. This means that the index does not include any of the environmental objectives. Of the 309
indicators that are available, R&D and AROPE have some significant gaps in data coverage. At the time of 310
30
40
50
60
70
80
90
100
EU28 CZ PL HR DK LV HU SK AT SI DE SE UK LT NL IT FI BG RO BE PT CY IE EE FR EL ES MT LU
2018 2010
Journal Pre-proof

writing, 2018 data for R&D was not available at all, and so was proxied using 2016 data. AROPE data had 311
slightly better coverage, but still around half of countries, including Belgium, Germany, France and the 312
UK, had only national-level data. The other three indicators (EMP, ESL and TERT) all had fairly good 313
coverage, with an occasional imputation necessary. 314
Figure 4 shows the regional index results in 2018 with respect to EU targets. It shows that the best-315
performing regions in the EU are the capital regions of Sweden
4
, Finland and two regions
5
in Germany, 316
while the worst performing regions are in Spain, Italy, Bulgaria and Romania. Italy in particular has 317
several regions with scores below 20. 318
The capital region tends to be the one of the highest scoring regions in each country, in particular for 319
East-European countries such as Romania and Bulgaria, whose capital regions are comparable with the 320
national average of Germany or the UK. However, a number of West-European countries have a capital 321
region with a score which is only around the national average, and Belgium stands out in that its capital 322
region is one of the lowest-scoring in the country. 323
324
Figure 4: Regional EU2020 index 2018 with respect to EU targets.
325
Note: NUTS2 regions are illustrated as blue circles), capital regions as red circles) and national values as black lines. The countries are sorted by
326
national scores (using the same limited set of indicators used at the regional level).
327
It is also clear that some countries have a considerably greater spread in EU2020 scores across their 328
regions than others, and this can also be a source of feelings of discontent and unfairness in certain 329
regions. Italy has the highest variance in EU2020 score over its regions of any EU country, and also the 330
widest range, with scores ranging from only 4.5 in Sicily to 83.8 in Emilia Romagna. Spain, France and 331
Hungary also stand out in this respect, but to a lesser degree. 332
Examining differences in scores at the regional level from 2010-2018, the large majority (88%) of NUTS2 333
regions have made positive progress – see Figure 5. Particularly large strides have been made by regions 334
4
Two other regions in Sweden score the maximum value of 100, East-Central Sweden and West Sweden.
5
Stuttgart and Upper Bavaria
0
20
40
60
80
100
EU SE DK AT FI DE SI NL BE FR CZ UK LU IE PL EE HU PT SK CY LT MT LV ES HR BG EL IT RO
EU2020 Score in 2018
National
Capital region
NUTS2 regions
Journal Pre-proof

in Hungary, Malta and Latvia, which have seen increases of 30+ points in EU2020 scores. At the other 335
end, however, a number of regions have gone backwards, mainly in Northern Europe, including regions 336
in Sweden, Germany and Finland. As Figure 5 shows, regions with negative progress include those with 337
relatively high scores (in Sweden and Germany), but also those with low scores, including regions in 338
France (Guadeloupe), Greece (both North and South Aegean), Spain (Ceuta) and Italy (Sardinia). These 339
latter regions, many of which are in isolated or periphery regions of the EU, should be of particular 340
concern to policy makers. 341
342
Figure 5: Changes in regional scores over 2010-2018 plotted against 2018 scores, for all 281 NUTS2 343
regions 344
The regional data allows a further interesting analysis to see the extent to which NUTS2 regions have 345
converged over the time period considered (2010-2018). Here, divergence is defined as the increase in 346
the dispersion (which is measured by the variance) of NUTS2 regions within each member state. For 347
example, if the variance of EU2020 scores has increased over the period 2010-2018, this would indicate 348
that regions have moved further apart from one another, i.e. diverged. Figure 6 shows this definition of 349
divergence, plotted against the change in the national level EU2020 score (using the limited set of 350
indicators used for the regional index) over the same period. This shows to what extent member states 351
are experiencing divergence or convergence (internally), and furthermore whether they are converging 352
to higher values or to lower values. 353
The results show that all countries (except Estonia, Latvia, Luxembourg, Cyprus and Malta, which do not 354
have a divergence score since they have only one NUTS2 region) have experienced an overall increase in 355
-20
-10
0
10
20
30
40
50
0 10 20 30 40 50 60 70 80 90 100
Change over 2010-2018
2018 Score
Ceuta Sardinia N. Aegean GuadeloupeS. Aegean S. Sweden
Chemnitz
Małopolskie
Észak-Alföld Dél-Alföld
North-central Bulgaria
Malta
NW Romania
Journal Pre-proof

EU2020 score, with respect to the indicators used in the regional index. However, around half of 356
member states have actually experienced internal divergence, which means that regions have become 357
further away from each other in terms of EU2020 targets. In particular, the regions of Italy, Romania and 358
Bulgaria have diverged significantly, whereas Slovakia and Slovenia show the greatest convergence. 359
360
361
Figure 6: Divergence (measured as the change in the variance over regions) against change in national score, over the period
362
2010-2018. Countries with only one NUTS2 region (Cyprus, Estonia, Latvia, Luxembourg and Malta) are not shown here since
363
divergence cannot be measured.
364
5 The effect of development 365
EU Cohesion Policy targets all regions and cities within the EU and has the objective to improve the 366
economic, social and territorial cohesion throughout the whole union [26]. EU regional policy goes back 367
to the beginning of the European Communities, in 1957 and the Treaty of Rome, where regional 368
differences were mentioned [27]. Cohesion Policy is in fact the EU’s main investment policy and 369
complements other policies concerning e.g. agriculture, education, employment, energy, environment, 370
single market, research and innovation. Almost one third of the total EU budget has been allocated to 371
Cohesion Policy for this programming period (2014-2020), 351.8 billion Euro. EU Cohesion Policy goes 372
hand in hand with the Europe 2020 strategy, as the former provides the investment framework to meet 373
the defined goals. 374
Regions are categorised according to their GDP in three categories, as “more developed”, “transition” or 375
“less developed”. Depending on the category, the Cohesion Policy funds can provide from half up to 376
85% of the total financing of a project. Most of the Cohesion Policy funding is dedicated to the less 377
SE DE
FI
FR
EL
UK
ES
IT
DK
SI
BG
NL
RO
CZ
BE
SK
AT
PT
HR
PL
IE LT
HU
-300
-200
-100
0
100
200
300
0 5 10 15 20 25 30 35 40 45
Divergence
Score change
Journal Pre-proof

developed regions. Out of the 281 NUTS 2 regions (v.2016) 73 (26%) are classified as less developed, 50 378
(18%) as transition and the remaining 158 (56%) as more developed. 379
The EU2020 Index by degree of development (DDev) is based on the regional index and uses these 380
development categories to look at how EU2020 scores are related to this definition of development, 381
both at the EU level, and at the national level. 382
To relate the EU2020 scores to development at the EU level, the approach here is to treat all regions 383
that fall into a given development category as a single aggregated region which has its own EU2020 384
score. However, to do this properly, the normalised scores of each region must be weighted for each 385
indicator by the appropriate statistic. For example, for EMP, the (normalised) distance of each region to 386
the target is weighted by the population that are aged 20-64, which ensures that each region 387
contributes to the overall score proportionately to its size in terms of each indicator. The quantities used 388
to weight each indicator are given in Table 5. 389
EMP R&D ESL TERT AROPE
Populat
ion of
region aged 20-64
in 2018
GDP of region in
2017
Population of
region aged 18-24
in 201
8
Population of
region aged 30-34
in 201
8
Total population of
region in 2018
Table 5: Weighting statistics used in calculation of EU2020 DDev index
390
The EU2020 index is defined as the weighted sum of normalised scores

(see Equations 1-3). Consider 391
a development category, “less developed”, and denote LD as the set of regions in the EU which fall into 392
that category. Now the EU2020 index over all less developed regions in the EU is defined as follows: 393
'(
))
-./0

*
#

1

2

*

3



-.
4


&

(4)
where 2

%3
-.
. In other words, the scores for each LD region are summed, weighted by their 394
relevant statistic, 3

, and divided by the sum of all the 3

. Then the EU2020 index is constructed in the 395
normal way, by a weighted sum of the scores in each indicator using the global weights w
i
. This 396
definition naturally extends to the other development categories e.g. by considering MD instead of LD. 397
At the national level the same approach is taken, except now we define e.g. LD
c
as the set of regions in 398
country c which are classed as less developed, and use LD
c
in place of LD in Equation 4. So this gives a 399
combined EU2020 score for all the regions in a given development category, for a given country. 400
At the EU level, the EU2020 scores by degree of development are plotted over time in Figure 7. 401
Evidently, there is an upward trend all three development groups, and more developed regions easily 402
score highest overall. It is also notable that less developed regions are improving more rapidly than 403
transition and more developed regions on the whole: this is likely due in part to receiving a greater 404
share of cohesion funding, and points to the success of these programmes. However, transition regions 405
have actually improved the least, and even experienced a slight decline in 2012 and 2013. This may 406
suggest that transition regions should receive a little more attention from regional policy-making. 407
Journal Pre-proof

408
Figure 7: EU2020 index by degree of development, grouping all EU regions into development categories; left is with respect
409
to national targets; right is with respect to EU targets
410
At the national level, Figure 8 show the EU2020 scores by degree of development in 2018 (top figure), 411
and also the change of each development category over 2010-2018 (bottom figure). The national scores 412
echo the scores at the EU level: in general, the highest scores are found in more developed regions, the 413
lowest in less developed regions, and transition regions are somewhere in between. The only exception 414
to this is in Portugal, where the Algarve region scores lower than the less developed regions, despite 415
being a transition region. 416
Across countries, the scores of in each development category vary significantly and overlap with other 417
categories. For example, transition regions in Italy, Greece and Spain have lower scores than most less-418
developed regions in other countries. The more developed regions of Greece (Attiki and Notio Aigaio) 419
have scores lower than most transition and less developed regions in other parts of the EU. 420
30
40
50
60
70
80
90
2010 2011 2012 2013 2014 2015 2016 2017 2018
Less developed Transition More developed
Journal Pre-proof

421
Figure 8: EU2020 scores with respect to EU targets, by degree of development in 2018 (top) and change over 2010-2018
422
(bottom). Sorted by 2018 overall national scores using regional indicators.
423
Note: Regions are categorised as more developed (MD, blue circles), transition (yellow circles) and less developed (LD, red circles).
424
425
Over the 2010-2018 period, there has been a general increase in EU2020 scores, however, some have 426
remained effectively static, or have decreased slightly. This is not limited to one particular degree of 427
development. The less developed regions of Spain, along with transition regions of Denmark and 428
Germany, and the more developed regions of Slovenia and Luxembourg, have all experienced small 429
decreases in their scores. Although these are small steps backwards, in the context of overall European 430
improvement, these regions should be examined closely. Again, this information could help to direct 431
regional policy from 2020 onwards. 432
0
20
40
60
80
100
EU28 SE DK AT FI DE SI NL BE FR CZ UK LU IE PL EE HU PT SK CY LT MT LV ES HR BG EL IT RO
2018 Score
-10
0
10
20
30
40
EU28 SE DK AT FI DE SI NL BE FR CZ UK LU IE PL EE HU PT SK CY LT MT LV ES HR BG EL IT RO
Change over 2010-2018
Less developed Transition More developed
Journal Pre-proof

6 Degree of urbanisation 433
The final angle of analysis for the EU2020 index is compare progress by degree of urbanisation, using 434
Eurostat data. Indicator data was available at the national level by degree of urbanisation, for four 435
indicators: EMP, ESL, TERT and AROPE. In this data set, there are three degrees of urbanisation: cities, 436
towns and suburbs, and rural areas. The degree of urbanisation classifies local administrative units, 437
which are much smaller than NUTS-2 regions. This data is different from NUTS2 data—i.e. it is not 438
simply NUTS2 regions classified into each degree of urbanisation and aggregated, as with the index by 439
degree of development. Data coverage is relatively good, with only a few entries missing. Some minor 440
imputations were necessary, done by using the nearest year for a small number of countries. 441
442
443
Figure 9: EU2020 index by degree of urbanisation with respect to EU targets in 2018 (top), sorted by total scores, and change
444
over 2010-2018 (bottom)
445
Figure 9 shows the results with respect to EU targets. Overall for the EU, the differences in scores are 446
fairly small, but cities have on average slightly higher scores than towns and suburbs, which in turn score 447
higher than rural areas. At the national level, the differences become greater: as the national score 448
decreases (moving right along Figure 9, top), the difference between degrees of urbanisation increases. 449
Countries such as Romania and Bulgaria, and other Eastern European countries, have particularly high 450
differences between city and rural scores with cities scoring 90 or above in most cases. Italy is an 451
exception in this respect, having similar scores for all degrees of development, which is more 452
characteristic of a high national average such as Sweden or the Netherlands. 453
Slovenia also stands out here in having the highest overall score. This is because the data by degree of 454
urbanisation only includes four indicators (EMP, ESL, TERT and AROPE). Slovenia has achieved the EU 455
targets for all of these indicators, in all of its degrees of urbanisation, and so gets a maximum score. 456
0
20
40
60
80
100
EU28 SI SE NL FI DK AT DE CZ PL UK FR IE SK LU HU EE BE PT CY MT LV LT HR ES BG EL RO IT
2018 score
Series1 Town/sub Rural Total
-20
-10
0
10
20
30
40
50
60
70
EU28 SI SE NL FI DK AT DE CZ PL UK FR IE SK LU HU EE BE PT CY MT LV LT HR ES BG EL RO IT
Change over 2010-2018
Urban Town/suburb Rural Total
Journal Pre-proof

Some countries go against the general trend, by having cities scoring worse than other areas: these are 457
the UK, France, Belgium and Austria. These countries, along with Slovenia, the Netherlands, and 458
Sweden, have the highest levels rural development, according to the available indicator data. 459
Examining the lower chart in Figure 9, it is clear that the large majority of areas have made progress over 460
2010-2018. That said, towns and suburbs in Luxembourg and Cyprus, as well as cities in Belgium, 461
Denmark and Greece, have moved overall further away from EU2020 targets. Cities in Bulgaria, 462
Romania, Croatia and Latvia have made particularly positive progress. At the EU level, rural areas have 463
made the most progress, with towns and suburbs and cities making slightly less. On the other hand, the 464
average score change (averaged across countries, for each degree of urbanisation) paints a slightly 465
different picture: the average score change for cities is +18, +17 for towns and suburbs, and +15 for rural 466
areas. This is likely due to the different sizes of urbanisation categories in different countries. 467
7 Discussion and Conclusions 468
Human development and sustainable development are complex multidimensional topics that cannot be 469
perfectly encapsulated within an indicator framework. Nevertheless, the EU set clear and measurable 470
targets to be achieved over 2010-2020, and this allows a fairly clear measurement of progress to these 471
specific targets. Even this simplified picture is rather complex, and can be viewed from many angles and 472
levels of disaggregation. 473
What progress has been made over the period of the EU2020 strategy, and what lessons might be 474
learned? Overall, there has been clear and tangible progress in each dimension: at the EU level, 475
employment rates have risen, R&D spending and renewable energy have increased, emissions are down, 476
there are more tertiary graduates, and the numbers of early school leavers and people at risk of poverty 477
or social exclusion have dropped. However, only one of these dimensions (tertiary graduates) has met 478
the target as of 2018. 479
At the national level, every EU country has made overall progress since 2010, however as of 2018, no 480
country has yet succeeded in meeting all EU-level targets. Sweden and Denmark have met six of seven 481
targets (for which data is available), but should further reduce greenhouse gas emissions to complete 482
the set. 483
While Scandinavian countries unsurprisingly occupy the top places, some countries have made 484
particularly strong progress since 2010: Hungary, Latvia and Poland in particular have made significant 485
improvements in employment rates and poverty reduction, and have seen large positive changes in 486
national-level scores, with some regions showing the greatest increases in the EU. 487
On the other hand, some areas have seen little progress, or even moved backwards. Luxembourg in 488
particular has remained relatively static, and has seen increases in poverty and social exclusion, despite 489
having the highest GDP/capita in the EU. Some more isolated regions of Spain (Ceuta), Italy (Sardinia) 490
and Greece (N. and S. Aegean) have moved backwards, and these are some of the lowest-scoring 491
regions in the EU. This should be of particular concern to policy-makers, all the more so because Italy 492
Journal Pre-proof

and Spain have the largest intra-regional differences of any country, and these differences are actually 493
increasing. Lagging regions contribute to the feeling of unfairness and inequality, and go against the 494
original EU2020 objectives of inclusive growth , and cohesion policy in general: this can in turn lead to 495
populism and distrust in institutions [28]. These countries and areas should be focused on in the next 496
policy cycle. 497
For the specific indicators, the greatest progress seems to have been made in education (tertiary 498
graduates and early school leavers, with 18 of 28 countries having achieved these targets in 2018), but 499
much less has been made in R&D spending, with only four countries reaching the 3% target. Worst of all, 500
no countries have yet met targets (either EU-level or national) in greenhouse gas emissions. Given the 501
pressing need to cut emissions, this is rightly a major focus area of the EU with the inception of the 502
Green Deal in the present decade. 503
How do these findings compare to recent studies on sustainable development in Europe? At the 504
national level, the UN Sustainable Development Solutions Network (SDSN) has produced a European 505
Sustainable Development Report based on their SDG Index and Dashboards [29]. Here, the overall 506
results are very similar: the top five countries in the SDG Index are the same as those in the EU2020 507
national index (with respect to EU targets), while the bottom five include three of the five bottom five in 508
the EU2020 index. They also note that the areas in need of the most attention are those related to 509
responsible consumption and production, climate and biodiversity, which reflects the EU2020 index 510
observations that the EU should focus on environmental targets in particular. This conclusion is also 511
reflected in the EU’s Sustainable Development Monitoring Report, which identifies climate action as the 512
goal with the least progress, apart from gender inequality [4]. The EU study also includes a rich set of 513
information about specific goals and countries, but does not explicitly rank countries. 514
The UN also publishes its own reports on SDG progress [30] – here it is somewhat hard to compare with 515
the EU2020 Index results, because this is a global report. It identifies major challenges in almost all 516
SDGs, and does not focus on national results in particular. 517
While none of these studies mentioned examine regional data, the UN SDSN report does provide an 518
interesting “Leave No One Behind Index”, which is based on indicators of inequality (including poverty, 519
income equality, gender equality and access to services). This draws somewhat similar conclusions to 520
the convergence study presented in this paper, showing that Bulgaria and Romania are of particular 521
concern. However, the EU2020 index particularly flags Italy as a country of concern, which is less evident 522
in the SDSN study – this is due to the specific use of sub-national data here rather than using national-523
level indicators. 524
As a final point, consider that the period of analysis here is 2010-2018, and that particular events can 525
alter or even reverse the trends observed in this study. In particular, the COVID crisis of 2020 has altered 526
caused economic upheaval and altered political priorities. However, the impacts on sustainable 527
development will only become clear when sufficient data is available. 528
Journal Pre-proof

References 529
[1] European Commission, Europe 2020: A Strategy for Smart, Sustainable and Inclusive Growth: 530
Communication from the Commission. Publications Office of the European Union, 2010. 531
[2] United Nations General Assembly, Transforming our world: the 2030 Agenda for Sustainable 532
Development, A/RES/70/1. 2015. 533
[3] United Nations General Assembly, ‘Work of the Statistical Commission pertaining to the 2030 534
Agenda for Sustainable Development, A/RES/71/313’. 2017. 535
[4] European Commission, ‘Sustainable development in the European Union: Monitoring report on 536
progress towards the SDGs in an EU context’. 2020, [Online]. Available: 537
https://ec.europa.eu/eurostat/documents/3217494/11011074/KS-02-20-202-EN-N.pdf/334a8cfe-538
636a-bb8a-294a-73a052882f7f. 539
[5] European Commission, ‘Communication from the Commission to the European Parliament, the 540
European Council, the Council, the European Economic and Social Committee and the Committee 541
of the Regions: The European Green Deal (COM/2019/640 final)’. Publications Office of the 542
European Union, Dec. 11, 2019, [Online]. Available: https://eur-lex.europa.eu/legal-543
content/EN/TXT/?uri=COM:2019:640:FIN. 544
[6] Statistical Office of the European Communities, Sustainable development in the European Union: 545
overview of progress towards the SDGs in an EU context : 2019 edition. 2019. 546
[7] P. Pasimeni, ‘The Europe 2020 Index’, Social Indicators Research, vol. 110, no. 2, pp. 613–635, Jan. 547
2013, doi: 10.1007/s11205-011-9948-9. 548
[8] G. Rappai, ‘Europe En Route to 2020: A New Way of Evaluating the Overall Fulfillment of the 549
Europe 2020 Strategic Goals’, Social Indicators Research, vol. 129, no. 1, pp. 77–93, Oct. 2016, doi: 550
10.1007/s11205-015-1092-5. 551
[9] M. S. Çolak and A. Ege, ‘An Assessment of EU 2020 Strategy: Too Far to Reach?’, Social Indicators 552
Research, vol. 110, no. 2, pp. 659–680, Jan. 2013, doi: 10.1007/s11205-011-9950-2. 553
[10] B. Walheer, ‘Decomposing the Europe 2020 Index’, Social Indicators Research, vol. 140, no. 3, pp. 554
875–905, Dec. 2018, doi: 10.1007/s11205-017-1797-8. 555
[11] F. Pasimeni and P. Pasimeni, ‘An Institutional Analysis of the Europe 2020 Strategy’, Social 556
Indicators Research, vol. 127, no. 3, pp. 1021–1038, Jul. 2016, doi: 10.1007/s11205-015-1013-7. 557
[12] A. Cobham, P. Janský, and M. Meinzer, ‘The financial secrecy index: Shedding new light on the 558
geography of secrecy’, Economic Geography, vol. 91(3), pp. 281–303, 2015. 559
[13] United Nations Development Programme, ‘Human Development Report 2015’. United Nations 560
Development Programme, 2015, [Online]. Available: 561
http://hdr.undp.org/sites/default/files/2015_human_development_report.pdf. 562
[14] T. Lung, A. Dosio, W. Becker, C. Lavalle, and L. M. Bouwer, ‘Assessing the influence of climate 563
model uncertainty on EU-wide climate change impact indicators’, Climatic Change, vol. 120, no. 1–564
2, pp. 211–227, Jul. 2013, doi: 10.1007/s10584-013-0825-1. 565
[15] J. G. Kelley and B. A. Simmons, ‘Politics by number: indicators as social pressure in international 566
relations’, American Journal of Political Science, vol. 59, no. 1, pp. 55–70, 2015. 567
[16] R. Bandura, ‘A survey of composite indices measuring country performance: 2008 update’, New 568
York: United Nations Development Programme, Office of Development Studies (UNDP/ODS 569
Working Paper), 2008. 570
[17] L. Yang, ‘An Inventory of Composite Measures of Human Progress’, United Nations Development 571
Programme Human Development Report Office, 2014. 572
[18] OECD, Handbook on Constructing Composite Indicators. Methodology and User guide. Paris: OECD, 573
2008. 574
[19] L. Dijkstra, ‘The Regional Lisbon Index’, DG for Regional Policy, European Commission, 2010. 575
Journal Pre-proof

[20] EEA, ‘EU Emissions Trading System (ETS) data viewer — European Environment Agency’, European 576
Environment Agency, 2016. http://www.eea.europa.eu/data-and-maps/data/data-577
viewers/emissions-trading-viewer (accessed May 31, 2016). 578
[21] C. Mayer, Smarter, greener, more inclusive?: indicators to support the Europe 2020 strategy. 579
Luxembourg: Publications Office of the European Union, 2018. 580
[22] European Commission and Statistical Office of the European Union, Smarter, greener, more 581
inclusive?: indicators to support the Europe 2020 strategy : 2019 edition. 2019. 582
[23] S. Alkire and J. Foster, ‘Counting and multidimensional poverty measurement’, Journal of Public 583
Economics, vol. 95, no. 7–8, pp. 476–487, Aug. 2011, doi: 10.1016/j.jpubeco.2010.11.006. 584
[24] M. Saisana, A. Saltelli, and S. Tarantola, ‘Uncertainty and sensitivity analysis techniques as tools for 585
the quality assessment of composite indicators’, Journal of the Royal Statistical Society: Series A 586
(Statistics in Society), vol. 168, no. 2, pp. 307–323, Mar. 2005, doi: 10.1111/j.1467-587
985X.2005.00350.x. 588
[25] S. Athanasoglou, L. Dijkstra, European Commission, and Joint Research Centre, The Europe 2020 589
regional index. Luxembourg: Publications Office, 2014. 590
[26] Official Journal of the European Communities, ‘Single European Act, OJ L 169’. 1987. 591
[27] European Union, ‘Treaty of Rome’. 1957, [Online]. Available: https://eur-lex.europa.eu/legal-592
content/IT/TXT/?uri=celex:11957E/TXT. 593
[28] D. Al Khudhairy, F. Campolongo, S. Langedijk, P. Benczúr, and Europäische Kommission, Eds., What 594
makes a fair society? Insights and evidence. Luxembourg: Publications Office of the European 595
Union, 2017. 596
[29] SDSN & IEEP, ‘The 2019 Europe Sustainable Development Report’, Sustainable Development 597
Solutions Network and Institute for European Environmental Policy: Paris and Brussels, 2019. 598
[30] United Nations Department for Economic and Social Affairs, Sustainable Development Goals 599
Report 2020. S.l.: UNITED NATIONS, 2020. 600
[31] European Parliament, and Council of the European Union, ‘Directive 2009/29/EC of the European 601
Parliament and of the Council’. Official Journal of the European Union, Apr. 23, 2009. 602
[32] A.-D. Barbu et al., Trends and projections in Europe 2015 tracking progress towards Europe’s 603
climate and energy targets. Luxembourg: Publications Office, 2015. 604
605
Appendix A: Targets 606
Table 6 lists the national Europe 2020 targets. All values here were directly obtained from Eurostat. As 607
with GHG data, GHG targets are a specific case because emissions consist of both ETS and ESD 608
contributions. ETS targets cannot be disaggregated to the country level, as emissions that are part of the 609
ETS trading scheme are auctioned and traded EU-wide. Therefore effectively there is only one single EU-610
level target for ETS emissions, which is set at 21% below 2005 levels [31]. The ESD emissions, on the 611
other hand, have national-level targets in the same way as the other indicators—see e.g. [32]. These are 612
given in Table 6. 613
In some cases, national targets were not available for certain countries. This means that either such 614
targets will need to be occasionally imputed, or some countries and/or indicators will need to be 615
omitted from the analysis. In this work, we have predominantly opted for the former option by imputing 616
national targets where they are not available, provided the corresponding EU-28 regional data are not 617
Journal Pre-proof

too sparse. While we have attempted to do so with care, such imputation introduces an unavoidable 618
degree of subjective judgment to the results. 619
AROPE national targets involve numerical goals regarding the reduction of the total number of people at 620
risk of poverty or social exclusion. However, given that the effort to reduce the number of people at risk 621
should be seen in light of the total population of country and its share of population at risk, we 622
transformed the national AROPE Europe 2020 targets into population percentages using 2009 national 623
data on total population 
55
, number of people at risk of poverty 6789'
55
, and the EU2020 target 624
reduction . The first two types of data we obtained from Eurostat, while the third by visiting each 625
country’s individual webpage at the Europe 2020 Commission website.
6
For the sake of analytic 626
precision, the AROPE target of country : expressed as a population percentage, denoted by 6789'
;
, 627
is equal to: 628
6789'
;
6789'
55;

;

55;

Looking at Table 6, none of the listed national targets are available for the United Kingdom (UK). A 629
handful of other countries have either not reported targets for certain objectives (Croatia for AROPE), or 630
have provided targets that are of a different nature than the Europe 2020 figures (the Czech Republic for 631
R&D and Sweden for AROPE). 632
633
Country EMP R&D REN ESL TERT AROPE GHG*
EFF
EU-28 75.0 3.0 20.0 10.0 40.0 19.5 90.0 1483
AT 77.0 3.8 34.0 9.5 38.0 17.6 84.0 31.5
BE 73.2 3.0 13.0 9.5 47.0 17.0 85.0 43.7
BG 76.0 1.5 16.0 11.0 36.0 42.0 120.0 16.9
CY 75.0 0.5 13.0 10.0 46.0 19.8 95.0 2.2
CZ 75.0 1.0 13.0 5.5 32.0 14.9 109.0 39.6
DE 77.0 3.0 18.0 10.0 42.0 19.5 86.0 276.6
DK 80.0 3.0 30.0 10.0 40.0 15.8 80.0 17.4
EE 76.0 3.0 25.0 9.5 40.0 18.0 111.0 6.5
EL 70.0 1.2 18.0 9.7 32.0 23.2 96.0 24.7
ES 74.0 2.0 20.0 15.0 44.0 21.3 90.0 119.8
FI 78.0 4.0 38.0 8.0 42.0 14.3 84.0 35.9
FR 75.0 3.0 23.0 9.5 50.0 15.0 86.0 219.9
HR 62.9 1.4 20.0 4.0 35.0 28.8 111.0 11.15
HU 75.0 1.8 14.7 10.0 34.0 23.3 110.0 24.1
IE 69.0 2.0 16.0 8.0 60.0 22.7 80.0 13.9
IT 67.0 1.5 17.0 16.0 26.0 22.0 87.0 158
6
http://ec.europa.eu/europe2020/europe-2020-in-your-country/index_en.htm
Journal Pre-proof

LT 72.8 1.9 23.0 9.0 48.7 23.6 115.0 6.5
LU 73.0 2.3 11.0 10.0 66.0 13.6 80.0 4.5
LV 73.0 1.5 40.0 10.0 34.0 28.2 117.0 5.4
MT 70.0 2.0 10.0 10.0 33.0 18.3 105.0 0.7
NL 80.0 2.5 14.0 8.0 40.0 14.2 84.0 60.7
PL 71.0 1.7 15.0 4.5 45.0 26.2 114.0 96.4
PT 75.0 2.7 31.0 10.0 40.0 24.2 101.0 22.5
RO 70.0 2.0 24.0 11.3 26.7 42.8 119.0 43
SE 80.0 4.0 49.0 7.0 45.0 13.8 83.0 43.4
SI 75.0 3.0 25.0 5.0 40.0 16.0 104.0 7.3
SK 72.0 1.2 14.0 6.0 40.0 17.5 113.0 16.4
UK 77.1 2.9 15.0 12.3 42.9 19.5 84.0 177.6
634
Table 6: Europe 2020 national and EU-28 targets; units as in Table 2. Imputed targets in grey.
635
*GHG targets represent EDS targets only.
636
When national targets for a particular country-indicator pair were not available, a reasonable estimate, 637
based on the national targets of countries with roughly similar “starting points”, was derived. To 638
illustrate this, take the example of the UK’s TERT target. In 2009 the UK had a TERT of 41.5, which was 639
similar to that of DK (40.7), NL (40.5), LT (40.6), PT (71.2), FR (43.2), and CY (43.9). Using the TERT targets 640
for the latter countries, the average distance of their 2009 rates to their corresponding targets was 641
computed
7
, which was equal to 1.4. This represents an average distance to target for countries with 642
similar TERT starting points to the UK. To impute the UK target, we added to its 2009 value this average 643
distance to target, resulting in a target of 41.5+1.4=42.9. This strategy was applied for all missing values 644
highlighted in Table 6. 645
The EFF targets are complicated and based on a 20% reduction with respect to a “business as usual” 646
projections of 2020 primary energy consumption made in 2007 [22]. Clearly, these projections can be 647
subject to considerable debate, therefore the EFF targets are not considered to be as clear and robust as 648
the other targets. For this reason, the EFF indicator is removed in a sensitivity analysis in Appendix B. 649
A final note on targets is that for the GHG indicators, the national ESD targets and EU ETS targets were 650
both used in the indexes with respect to EU targets and with respect to national targets. This is because 651
there are no national targets available for the ETS emissions, and ESD emissions targets anyway add up 652
to the EU total and are legally binding. By the same argument, the REN national targets are used in both 653
the EU-targets versions of the index, and the national-targets versions. 654
7
Where this distance was negative (as in the case of DK and NL), meaning that a country had already attained its
target in 2009, we truncated it to 0.
Journal Pre-proof

Appendix B: Sensitivity Analysis 655
In this section the EU2020 index is tested against varying some selected assumptions. This amounts to a 656
conceptually simple sensitivity analysis to demonstrate the impact of selected important methodological 657
choices. A more detailed sensitivity analysis is available in [25]. 658
659
Figure 10: EU2020 national index with EU targets, plotted against alternative formulations: with equal weighting for each
660
indicator (left); and with removing EFF (right).
661
Two particular assumptions are tested here. The first is based on weights: the EU2020 index uses equal 662
weighting across the five policy areas (see again Table 1), and since different policy areas have different 663
numbers of indicators, this means that the indicators themselves are not equally weighted. An 664
alternative formulation is to equally weight the indicators themselves. This amounts to increasing the 665
relative weight of the environmental and education policy areas, since both have more than one 666
indicator associated with them. Figure 10 (left) shows the effect of this variation: the alternative 667
weighting has a significant impact on the rankings, with a few countries dropping up to 7 places in the 668
rankings (the Netherlands and France). On the other hand, the top and bottom countries are relatively 669
stable. 670
The second assumption to test is removing EFF. As mentioned previously, EFF has an unusual target 671
system based on projections, and arguably this does not amount to a clear and measurable target. 672
Figure 10 (right) shows the effect of recalculating the index without this indicator. Here, the impact is 673
very minor, with the majority of countries unchanged in the rankings. Poland is a slight exception, 674
dropping three places as a result of this change. 675
Overall, the choice of weighting does have a significant impact on the results of (particularly) mid-ranked 676
countries, and this should be accounted for in the conclusions of the study. Weighting by policy areas 677
represents one possible perspective of progress towards EU2020 targets. 678
EU28
AT
BE
BG CY
CZ
DE
DK
EE
EL ES
FI
FR
HR
HU
IE
IT
LT
LU
LV
MT
NL
PL
PT
RO
SE
SI
SK
UK
0
5
10
15
20
25
30
051015202530
Rank (equal weighting)
Nominal rank
EU28
AT
BE
BG
CY
CZ
DE
DK
EE
EL
ES
FI
FR
HR
HU
IE
IT
LT
LU
LV
MT
NL
PL
PT
RO
SE
SI
SK
UK
0
5
10
15
20
25
30
051015202530
Rank (equal weighting)
Nominal rank
Journal Pre-proof

Declaration of interests
☒
The authors declare that they have no known competing financial interests or personal relationships
that could have appeared to influence the work reported in this paper.
☐The authors declare the following financial interests/personal relationships which may be considered
as potential competing interests:
Journal Pre-proof