![Sample our Built Environment journals, sign in here to start your access, Latest two full volumes FREE to you for 14 days]({"type":"image" , "src" : "/sda/5926/TOC-Subject-Sample-2021-Built-Environment.jpg"})
Abstract
Using the European Community Household Panel data survey and confirmatory factor analysis, this paper explores the causal relationship between the multidimensional concepts of urbanization and socioeconomic status in the European Union. It shows that income, education and occupation are strongly related to status, and the density of population area and the town size, to urbanization. The relationship between urbanization and status is positive. This means that urban areas contain the residences of the elite, such as the rich, those with high levels of educational attainment and people with high-calibre jobs. This paper does not provide evidence to support the idea that individuals lose socioeconomic status as they age and that status is higher in capital city regions. The relationship between urbanization and status is relatively stronger for the old working-age cohort and for the non-capital city regions. These results have policy implications with regard to social welfare and urban and regional planning and development.
Notes
1. This paper does not aim to test whether there is a unidirectional causality either from urbanization to status or from status to urbanization, because theory (see Section 2) supports the idea that there is a bilateral (feedback) causality between urbanization and status.
2. The terms capital city regions, capital regions and metropolitan regions are used interchangeably. For a review of these concepts see Rodríguez-Pose (Citation2008).
3. For example, the capital city region of Belgium is Région Bruxelles-capitale and that of Greece is Attiki.
4. There is more recent information on income and living conditions in the Statistics on Income and Living Conditions (SILC) survey. However, the SILC survey does not provide data on the degree of densely populated area and on the town size.
5. Therefore, the terms constructs, components, factors and indicators are used interchangeably.
6. This paper does not use explanatory factors analysis, where all items lead on all factors, because theory specifies which items load on which factors.
7. A CFA tests the causal relationship between urbanization and socioeconomic status. The structural equation model (SEM) allows researchers to test theoretical propositions regarding how constructs are theoretically linked and the directionality of significant relationships (Schreiber et al., Citation2006). In this context, SEM can test either the impact of urbanization on socioeconomic status or the impact of socioeconomic status on urbanization. Causation in the recursive SEM model is unidirectional as there are no reciprocal paths. Since theory confirms a bilateral relationship between urbanization and status, this paper uses CFA rather than SEM. The estimation of a non-recursive SEM is no less problematic than it is in a single-equation regression model as it should control for causality using, for example, instrumental variables estimation. According to the theory, there is no clear evidence on direction in which the causal relationship between urbanization and status runs. However, both CFA and (recursive and non-recursive) SEM explain how changes occur in various constructs and in how various constructs are related to each other, and are used to establish the validity of measurements and answer substantive questions.
8. Unfortunately, there are no data on urbanization for the remaining European countries.
9. Personal income includes wages and salaries, income from self-employment or farming, pensions, unemployment and redundancy benefits, any other social benefits or grants, and private income (ECHP documentation).
10. Nevertheless, the LISREL program (Jöreskog & Sörbom, Citation1996; Jöreskog et al., Citation1999), which is used for this study, has an option to normalize variables before analysis, thus providing a way to deal with non-normality in samples of small and moderate size (Jöreskog et al., Citation1999). Hence, the LISREL program can take ordinal or skewed data and transform them into forms that can be used. LISREL has the feature of calculating scaled test statistics which appear to be asymptotically robust against deviations from normality (e.g. Browne, Citation1984; Satorra, Citation1990). Thus, the ML estimation seems to be fairly robust to violations of normality.
11. Absolute indices determine if the proposed model is consistent with the data without the use of a reference model, while comparative indices judge “proportionate improvement in fit” by matching the hypothesized model with a nested baseline model (Holbert & Stephenson, Citation2002; Hu & Bentler, Citation1995).
12. The input of the analysis is a correlation matrix of the measured variables (see Appendix). Due to data availability, Ireland and Denmark are considered as single capital regions. However, even if these countries are excluded from the study, the results are still robust. The GLS and OLS estimators of the CFA for the whole of the population show the robustness of the results to violations of the statistical assumptions of potential estimation methods (Boomsma, Citation2000; Hoogland & Boomsma, Citation1998). All these results can be provided upon request.
13. The results are not reported because of space constraints but can be obtained on request.
14. These CFA results can be provided upon request.