manner, one sees new factors arise, either by splitting a factor or by forming new factors
alongside the existing ones. By following this process step by step, one gains insight in
the structure of the factors. In factor analysis, the identity of the factors is usually
determined on the basis of the matrix of loadings. In this case, the interpretation of the
results is facilitated by the fact that each of the variables is linked to a location. This
makes it possible to depict the factor loadings in maps.
For the 1983 and 1993 surveys, the principal components analysis led to similar results.
In both cases, the result of the rotation of three components lends itself best to
interpretation in terms of location factors. For that reason, this variant was selected for
further analysis (Meester 1994). Figure 4 shows the loadings on the three rotated
components for the 1993 survey. The patterns for 1983 are almost identical.
The first component, depicted in Figure 4a, expresses an opposition between the center
of the country and the periphery. If we want to interpret this component in terms of
location factors, an interpretation as ‘relative location with respect to the national
market’ is self-evident.
The second component shows an opposition between the coastal provinces and the East
and South of the country (Figure 4b). The pattern displayed here shows remarkable
similarities to the pattern of residential preferences in the Netherlands. Therefore, an
interpretation as ‘residential environment’ seems appropriate.
The third component displays a pattern of high loadings in the West of the Netherlands
and low loadings in the eastern periphery (Figure 4c). The area that is bounded by the
0.6 isopleth coincides almost precisely with the Randstad. Apparently, agglomeration
effects manifest themselves in this component as a location factor. Under this heading
fall the advantages of agglomerations but also the disadvantages, such as congestion,
lack of space, high land prices, etc.
The correspondence of the results of the surveys of 1983 and 1993 is remarkable. The
identity and the order of the three rotated components are the same. Also the proportion
of explained variance is virtually unchanged (59 and 61 %, respectively), and even the
factor loadings are essentially the same. All these results must be interpreted in light of
the fact that the respondents in the second survey are not the same ones as in the first.
Because the general pattern of locational preferences is still the same in 2003, one might
expect to see similar results for the principal components analysis as well. This turns out
not to be the case, however. The main results, which are not shown in maps here, can be
summarized as follows.