![Sample our Urban Studies journals, sign in here to start your FREE access for 14 days]({"type":"image" , "src" : "/sda/110826/TOC-Subject-Sample-2021-Urban-Studies.jpg"})
Abstract
In this article, we argue that similarly to the literature on urban neighbourhood effects, the idea of disadvantaging residential environment can be used to explain the socio-economic disadvantage of residents in rural peripheries. We present arguments as to why it is appropriate to consider the effects of the residential environment in a micro-regional perspective and outside of metropolitan areas. These effects are empirically analysed using income related data from the Czech Republic. We ask whether income is negatively affected by housing in peripheral micro-regions. The results confirm that the concept of residential disadvantage is also relevant in rural areas, and that in the Czech Republic residence in remote rural areas with limited educational levels and a high concentration of low-pay jobs has negative yet weak contextual effects, whereas increased levels of deprivation and social exclusion, typical also of many urban regions, are not decisive. A stronger negative impact of living in peripheries has been found on higher-skilled workers and women.
Disclosure statement
No potential conflict of interest was reported by the authors.
Notes
1 The extensive literature dealing with neighbourhood effects on children and youth is left aside in this overview.
2 The proposed procedures for solution consist either in the use of longitudinal data (Galster et al., Citation2010; Hedman et al., Citation2015; Musterd et al., Citation2012) analysed by a fixed effect approach, or multiple individuals within groups, e.g. dyadic data for siblings, or with the aid of an instrumental variable estimator (Fisher Citation2005; Galster et al., Citation2007; Hedman & Galster, Citation2013). The fundamental limitation of the fixed effect approach is the need for very extensive longitudinal data sets, with sufficient variability of the dependent and independent variables. As a rule extensive data obtained from the population register can best meet such demands. However, this is available for research purposes only in a small number of countries. In the Czech case, the use of fixed effect models is not realistic due to the non-existence of a usable longitudinal set.
3 Source: Demographic statistics of the Czech Republic, Czech Statistical Office
4 .Instrumental variable estimation uses exogenous variables (instruments) that serve as proxies for endogenous ones and thus provides a general solution to the problem of an endogenous explanatory variable (Wooldridge, Citation2010, p. 89). As instruments we used two conceptually different types of variables. These were on one hand variables measured on the level of residence, linked to spatial characteristics preceding the observed characteristics of peripherality, specifically the rural nature of the environment operated via rurality (the low population density), and on the other variables relating to the socio-economic status of the partner in the household, specifically three markers relating to his/her position on the labour market. We discuss the validity of these instruments in the corresponding part.
5 We limited our sample to employees only because cash benefit related information from the self-employed is of limited reliability and comparing employee cash income and cash benefits from employment for the self-employed is a problematic procedure (Večerník Citation2011). One could argue, that in rural areas, there is higher probability of individuals being engaged in farming, family business and other self-employed business and removing this group from the study could result in bias. In fact the share of self-employed is practically the same in cities and rural areas (15.1% and 14.5%, respectively).
6 All analyses were also carried out with the minimum of 10 cases per a micro-regional-unit (resulting in 18% reduction of micro-regional-units sample and 4% of individual cases for the whole sample, and in case of couples data 56% and 24%, respectively). The results were virtually the same (data not shown).
7 Survey Social disadvantage in peripheral areas of Czechia, 2016, representative for Czech adult population (N = 2448).
8 We also performed a PCA on all area-level variables together, which confirmed the allocation of these two dimensions.
9 The metropolitan area of the capital city of Prague does not correspond to a NUTS2 region, as extensive outer parts of it belong to the Central Bohemian region. For this reason, we assigned the parts of the Central Bohemian region most densely interconnected with the capital (daily commuting to Prague by more than 50% of economically active inhabitants) to Prague. There are important analytical reasons for this, due to the large wage differences between Prague and the other regions.
10 As we expect wage differences to be influenced by accessible employment opportunities, including the location of the employment in the analysis would help determine how commuting affects wage. This seems important in particular as rural peripheral localities exhibit relatively low shares of commuters, whereas in peri-urban localities commuting is more frequent. Unfortunately, EU-SILC data lack information on employment localisation.
11 Whereas the second condition can be directly tested, the first needs rather a theoretical justification. There is always a question of plausibility of the exclusion restriction, i.e. whether our instrument is valid in terms of whether it transmits is influence onto the income outcome only through the mediating endogenous variable – peripherality. We use two different sets of instrumental variables and a combination thereof and triangulate the results of different models to enhance the plausibility of our results.
12 There is a correlation between the income of the partners, which is due to marital homogamy, but after we control for their individual characteristics (e.g. education and occupation), only the influence of residence remains.
13 The assumption of exclusion restriction can be further tested. If a researcher has more instruments than endogenous variables, a test for over-identifying restriction can be performed. This will verify the validity of the instruments, but only as a whole for the considered set of instrumental variables. With only a single endogenous regressor – rural low qualification peripherality, models 6 and 8 are over-identified whereas model 7 with rurality as the only instrument is exactly identified. This allows us to perform tests of over-identifying restrictions for models 6 and 8, i.e. we can test whether the instruments are uncorrelated with the error term. The results for the test are presented in the bottom section of table 4. The null hypothesis that instruments are exogenous (i.e. are uncorrelated with the error term of the wage equation) cannot be rejected in both models 6 and 8. This increases the credibility of the instrument sets. In addition, we test for the presence of weak instruments, which can be carried out with the first-stage regression F-statistics. In all three models (6, 7, 8) it far exceeds the value of 10, which is commonly considered the minimum limit, thus indicating a reasonably strong set of instruments.
14 We use the algorithm EC2SLS implemented within the command xtivreg in STATA 14.
Additional information
Funding
Notes on contributors
Josef Bernard
Josef Bernard and Jiří Šafr are researchers in the department Local and Regional Studies at the Institute of Sociology of the Czech Academy of Sciences.
Jiří Šafr
Josef Bernard and Jiří Šafr are researchers in the department Local and Regional Studies at the Institute of Sociology of the Czech Academy of Sciences.