Typology of cities · 2015. 8. 12. · achieve the sustainable, energy–efficient, smart city. By coordinating strategies and combining best practices, PLEEC develops a general model

Typology of cities WP 2

Deliverable 2.2

Report created by TUWIEN team

Rudolf Giffinger Gudrun Haindlmaier Florian Strohmayer

January 2014

Report abstract – Typology of cities (D2.2)

Main aim of report

City profiles will reveal the specific opportunities and threats that will confront cities on their innovation journey

towards an energy-efficient development. Deliverable 2.2 analyses city performance on certain key aspects in a

European perspective and in comparison. This typology allows for benchmarking and an easy identification of

comparable cities in order to look for best practices and to gain some input for WP3-5 of the PLEEC project.

Target group

PLEEC cities, administration authorities, PLEEC partners in WP3-5

Main findings/conclusions

• The typology of European cities shows up with three big clusters (comprising 16-25 cities) arranged along

three axes across Europe:

o a first axis runs down from Scandinavia in the North to the South of France

o a second one forms a north-east curve leading to the south-west of Europe

o a third axis running from the West to the East of Europe

• The other three clusters of the typology are smaller ones (consisting of 3-7 cities) and are situated in the

heart of Europe as well as in the outermost East

• Stoke-on-Trent as well as Santiago de Compostela go in the same cluster (no. 5); Turku, Jyväskylä and

Eskilstuna perform similar (cluster 6); Tartu is situated in another cluster (no. 3)

• To conduct a specific energy efficiency cluster analysis, a common data base of sufficient indicators is (still)

needed in Europe

WP7: Dissemination

WP

2:

Sm

art

cit

y p

rofi

les

WP

6:

Syn

erg

y o

f

pe

rsp

ec

tiv

es a

nd

acti

on

pla

n

WP3: Technology driven efficiency

potentials

WP4: Structure driven efficiency

potentials

WP5: Behaviour driven efficiency

potentials

Activities carried out including methodology used

• Multiple correspondence analysis along 6 Smart City key fields for 77 European cities resulting in a typology

of cities

• Guidelines for cities how to use typology

Introduction PLEEC project

Energy efficiency is high on the European agenda. One of the goals of the European Union's 20-20-20 plan is to

improve energy efficiency by 20% in 2020.

However, holistic knowledge about energy efficiency potentials in cities is far from complete. Currently, a variety of

individual strategies and approaches by different stakeholders tackling separate key aspects hinders strategic energy

efficiency planning.

For this reason, the PLEEC project – "Planning for Energy Efficient Cities" – funded by the EU Seventh Framework

Programme uses an integrative approach to achieve the sustainable, energy–efficient, smart city. By coordinating

strategies and combining best practices, PLEEC will develop a general model for energy efficiency and sustainable

city planning.

By connecting scientific excellence and innovative enterprises in the energy sector with ambitious and well-

organized cities, the project aims to reduce energy use in Europe in the near future and will therefore be an

important tool contributing to the EU's 20-20-20 targets.

How to make use of the Smart City typology

1) Identify your city and your cluster

2) Have a look at cluster characteristics

0 = average (all cities)

positive values mean smart performance (the higher the better)

negative values indicate performance below average

Cluster

no.

Smart

Economy

Smart

Environ-

ment

Smart

Governance Smart Living

Smart

Mobility

Smart

People

1 -0,73 -0,84 -0,44 -0,57 -0,92 -1,08

2 -0,44 -0,17 -0,71 -0,67 -0,51 -0,55

3 -0,39 -0,1 -0,29 -0,13 -0,28 -0,45

4 0,68 0,22 0,01 0,88 0,60 0,45

5 0,27 0,02 0,10 0,19 0,26 0,24

6 0,13 0,46 0,65 0,21 0,15 0,62

your city/your cluster

Cluster 1

Cluster 2

Cluster 3

Cluster 4

Cluster 5

Cluster 6

Sources of basemap: TUWIEN 2013 based on Esri, GEBCO, NOAA, National Geographic, DeLorme,

NAVTEQ, Geonames.org and other contributors; EuroGeographics

⇒ Map on page 9

3) Compare your city’s profile with cluster

Compared to an average city of your type: where do you do better than others? What could you improve?

4) Search for others like you

City Cluster no.

…

RZESZOW 3

BYDGOSZCZ 3

GRAZ 4

LINZ 4

SALZBURG 4

LUXEMBOURG 4

INNSBRUCK 5

GENT 5

…

5) Go for best practices!

Cluster 1

Cluster 2

Cluster 3

Cluster 4

Cluster 5

Cluster 6



⇒ Map on page 9 ⇒ city list on

page 16

PLEEC Deliverable 2.2

Typology of cities

Contents

Introduction PLEEC .................................................................................................................................7

Aims WP2/D2.2. Typology of cities ....................................................................................................9

The methodology – clustering cities in Europe ......................................................................................9

General remarks on cluster analysis .............................................................................................10

Multiple correspondence analysis ....................................................................................................10

Results – typology of cities ...............................................................................................................10

The typology of Smart Cities in Europe in a spatial perspective ...................................................11

Cluster profiles .............................................................................................................................14

Energy efficient clustering ................................................................................................................16

Appendix: Smart City list ......................................................................................................................18

Appendix: Methodology .......................................................................................................................19

References on methodology ............................................................................................................19

Multiple Correspondence Analysis – Statistics output .....................................................................19

Introduction PLEEC

As energy efficiency is high on the European agenda (one of the goals of the European Union’s 20-20-

20 plan is to improve energy efficiency by 20% in 2020), the PLEEC project – “Planning for Energy

Efficient Cities” – funded by the EU Seventh Framework Programme uses an integrative approach to

achieve the sustainable, energy–efficient, smart city. By coordinating strategies and combining best

practices, PLEEC develops a general model for energy efficiency and sustainable city planning.

In work package 2 (WP2, see figure on project work plan below) the performance of six model mid-

sized cities on key aspects is analysed. These city profiles (detailed elaboration and discussion see

Deliverable D2.1) will reveal the specific opportunities and threats that will confront cities on their

innovation journey towards an energy-efficient development. Deliverable 2.2 now analyses city

performance on certain key aspects in a European perspective and in comparison. Based on WP2,

three parallel work packages the will elaborate energy efficiency recommendations.

Figure 1: Work packages of PLEEC

Figure 2: Location of analysed European Smart Cites



9

Aims WP2/D2.2. Typology of cities

The hierarchical smart city approach as displayed in the figure below allows on a first glance for

ranking and positioning of cities. To do so, 81 components are aggregated to 28 domains and again

to 6 key fields which then define a city’s position within the European Smart City system. But this

approach not only allows for a “beauty contest” of the final rank, but also for more detailed

evidence-based benchmarking on the level of key fields and domains. In order to facilitate

orientation among many other European cities, a typology of European small and middle sized cities

is established to help the 6 PLEEC model cities to find other cities to compare with. By delivering

clusters of cities, those cities with similar in particular characteristics and similar conditions for

energy efficient urban development can be identified. This allows for the identification of meaningful

best practice examples, which is on the one hand important for further PLEEC work packages and on

the other hand for the cities themselves. Specific strengths and weaknesses as well as (potential)

action fields can be identified for each city (see also “How to make use of this typology”-section in

this report).

Figure 3: Smart city model – a hierarchical approach

The methodology – clustering cities in Europe

To analyze differences among cities and to create a spatial typology there are basically several

procedures or methods to do so. Within the PLEEC project the available and compiled data allows for

a quantitative analysis based on the SC indicators for 77 cities (despite some potential for

improvement of available data on European cities in general). Since the characteristics of the

indicators are to some extent linked to the city size, standardized data is used in the analysis.

However, the overall data situation is to some extent problematic as the different data collection

procedures and standards in the individual countries and cities have to be taken into consideration.

Furthermore, some data is only available at NUTS2 or NUTS0 level. These respective indicators hardly

allow for conclusions about the city in its administrative border and assembled structure but only for

conclusion on a regional or national level. A cluster analysis using SC-indicators is only meaningful

and unbiased if variables are available at city level (e.g. Urban Audit data). Out of these reasons a

typology on European Smart cities can’t be conducted at the lowest level of the components (see

figure above) at this point of the PLEEC project. To overcome these data problems and to provide a

good typology in terms of methodology the 6 key fields have to be used for clustering.

10

General remarks on cluster analysis

In order to differentiate groups within a complex data set and to reveal underlying structures

therein, very often a cluster analysis is conducted. This is an analysis tool to identify groups within a

heterogeneous collection of objects according to their similarity in characteristics (measured

variables). The aim is to detect and bundle subsets of objects as homogeneous as possible. Thereby,

so called proximity or distant functions indicate the similarity or dissimilarity of two objects in

relation to each other (this is checked for every two objects in the dataset). Depending on the level of

measurement of features to describe the objects different distant functions are to be applied. Due to

their similarity, the objects, persons or entities are summarized into groups with largely coincident

characteristics.

Depending on the applied cluster algorithm quite diverging results may occur. Most often single-

linkage (based on shortest distance (nearest neighbour method), enables for identification of

outliers) or ward (objective function is the error sum of squares) procedures are used for clustering.

However, the problem is that cluster analysis occasionally is very sensitive to structures within the

data set without any substantive statement (but, for example depend on missing values and other

methodological artifacts). Therefore, the PLEEC project makes use of the multiple correspondence

analysis to group cities according to their key characteristics.

Multiple correspondence analysis

Multiple correspondence analysis arranges objects or data measurements according to their

similarity or dissimilarity along certain dimensions (see Blasius 2001). It detects and represents

underlying structures in a data set by representing data as points geometrically in a low-dimensional

Euclidean space. MCA is an extension of simple correspondence analysis for more than 2 variables

(see Adbi/Valentin 2007). Originally the MCA has been developed for categorical data, thus it reacts

less to outliers than cluster analysis does. Furthermore it is able to map both variables and

individuals, so allowing the construction of complex visual maps1 for interpretation and therefore

offers the potential of linking both variable-centred and case-centred approaches.

To take Luxemburg as illustrative example: when conducting a cluster analysis, Luxembourg is

always a stand-alone cluster due to its exceptionally high values in the key field of "Smart Economy"

and the fact that it represents a state of its own. This leads to the point that among the other 76

countries a very large cluster remains. This cluster has small significance because it is blurred and

mixed up to a large part along national borders.

Results – typology of cities On the level of 6 key fields (Smart Economy, Smart Living, Smart Environment, Smart Mobility, Smart

People and Smart Government) a multiple correspondence analysis has been conducted. Although

similar to the results of a classical cluster analysis, this method shows up with a (more) stable

calculation and plausible division of European Smart cities into 6 different types of cities (details on

statistics and results see appendix page 17/18).

1 In the indicator matrix approach, associations between variables are uncovered by calculating the chi-square distance

between different categories of the variables and between the individuals (or respondents). These associations are then

represented graphically as "maps", which eases the interpretation of the structures in the data. Oppositions between rows

and columns are then maximized (see http://en.wikipedia.org/wiki/Multiple_correspondence_analysis).

11

The typology of Smart Cities in Europe in a spatial perspective

Cluster

Number of

cities

in cluster

1 3

2 7

3 22

4 4

5 25

6 16

total 77

Figure 4: Spatial distribution of 6 types of Smart cities

As a start, the typology of European cities shows up with three big clusters (comprising 16-25 cities)

arranged along three axes across Europe: a first axis runs down from Scandinavia in the North to the

South of France (cluster 6 marked by a blue diamond), a second one form a north-east curve leading

to the south-west of Europe (cluster 3 indicated by a yellow dot) as well as a third axis running from

the West to the East of Europe (cluster 5 marked by green squares doted in the centre).

The other three clusters of the typology are smaller ones (consisting of 3-7 cities) and are situated in

the heart of Europe (Cluster 4 going by beige squares) as well as in the outermost East (Cluster 1 in

purple cycles und cluster 2 marked by red triangles).

Cluster 1

Cluster 2

Cluster 3

Cluster 4

Cluster 5

Cluster 6



12

Figure 5: Clusters values according to the 6 Smart City key fields

Interpretation/How to read table above:

• 0 = mean of all cities

• Negative values refer to a cluster performance below average in the respective key field;

positive values indicate that this cluster performs higher than the European average city.

• The higher the amount, the greater the deviation from the average value.

Tartu can be found within cluster 3, Stoke-on-Trent as well as Santiago de Compostela go in Cluster 5.

All other PLEEC model cities are located in Cluster 6 (see also detailed list of cities at page 16).

Cluster 1 and 2 manifest values clearly below average in all 6 key fields. In opposite to cities of cluster

1, those belonging to cluster 2 show up with a rather Smart Environment not differentiating much

from the performance of an average European city (iinterestingly, the lowest values in Smart

Governance go along with relatively good values in Smart Environment within this cluster). Also

cluster 3 shows up with a performance slightly below average. Thereby, Smart Economy and Smart

People turn out with the highest need for improvement.

On the other hand, Clusters 4 to 6 feature values above average in all key fields. Cities of cluster 5

orientate very closely to the average European city, while those in cluster 4 and 6 show up with

highlights in certain areas. Particular strengths of cluster 4-cities can be found mainly in Smart

Economy, Smart Living and Smart Mobility; those of cluster 6-cities are to be located in terms of

Smart Environment, Smart Governance und Smart People.

These finding are also evident from the spider net charts shown on next page:

Cluster No. Smart

Economy

Smart

Environment

Smart

Governance

Smart

Living

Smart

Mobility

Smart

People

1 -0,73 -0,84 -0,44 -0,57 -0,92 -1,08

2 -0,44 -0,17 -0,71 -0,67 -0,51 -0,55

3 -0,39 -0,10 -0,29 -0,13 -0,28 -0,45

4 0,68 0,22 0,01 0,88 0,60 0,45

5 0,27 0,02 0,10 0,19 0,26 0,24

6 0,13 0,46 0,65 0,21 0,15 0,62

13

Figure 6: Representation of ideal-typical city profile for each cluster

14

Cluster profiles

The following profiles provide detailed information on the performance for each cluster:

“Bad boys” among the cities.

However, Smart Governance

seems the anchor point (in all

three cities of this cluster this key

fields shows up with relatively

the least low cuts) - possibly

government can be the crucial

point to make a difference, as

functioning structures seem to

be present. Quite clearly this

cluster is missing the people who

could make good use of these

structures.

Craiova, Sibiu, Timisoara

Cluster 2 shows up with

particularly under-average

performance in terms of Smart

governance which is associated

with low values in Smart Living

and lack in Smart People.

Although mobility is also below

average, an notably good

development in the field of

Smart Environment can be

observed (several cities have

even values slightly above

average in this area).

Liepaja, Kaunas, Kosice, Pleven, Ruse, Larisa, Patrai

Overall, there is a slightly below-

average performance to be

observed in this cluster:

(somehow the findings in Cluster

2 are to be continued): even if

Smart Economy and Smart

People (still?) leave a lot to be

desired, the values regarding

Smart environment are quite

decent. Also Smart Living is

found to be reasonably in many

cities of this cluster, whilst Smart

Mobility and Smart Governance

oscillate around the European

average Ancona, Banska Bystrica, Bialystok, Bydgoszcz, Coimbra, Gyor, Kielce, Miskolc, Nitra, Oviedo, Padova, Pecs,

Perugia, Rzeszow, Suwalki, Szczecin, Tartu, Trento, Trieste, Usti nad labem, Valladolid, Venezia

15

Very differentiated profile: on

the one hand Smart Living, Smart

Economy as well as Smart

Mobility are evaluated quite high

compared to the average. On the

other hand, Smart Governance

and Smart Environment (ie

everything that is regulated by

the state) is disproportionately

well performing / less efficient.

Graz, Linz, Luxembourg, Salzburg

Flagship-average-cluster concerning

each key field

Aalborg, Aarhus, Aberdeen, Brugge, Cardiff, Cork, Eindhoven, Enschede, Erfurt, Gent, Innsbruck, Kiel, Leicester,

Ljubljana, Madgeburg, Maribor, Pamplona, Plzen, Portsmouth, Regensburg, Rostock, Santiago de Compostela,

Stoke-on-Trent, Trier, Verona

This cluster is characterized by a

strong state (France e.g. has high

ratio of childcare age 0-3; same

as Scandinavian cities which are

also found in this cluster). Both

Smart Governance as well as

Smart People perform far above

average (only within this cluster

that degree of clarity can be

observed regarding the

combination of these key fields).

Furthermore, Smart Environment

ranges above average; the other

three key fields appear to have

still potential for improvement. Clermont-Ferrand, Dijon, Ekilstuna, Goettingen, Groningen, Joenkoeping, Jyväskylä, Montpellier, Nancy,

Nijmegen, Odense, Oulu, Pointiers, Tampere, Turku, Umeaa

16

A detailed guide for cities how to work with these analysis results to identify their respective position

in Europe and, consequently, how to search for suitable best practices can be found in the beginning

of this report (“how to make use of typology” see page 2/3).

Initially, the data used here do not directly inference to energy efficiency. Therefore, an approach of

an energy efficient clustering with selected indicators would certainly be most effective. However,

such a model is seriously challenged by poor data availability (as not all of the relevant dimensions of

energy efficiency can be covered by the available data by far). This has to be elaborated further on in

PLEEC (see next section).

Energy efficient clustering Energy efficient urban development can be delineated along three areas, namely its technological

dimension, the energy efficient behaviour of a city’s inhabitants and the prevailing energy-efficient

urban structures. Policy instruments dealing with establishing or improving of energy efficient urban

development have to allow for incorporating energy efficiency in land management, assessment of

energy efficiency through monitoring and benchmarking, decoupling growth and energy

consumption etc. However, it is important to consider that energy efficiency is not simply a reduction

of energy use but needs to consider energy-efficiency in a wider systemic perspective. In the context

of this project, energy efficiency is understood as the balanced relation between input (energy

resources) and output in form of services (through reduction of input whereby quality of services

keeps the same and through increase in quality of services whereby input of used energy keeps at

least stable).

Policy relevance of smart EE-definition is expressed by the 20-20-20 targets. The PLEEC project deals

with the elaboration of prioritized strategies for positioning of city as result of:

• place based analysis and discussion through surveys (WP 2)

• Identification of strengths and weaknesses (WP 2, 3, 4 and 5)

• Elaboration of effective measures (based on WP 2 in WP 3, 4 and 5)

Figure 7: Smart city key fields and domains on energy efficiency

17

The domains depictures in figure 7 serve as input for the next steps to be taken in the PLEEC-project

(see also figure 1 on work packages on page 5). As an insight within WP2 (especially on creating a

typology of cities), only few indicators of the original Smart cities model seem directly relevant for

the new model on Energy Smart Cities. This is one of the reasons why it is not reasonable to do a

cluster analysis based on indicators or one with special focus on energy efficiency by using the

prevailing data base on European cities. The indicators meet energy efficiency only marginally or not

at all (however, this has never been the aim of this work package).

As a result of WP2 (D2.1 and D2.2) a list of indicators resp. components as basis for discussion has

been compiled and already sent to the PLEEC model cities. These components are complemented

and developed further on in the next PLEEC work steps by taking advantages of a stakeholder’s

perspective in 6 model cities across Europe (see D2.3 and D2.4.). How to measure energy efficiency

and how to develop a synergized model for energy efficiency planning by considering the energy

efficiency potential of city key aspects will be developed further in WP4.

18

Appendix: Smart City list

List of analysed Smart cities in alphabetical order with number of corresponding cluster:

City cluster

number

AALBORG 5

AARHUS 5

ABERDEEN 5

ANCONA 3

BANSKA BYSTRICA 3

BIALYSTOK 3

BRUGGE 5

BYDGOSZCZ 3

CARDIFF 5

CLERMONT-

FERRAND 6

COIMBRA 3

CORK 5

CRAIOVA 1

DIJON 6

EINDHOVEN 5

ENSCHEDE 5

ERFURT 5

ESKILSTUNA 6

GENT 5

GOETTINGEN 6

GRAZ 4

GRONINGEN 6

GYOR 3

INNSBRUCK 5

JOENKOEPING 6

JYVÄSKYLÄ 6

City cluster

number

KAUNAS 2

KIEL 5

KIELCE 3

KOSICE 2

LARISA 2

LEICESTER 5

LIEPAJA 2

LINZ 4

LJUBLJANA 5

LUXEMBOURG 4

MAGDEBURG 5

MARIBOR 5

MISKOLC 3

MONTPELLIER 6

NANCY 6

NIJMEGEN 6

NITRA 3

ODENSE 6

OULU 6

OVIEDO 3

PADOVA 3

PAMPLONA 5

PATRAI 2

PECS 3

PERUGIA 3

PLEVEN 2

City cluster

number

PLZEN 5

POITIERS 6

PORTSMOUTH 5

REGENSBURG 5

ROSTOCK 5

RUSE 2

RZESZOW 3

SALZBURG 4

SANTIAGO DE

COMPOSTELA 5

SIBIU 1

STOKE-ON-TRENT 5

SUWALKI 3

SZCZECIN 3

TAMPERE 6

TARTU 3

TIMISOARA 1

TRENTO 3

TRIER 5

TRIESTE 3

TURKU 6

UMEAA 6

USTI NAD LABEM 3

VALLADOLID 3

VENEZIA 3

VERONA 5

19

Appendix: Methodology

References on methodology

Abdi, H., & Valentin, D. (2007). Multiple correspondence analysis. In N.J. Salkind (Ed.): Encyclopedia

of Measurement and Statistics. Thousand Oaks (CA): Sage. pp. 651-657.

Blasius, J. (2001): Korrespondenzanalyse. Munich: Oldenburg.

Multiple correspondence analysis: http://en.wikipedia.org/wiki/Multiple_correspondence_analysis

(last access: 30.01.2014)

Multiple Correspondence Analysis – Statistics output

Case processing Summary

Valid processed cases 77

Processed cases with missing values 0

Additional cases 0

total 77

Overview on model

Dimension Cronbach-Alpha

variances

total

(Eigenvalue) Trägheit % of variance

1 ,913 4,186 ,698 69,766

2 ,749 2,658 ,443 44,301

sum 6,844 1,141

mean ,849 3,422 ,570 57,034

Correlations of transformed variables

Dimension: 1

Smart

Economy

Smart

Environment

Smart

Governance

Smart

Living

Smart

Mobility

Smart

People

Smart Economy 1,000 ,498 ,560 ,684 ,707 ,816

Smart Environment ,498 1,000 ,571 ,553 ,550 ,639

Smart Governance ,560 ,571 1,000 ,580 ,560 ,732

Smart Living ,684 ,553 ,580 1,000 ,580 ,713

Smart Mobility ,707 ,550 ,560 ,580 1,000 ,762

Smart People ,816 ,639 ,732 ,713 ,762 1,000

Dimension 1 2 3 4 5 6

Eigenvalue 4,186 ,571 ,434 ,418 ,259 ,133

20

MCA objects plot (displaying case numbers)

Normalization using variable principal

Documents

Typology of cities · 2015. 8. 12. · achieve the sustainable, energy–efficient, smart city. By coordinating strategies and combining best practices, PLEEC develops a general model