EUROSTAT
Directorate E: Sectoral and regional statistics
Unit E-4: Regional statistics and geographical information
Task Force on harmonised Labour Market Areas
Notes of the 2nd TF meeting on
24th March 2014
Luxembourg
Date: / 05/06/2014Version: / 1.1
Authors: / Oliver HEIDEN
Revised by:
Approved by:
Public:
Reference Number:
Document History
Version / Date / Comment / Modified Pages1.0 / 02/06/2014 / Document created by Oliver HEIDEN
1.1 / 05/06/2014 / Document revised by Oliver HEIDEN / all
Contact: Oliver HEIDEN, Telephone: +352 4301-31961,
Agenda of the meeting
1.1 / Welcome / Oliver HEIDEN
1.2 / Approval of the agenda
(for information and adoption) / Agenda / All
Item 2 / Review of the result of the sensitivity analysis of the Euro (TTWA) method
2.1 / TTWA
(for information and discussion) / PowerPoint presentation / Oliver HEIDEN, all
2.2 / Conclusions, distribution of tasks
(for discussion and adoption) / All
Item 3 / Review of data source for the method
3.1 / Coverage (statistical population vs. sample)
(for information and discussion) / PowerPoint presentation / S. Cruciani, all
3.2 / Specifications of data (population: employed vs. active population), (commuting: daily vs. total) etc.
(for information and discussion) / PowerPoint presentation / S. Cruciani, all
3.3 / Conclusions, distribution of tasks
(for discussion and adoption) / All
Item 4 / Conclusions of the meeting / All
1.Participants
Andy BATES – Office for National Statistics (ONS, UK)
András KEZÁN – KözpontiStatisztikaiHivatal (KSH, HU)
Hilde KEUNING van OIRSCHOT – Centraal Bureau voor de Statistiek (CBS, NL)
Paweł STOPIŃSKI – GłównyUrządStatystyczny (GUS, PL)
Sandro CRUCIANI – Istitutonazionale di statistica (ISTAT, IT)
Stéphanie MAS – Ministère du Travail, de l'Emploi, de la Formation professionnelleet du Dialogue social (FR)
Gunter SCHÄFER – Eurostat
Oliver HEIDEN – Eurostat
Document repository
All documents of the task force are stored in a CIRCABC repository i.e. in the Eurostat Library of Regional and Urban Statistics (LMA Task Force) and are publicly available:
2.Organisational topics
The Task Force members complained that travelling to Luxembourg is very time consuming and expensive. The option has been discussed to hold the 3rd meeting of the Task Force in a city that is easier to reach by all participants. Preferably, the meeting place shall be the home institute of one of the participants. Therefore, Eurostat proposed to investigate the possibility to hold the next meeting in Paris or London.
The Task Force members decided that a fourth meeting will be necessary in order to have the possibility to draft the conclusions and recommendations of the Task Force for the Regional Statistics Working Party on 20/21 October 2014 in Luxembourg. The 4th meeting of the Task Force shall preferably take place in September 2014.
3.Conclusions of the 1st Task Force meeting
The Task Force agreed in its first meeting on 18th November 2013 to aim at a simple, transparent, reproducible, consistent and policy independent bottom-up method of comparable LMAs for the EU that focuses on larger LMAs with European relevance on the basis of self-containment constraints measured by commuting flows.
In order to understand the functioning and robustness of the two existing national methods advocated in the Eurostat Study on comparable Labour Market Areas (hereafter:the Study), members of the Task Force requested Eurostat to carry out sensitivity analyses of the LAM[1] and TTWA[2] methods on EU commuting data. The purpose of the sensitivity analyses was set to measure the impact of the parameters on the resulting grid of LMAs.
4.Sensitivity analysis of the Euro (TTWA) method
4.1.Preliminary aspects
The closer analysis of the algorithms of both, the LAM and the TTWA methods revealed that the LAM method cannot be considered to be a bottom-up approach. The LAM method starts with the definition of central municipalities, and follows a repetitive concentration method to attach other municipalitiesto the centre. Given that the Task Force has explicitly advocated a bottom-up approach, the LAM method was not further analysed and the sensitivity analysis focused on the Euro (TTWA) method.
The sensitivity analysis required at first place commuting data. For the majority of the countries commuting data is only available for the Census year 2001. Though, the 2001 tables are not complete, especially the EU10 tables are often erroneous or lacking content. Nonetheless, in order to minimise the GIS work related to the geographical visualization of the results, Eurostat concentrated on the 2001 tables.
As a further requirement a software program that is able to run the TTWA algorithm on a substantial amount of commuting flows was needed. Unfortunately, due to licencing issues the script used by the experts for the calculations presented in the Study was not available.Statistics Netherlands (CBS) has kindly shared with Eurostat a script written in R, however, the documentation of the algorithm left some questions open. Finally, the former external contractor managing the project on the Study provided a script written in JAVA. The sensitivity analysis was carried out on Eurostat machines with the JAVA script.
The main advantage of the TTWA method, as last revised in 2007[3], is the limited requirement for input data. The method runs on sheer commuting flows (incl. the number of non-commuters), and calculates the LMAs without a need for any further data or constrains. This characteristic is very striking because in the targeted scenario the method has to be applied to all Member States with very different commuting patterns.
4.2.The TTWA algorithm
The TTWA method operates with two properties: (1) self-containment and (2) number of workers (economically active population). For both properties minimum and target values have to be defined to run the algorithm. Consequently, four parameters set the constraints to define what is considered to be an LMA:
- minimum self-containment (minSC),
- targetself-containment (tarSC),
- minimum number of workers (minSZ),
- target number of workers (tarSZ).
Self-containment is measured for both, the supply and the demand side. Supply side self-containment (SSC) is the number of people living and working in an area divided by the number of residents in the area. Demand side self-containment (DSC) is the number of people living and working in an area divided by the number of jobs in the area.
The TTWA method considers a cluster of municipalities to be an LMA ifthree conditions are simultaneously fulfilled:
This can be translated into following figure:
Figure 1
In practice, the algorithm starts with assessing everymunicipality (Local Administrative Units (LAU)).Against the conditions outlined above and selects the LAU that is furthest from being an LMA. This LAUAis then assessed against all other LAUs to find the one LAUBwhich has the most important commuting flows according following formula:
where:‘Commuters’ stands for the commuting flow,
‘Workers’ stands for the economically active resident population of a municipality and
‘Jobs’ stand for the population economically active in a municipality.
The LAUB with the most important commuting links to LAUA is the one LAUA is grouped with. The grouping of LAUA and LAUB is now considered as one entity and the joined commuting flows to the other LAUs will be recalculated. Than the procedure restarts, while LAUs and groupings of LAUs are equally treated. The process stops when all LAUs or groupings fulfil the three conditions of an LMA.
4.3.Consequences for the application of the TTWA method
Given that the size of the economically active population both resident and working in a given municipality is an important input of the algorithm, the calculation is highly dependent on the size of the LAUs. Moreover, not only the population size but also the area of the LAUs will seriously impact the result of the algorithm, assuming, as it is likely, that a larger area will have a higher self-containment. Therefore the municipal structure of a country will have a substantial impact on the resulting grid of LMAs.
Other substantial impacts on the result of the algorithm can be summarised under the term ‘commuting behaviour’. Contrary to the population and area size of the LAUs, the commuting behaviour is an unknown feature. It can depend on the transport infrastructure, petrol prices, commuter allowance and many more attributes of a region or a country.
In order to minimise the impact of the size of LAUs Eurostat chose – from the Member States where data was available – two pairs of countries with similar LAUs in terms of population and area sizes:
- Netherland (NL) and Sweden (SE)
- average LAU-2 population approx. 33.000 inhabitants
- average LAU-2 area 100-1500 km2
- Austria (AT) and Hungary (HU)
- average LAU-2 population approx. 3.300 inhabitants
- average LAU-2 area 33 km2
NL and SE are characterised by rather large LAUs, while AT and HU have typically small LAUs. In addition, the former pair has some kind of official LMAs in national use, while the latter pair has no national methodology established.
4.4.The analysis
The sensitivity analysis was carried out by starting with the four default values of the TTWA method as applied in the UK and changing the parameters one-by-one keeping the remaining three parameters constant.
TTWA default (0) / Sensitivity analysisminSZ-var. (1) / tarSZ-var.
(2) / minSC-var. (3) / tarSC-var.
(4)
minSZ / 3.500 / 1.000 / 3.500 / 3.500 / 3.500
tarSZ / 25.000 / 25.000 / 50.000 / 25.000 / 25.000
minSC / 66.6 % / 66.6 % / 66.6 % / 70.0 % / 66.6 %
tarSC / 75.0 % / 75.0 % / 75.0 % / 75.0 % / 85.0 %
Table 1
The default and all four variants of parameters have been applied to every country. Following table summarises the results. Annex 1 shows the maps created for the TTWA default and the four variant for all countries.
Results of the Sensitivity analysisNumber of LAUs / Total workers / Variant / Number of LMAs / Average number of LAU per LMA / Largest LMA (workers) / Smallest LMA (workers) / Average size of LAM (workers)
NL / 504 / 6.860.684 / TTWA (0) / 53 / 9.5 / 714.025 / 5.838 / 129.446
minSZ-var. (1) / 55 / 9.2 / 782.539 / 1.504 / 124.739
tarSZ-var. (2) / 46 / 10.9 / 973.544 / 5.838 / 149.145
minSC-var. (3) / 37 / 13.6 / 872.464 / 5.838 / 185.423
tarSC-var. (4) / 52 / 9.7 / 785.702 / 5.838 / 131.936
SE / 289 / 3.885.151 / TTWA (0) / 133 / 2.2 / 811.244 / 3.851 / 29.211
minSZ-var. (1) / 157 / 1.8 / 811.244 / 1.018 / 24.746
tarSZ-var. (2) / 139 / 2.1 / 788.168 / 3.851 / 27.950
minSC-var. (3) / 123 / 2.3 / 811.244 / 3.851 / 31.586
tarSC-var. (4) / 121 / 2.4 / 811.244 / 3.851 / 32.108
AT / 2363 / 3.378.335 / TTWA (0) / 63 / 37.4 / 991.154 / 4.078 / 53.624
minSZ-var. (1) / 72 / 32.8 / 994.412 / 1.306 / 46.921
tarSZ-var. (2) / 54 / 43.8 / 959.672 / 4.056 / 62.561
minSC-var. (3) / 54 / 43.8 / 1.015.278 / 4.078 / 62.561
tarSC-var. (4) / 47 / 50.3 / 964.035 / 4.548 / 71.879
HU / 3135 / 3.480.428 / TTWA (0) / 150 / 20.9 / 964.034 / 3.595 / 23.202
minSZ-var. (1) / 250 / 12.5 / 963.487 / 1.000 / 13.921
tarSZ-var. (2) / 149 / 21.0 / 992.382 / 3.595 / 23.358
minSC-var. (3) / 143 / 21.9 / 960.362 / 3.676 / 24.338
tarSC-var. (4) / 144 / 21.8 / 967.757 / 3.595 / 24.169
Table 2
Further resultsrelated to the impact on the numbers of LMAs were illustrated by two graphs prepared by Eurostat.
4.5.Discussion
The Task Force discussed the results of the sensitivity analysis intensively. One substantial finding was that countries react with different sensitivity of the change of different parameters. As an example the number of LMA skyrocketed in Hungary when the minimum self-size was set on 1000 workers, while in other countries the number of LMA stayed rather stable with this variant. Netherlands reacted sensitively to the increase of the minimum self-containment and Sweden to the increase of the target self-containment.
5.Conclusions
The Task Force endorsed,that for every country a different set of parameters can be optimal. Therefore, for the establishment of the harmonised EU method it is proposed that countries shall have the possibility to define their own set of parameters while using the same algorithm. The Task Force decided that the participating Member States shall have a first try to identify the set of parameters that fits best to the characteristics of their country. At the same time it was stressed that the Task Force shall focuses on larger LMAs with European relevance.
6.Planning until the next meeting of the task force
In order to allow the members of the Task Force to carry out the sensitivity analysis own their own country, first the script of the algorithm needs to be circulated. As not all of the members are familiar with JAVA scripts the Task Force decided to use the R-script. In order to finalise, document, and, if necessary correct the script ISTAT and CBS volunteered to check whether the R-script is following the TTWA algorithm described by ONS by end of April.
Once the script is final and ready to use the Statistical Offices of Poland, Hungary, France, Netherlands, Germany (?) and Italy will try to define most suitable parameter sets. Countries with official LMAs in use (FR, IT, (NL)) shall preferably define parameters leading to results that are very similar to the results of their own methodology. Other countries shall start with defining ranges where the change of the parameter has little effect on the outcome. It is worth mentioning that is shall be possible to apply manual changes on the results, if necessary. In addition, it is vital to concentrate on large LMAs with European relevance; in case of smaller LMAs rather the structure shall be checked for plausibility. Given that experience shown that the script can run for several hours or days, large countries shall considerstarting to work on a part of the country. First result shall be circulated at the end of May.
ANNEX 1
Results of the Sensitivity analysis (maps)
Netherlands
TTWA (0) / minSC-variant (1)tarSC-variant (2) / minSZ-variant (3)
tarSZ-variant (4)
Sweden
TTWA (0) / minSC-variant (1)tarSC-variant (2) / minSZ-variant (3)
tarSZ-variant (4)
Austria
TTWA (0) / minSC-variant (1)tarSC-variant (2) / minSZ-variant (3)
tarSZ-variant (4)
Hungary
TTWA (0) / minSC-variant (1)tarSC-variant (2) / minSZ-variant (3)
tarSZ-variant (4)
Page1 / 12
[1]Swedish methodof Labour Market Areas: Lokala arbetsmarknader (LAM):
[2] UK method of Labour Market Areas: Travel to Work Areas (TTWA): , the Euro method presented in the Study is basically identical to the TTWA method. The sole difference is in applying other parameters for the size of the LMAs.
[3]