Labour market sector specialisation at regional level
From Statistics Explained
- Data from March 2010. Most recent data: Further Eurostat information, Main tables and Database.
This article takes a look at employment and unemployment in the regions of the European Union (EU), clustering them into different groups according to the main sector of activity. Taking this factor into account is a useful and meaningful way to complement the analysis of regional labour markets.
Regions have been affected in different ways by the crisis at the end 2008 because they have specific characteristics and as a result they face different challenges. A better understanding of regional labour market specificities linked to their sector specialisation may lead to more efficient policies better tailored to different needs.
Main statistical findings
Regional sector specialisation
Regional sector specialisation is broadly understood to be the extent to which particular economic sectors attract larger shares of employment or output in one region as compared with another.
The sectoral composition of the regional economy affects employment patterns in several ways. For example, sectors have different rates of growth in production and demand, different employment intensities, different regulations and policies, different capital intensity or different patterns of technological change. All of these factors will influence employment in each sector differently.
Two regions belonging to the same country with similar macroeconomic conditions can have different employment patterns which can be partly explained by their degree of specialisation in the different sectors.
Regions have differing degrees of sector specialisation and, therefore, a comparison of regional labour markets which takes into account their sector composition can shed some light on the analysis.
In order to take into account the degree of sector specialisation, the first question to answer is about how this factor can be measured in a given region.
Several approaches are found in the literature, but probably the most widely used is the location quotient approach, which compares the local economy with a reference economy, in an attempt to identify specialisations in the former. The location quotient is defined as the ratio between the share of regional employment in one sector and the share of employment in that same sector in the reference economy.
The reference economy could be either the EU as a whole or the national economy of which that region is part. In this article, each region is compared with its respective country, since there are different levels of technology in the various Member States, which entail different employment intensities for the same sector in different countries. As such, comparing regions with the EU average would take precedence over the different levels of technology. This choice between EU economy and national economy inevitably gave rise to a new problem, namely that it is impossible to compute the location quotients for Member States with a single NUTS 2 region, like Luxembourg or Malta. Further on in the text, we will postulate a different approach to deal with these Member States.
The location quotient for a specific sector and a specific region is greater than 1.0 when employment in that sector tends to be over-represented in that region, and is therefore regarded as being specialised in that sector. If the location quotient is less than 1.0, local employment is less than is expected for that given sector. Therefore, that sector is not even meeting the local demands for the particular goods or services.
The underlying data used to cluster regions according the degree of specialisation are data on employment by economic activity, at NUTS levels 1 and 2 according to NACE Rev. 1.1. This is not the most recent version of NACE (the statistical classification of economic activities in the European Community), but since only three sectors were used (agriculture and fisheries, industry and services) there are no significant changes to the most recent version. In addition, longer time series are available in the old NACE classification at regional level.
The Labour force survey (LFS) measures resident employment. For regions with high levels of commuters, i.e. employed persons who work in a different region from where they live, the location quotient based on resident employment may be quite different from the one obtained using domestic employment. Nevertheless, three things attenuate this difference in the analysis that is being carried out. First, there is in general a very high share of persons who work in the same NUTS 2 region as that in which they live. Second, only three sectors are taken into account (a more detailed analysis would be more exposed to the fact that resident employment is being used instead of domestic employment). Third, the purpose of the exercise is to create only a rough and approximate classification that should not be taken as a definitive indicator of sector specialisation.
Given the share of employed persons working in agriculture and fisheries, industry and services, location quotients for each of these sectors were computed for each NUTS 2 region.
Several model-based statistical clustering techniques were used and the number of clusters was chosen according to the Bayesian information criteria. Five clusters were identified as the best choice for this data set. Each of the five clusters was characterised according to its main characteristics and this classification has been used as the starting point for grouping the NUTS 2 regions in different clusters.
Another alternative approach was to look at each region’s location quotients for agriculture, industry and services, and to decide on the minimum threshold at which a region was to be considered as specialised in a particular sector. The chosen threshold was 1.1, which means that if a region has, for example, a location quotient in agriculture of 1.1 or higher, it is labelled as being specialised in agriculture, since the relative share of employment in agriculture is at least 10 % higher than the country average. If that location quotient was less than 0.9, the region was considered as being under-represented in agriculture, while regions with location quotients between 0.9 and 1.1 were considered to be ‘balanced’.
Since the most suitable number of clusters identified for this data set was five, regions have been classified into one of the following five categories:
- specialised in services: location quotient of services greater than 1.1 and location quotients of agriculture and industry below 0.9;
- specialised in industry: location quotient of industry greater than 1.1 and location quotients in agriculture and services below 1.1;
- specialised in agriculture and industry: location quotients of agriculture and industry greater than 1.1 and location quotient of services below 1.1;
- specialised in agriculture: location quotient of agriculture greater than 1.1 and location quotients of industry and services below 1.1;
- balanced: all the remaining regions, i.e. no location quotients on agriculture, industry or services below 1.1.
The classification described above bears some similarity to the classification obtained using the model-based clustering technique described above.
Since this latter approach for clustering gives similar results to the clusters obtained using the more complex model-based cluster techniques, the first approach was chosen. The classification rules are easy to understand and the results are similar to those obtained using more advanced cluster techniques.
Finally, countries with only one or two NUTS 2 regions, such as Luxembourg or Ireland, were included in the most similar cluster, i.e. the one which has the closest distance between the region’s location quotients to be classified and the cluster average.
The classification resulting from this method is presented in Map 1.
As expected, the majority of the NUTS 2 regions in which the capital city of the respective country is located were classified as specialised in services. A closer examination of how sector specialisation is distributed geographically enables us to identify a well-defined distribution of sectors in some Member States. Hungary is divided in half, with the south-east regions specialising in agriculture and the north-west regions specialising in industry; the exception is the region of Közép-Magyarország, which includes the capital city of Budapest and specialises in services.
Italy also shows a well-defined distribution of sector specialisation, with the southern regions specialised in agriculture, and the northern regions mainly dominated by industry. Eastern Germany is basically dominated by agriculture, except for the region of Berlin, which is specialised in services; western Germany, on the other hand, is mainly dominated by services and industry.
Clustering regions according to the type of sector specialisation can now be used in regional labour market analysis. As mentioned previously, the composition of the sector can have a significant influence on regional employment patterns, and taking this factor into account will provide an additional perspective for the analysis.
High education levels in the regional labour market
To demonstrate more clearly the usefulness and relevance of taking account of sector specialisation in regional labour markets, this section will look more closely at the number of employed persons with higher education (ISCED 5 and 6) as a percentage of total employment.
As expected, higher levels of education tend to be located in regions that are specialised in services, while in regions specialised in agriculture the share of higher-educated employment tends to be below the EU average. Figure 1 shows the average share of higher education levels in employment according to the sector specialisation.
By ranking all regions according to the share of employed persons with higher education in the regional labour market, we can see that the top-three regions in terms of higher shares of employed persons with higher education are Inner London (United Kingdom) with 55.0 %, Prov. Brabant Wallon (Belgium) with 51.0 % and Brussels (Région de Bruxelles-Capitale/Brussels Hoofdstedelijk Gewest, also in Belgium) with 49.1 %. The three regions having the lowest shares are Região Autónoma dos Açores (Portugal) with 8.0 %, Severozápad (Czech Republic) also with 8.0 % and Sud - Muntenia (Romania) with 9.5 %.
While two out of the top-three regions are specialised in services (Inner London and Brussels), two out of the bottom three regions are specialised in agriculture (Região Autónoma dos Açores and Sud - Muntenia).
As Figure 1 shows, there are different levels of higher education depending on the sector of specialisation, and therefore the fact that Inner London is highly specialised in services also contributes to that high level.
To take into account the effect of both the sector of specialisation and the country in which the region is located, a linear model with two explanatory variables will be used. The linear model is significant and explains 70 % of the variability. This means that a large amount of the information available concerning the employment of persons with a higher level of education in the regional labour markets can be explained by reference to the sector of specialisation and the country to which a region belongs. In other words, it is possible to make a fair estimate of the share of higher education in one region simply by knowing that country’s share of higher education and the sector(s) in which that region is specialised.
Having a closer look at the difference between the share of higher education in employment and the estimate based on the country’s share and the sector in which that region is specialised is to put any comparison among different regions into perspective, since the influences of sector and country have been removed from the analysis. In short, this approach treats the country and sector influences separately and focuses on other regional aspects.
Table 3 shows the top 10 and bottom 10 regions in absolute terms and after subtracting the effect of country and sector of specialisation.
In absolute terms, Região Autónoma dos Açores (Portugal) has the lowest share of employed persons with higher education in the EU. However, if we take into consideration the generally low share of persons with a high level of education that is characteristic of the Portuguese labour market (the lowest in the EU) and also the fact that this region specialises in agriculture, which tends to have lower shares of people with higher education, a different scenario is revealed. If we abstract the country and sector effects on specialisation, it is the Greek region of Notio Aigaio which now ranks the lowest. The figure of 14.8 % of employed persons with a high level of education in that region stands in marked contrast to the country’s average of 25.8 % and also to the 30.3 % of all EU regions that are specialised in services.
The approach adopted in this section shows that by taking regional sector specialisation into account we can gain a different view of employment patterns. Its purpose is not to substitute or lower the absolute values published, but rather to show that there is in fact a lot of information that can be extracted from the regional labour market data available, thus allowing a more thorough regional analysis to be performed.
The results presented in the first part of this article show that in 2008 we were still seeing rising employment and falling unemployment, but to a lesser extent than in previous years. Since the labour market began to be affected by the economic crisis in late 2008, the annual averages are still in positive territory.
The regions’ success in dealing with the crisis will determine the degree of cohesion of the regional labour market in the future. The dispersion of employment and unemployment rates has already started to show small increases, breaking with the pattern of the last six years. In the years to come we are likely to see a deterioration not just in the labour markets themselves, but possibly also in regional labour market cohesion.
This article also shows that taking into account the type of region in terms of its main sector of activity gives a different and complementary view of the regional labour market. The share of employment of persons with higher education has been analysed as a way to measure the importance of the region’s own characteristics. The number of highly-educated people in a region is to a very large extent determined by the country in which that region is situated, since all regions in that country are likely to share the same education system and facilities. On the other hand, a region that specialises in agriculture is less likely to have a large share of employed people with higher education, compared to a region that is specialised in services. Therefore, it is important to take these two factors into account when making regional comparisons.
The exercise of clustering regions according to their sector of specialisation is an additional tool for producing better and more detailed regional analyses. Although it has certain intrinsic limitations due to the level of detail of the data available, clustering definitely helps to increase our knowledge of regional labour markets.
Data sources and availability
Labour force survey
The source for regional labour market information down to NUTS level 2 is the EU Labour force survey (LFS). This is a quarterly household sample survey conducted in the Member States of the European Union.
The LFS target population is made up of all members of private households aged 15 or over. The survey follows the definitions and recommendations of the International Labour Organization (ILO). To achieve further harmonisation, the Member States also adhere to common principles on the construction of questionnaires.
All regional results presented here concern NUTS 2 regions and all regional figures are annual averages of the quarterly surveys.
For further information about regional labour market statistics, see the metadata on the Eurostat website.
Cluster analysis was conducted using model-based clustering techniques based on the Bayesian information criterion (BIC) in comprehensive strategies for clustering, density estimation and discriminant analysis.
A linear regression was used to check the amount of variability in regional higher education in the labour markets that is due to the country which that region belongs to and the predominant sector of activity. The dependent variable is the regional share of higher education and the independent variables are the country’s share of higher education and the cluster to which that region was assigned. The regression is significant with an adjusted R-squared of 70 %.
Population covers persons aged 15 and over, living in private households (persons living in collective households, such as residential homes, boarding houses, hospitals, religious institutions and workers’ hostels, are therefore not included). This category comprises all persons living in the households surveyed during the reference week. The definition also includes persons who are absent from the households for short periods due to studies, holidays, illness, business trips, etc. (but who have maintained a link with the private household). Persons on compulsory military service are not included.
Employed persons are persons aged 15 years and over (16 years and over in Spain, United Kingdom and Sweden (1995–2001); 15–74 years in Denmark, Estonia, Hungary, Latvia, Finland, Sweden and Norway (from 2001 onwards); 16–74 years in Iceland) who during the reference week performed work, even for just one hour a week, for pay, profit or family gain or were not at work but had a job or business from which they were temporarily absent for example due to illness, holidays, industrial dispute and education and training.
Unemployed persons are persons aged 15–74 (in Spain, Sweden and Norway 1995–2000), and aged 16–74 in the United Kingdom and Iceland, who were without work during the reference week, were currently available for work and were either actively seeking work in the past four weeks or had already found a job to start within the next three months.
Employment rate represents employed persons as a percentage of the population.
Old-age employment rate represents employed persons aged 55–64 as a percentage of the population aged 55–64.
Unemployment rate represents unemployed persons as a percentage of the economically active population. The unemployment rate can be broken down further by age and sex. The youth unemployment rate relates to persons aged 15–24.
Dispersion of employment (unemployment) rates is the coefficient of variation of regional employment (unemployment) rates in a country, weighted by the absolute population (active population) of each region.
Location quotient expresses the relationship between an area’s share of a particular industry or sector and the national share.
Due to the late 2008 crisis regions now have to face the huge challenge of picking themselves up and getting back on track, which will certainly present them with a whole range of difficulties. Regions have been affected in different ways and they display different characteristics.
Understanding that some regions are in fact different from others, and that they are therefore likely to be confronted by different challenges, is a first step towards becoming more policy efficient, by taking measures that are tailored to the different needs.
This article takes a closer look at employment and unemployment. Regions will be clustered into different groups according to the main sector of activity and we will show that taking this factor into account is a useful and meaningful way to complement the analysis of the regional labour market.
Further Eurostat information
- Eurostat regional yearbook 2010, chapter 3
- Regions and cities, see:
- Regional statistics
- Main tables
- Regional statistics (t_reg)
- Regional labour market statistics (t_reg_lmk)
- Employment rate of the age group 15-64, by NUTS 2 regions (tgs00007)
- Unemployment rate, by NUTS 2 regions (tgs00010)
- Dispersion of regional employment rates by gender (tsisc050)
- Share of long-term unemployment (12 months and more), by NUTS 2 regions (tgs00053)
- Employment rate of the group 55-64 years, by NUTS 2 regions (tgs00054)
- Regional labour market statistics (t_reg_lmk)
- Regional statistics (t_reg)
- Main tables
- Regions and cities, see:
- Regional statistics
- Regional statistics (reg)
- Regional labour market statistics (reg_lmk)
- Regional economically active population - LFS series and LFS adjusted series (reg_lfpop)
- Regional employment - LFS series (reg_lfemp)
- Regional unemployment - LFS adjusted series (reg_lfu)
- Regional socio-demographic labour force statistics - LFS series (reg_lfsd)
- Regional labour market disparities - LFS series and LFS adjusted series (reg_lmd)
- Regional labour market data based on pre-2003 methodology (data up to 2001) - LFS adjusted series (reg_lfh)
- Regional statistics (reg)
- Regions and cities, see:
- Regional statistics
Methodology / Metadata
- Regional employment - LFS series (ESMS metadata file - reg_lfemp_esms)
- The European Union labour force survey: main characteristics of the national surveys