1.2. Used data/ approach
The database is provided by the social security statistics.4 The used data record is a full
capture of all employees who required social security insurance in Germany in the year of
2003. The statistics allow a much better regional differentiation than frequently used surveys.
Individual as well as regional connections can be examined.
In this paper commuters are examined as employees working beyond their residential
municipality (LAU5 2, formerly NUTS 5). In other words they have to cross a boundary. If no
boundary is crossed they remain inside commuters. The distance between work and
residential location can be measured when commuters are crossing boundaries ‘as the crow
flies’. In the case that commuters are not crossing boundaries no distance can be measured.
Results of the German sample census show that the distances covered by inside commuters on
average are shorter than the distances covered by commuters that cross municipality
boundaries. 75% of all inside commuters commute less than 10 km. However only 20% of the
boundary crossing commuters commute less than 10 km (BBR 2003: 129). As a result short
distances are systematically underrepresented by capturing boundary crossing commuters.
The probability that commuters leave their municipality in order to commute to work is quite
high in regions with a small average area size. In regions with an extensive surface area this
probability is much lower. This is the reason why boundary crossing commuters with a
distance of under 10 km are taken out of the analysis (Saviranta 1970: 7).
On the other hand commuters differ in the frequency of travel and thus the work commuting
distance covered. Where a commuting distance of 30 km appears to be covered daily, it is
assumed that a distance of 150 km is covered weekly. To examine a homogeneous object of
research, a distance value is placed. It separates daily from weekly commuting. Based on
appropriate investigations (Kalter et al. 2001, BBR 2003) a value of 100 km is assumed to
delimit daily from weekly commuting. Commuters are defined as boundary crossing
employees who cover a distance of 10 to 100 km.
Commuting behaviour is measured by the commuting rate in percent and the average
commuting distance in km. Further a distinction in spatial and individual influences
associated with commuting behaviour is made similar to Coombes; Raybould in 2001. They
distinguish “place factors” and “people factors” (Coombes; Raybould 2001). On the spatial
level the focus of investigation lies with the region which is chosen as an observation unit.
The first objective is to find out if there are spatial differences in commuting behaviour
between urban and other types of regions. Therefore commuting rates and average commuting
distances are examined based on a differentiation of Germany in agglomeration centers, urban
fringes, low density and peripheral regions.6 The influence of supply of employees on
commuting behaviour is expressed by employment density. It is calculated from the
accumulated employee's interests. The higher this Gini coefficient7 is, the higher is the
4 The German social security statistics contain all social security insurance requiring employment. Since 1993 employees are
registered with their place of home and place of work. The statistics include employees, workers and trainees. Government
officials and public servants, self-employed, insignificant employees and family workers do not require social security
insurance and are not included in the statistics. It is estimated that the data covers about 75% to 80% of all employed
persons in Germany.
5 Local Administrative Units (LAU) are the basic components of NUTS regions (Nomenclature of Territorial Units for
Statistics).
6 The classification is based on four spatial categories according to their centrality. The starting point is the main
agglomerations of Germany that are further divided into agglomeration centers and urban fringes. Areas with low
accessibility and density are classified as low density and peripheral regions (Bade; Niebuhr 1999: 138, 152).
7 The Gini coefficient is a measure of inequality of a distribution.