Quantitative Spatial Analysis of Micro and Macro Data: Methodological Challenges and Solutions 2 |
|
| Chair | Professor Henning Best (TU Kaiserslautern ) |
| Coordinator 1 | Professor Corinna Kleinert (Leibniz Institute for Educational Trajectories) |
| Coordinator 2 | Mr Tobias Ruettenauer (TU Kaiserslautern) |
| Coordinator 3 | Dr Michaela Sixt (Leibniz Institute for Educational Trajectories ) |
Previous research has shown that individual labor market outcomes con-siderably depend on regional opportunity structures, such as district labor market conditions. However, the regional units usually available for data analysis are often artificial and, hence, do not adequately represent individuals’ opportunity structures: They may not be point-located but rather distributed in space, and neglecting this spatial dimension may result in underestimating regional effects.
However, when analyzing such spatial externalities, the crucial question is how to model these effects, i.e., how the wider spatial context has to be operationalized in order to capture its significance for labor market outcomes. There are two significant dimensions: The first concerns the extent of opportunity structures and the issue of whether it is sufficient to only consider immediately adjacent districts or if a much wider “zone of influence” must be taken into account. This issue is related to the Modifiable Areal Unit Problem which states that the results of spatial data analyses are generally sensitive to the spatial scale and spatial zoning system used. The second dimension concerns the problem of spatial autocorrelation of characteristics at the context level, which arises due to social interrelation and exchange between areal units. Spatial autocorrelation of indicators at the context level may lead to underestimating the effects of local and non-local opportunity structures.
We will tackle these issues by taking youths’ school-to-training transitions into account and testing for different ‘zones of influence.’ For this purpose, we link retrospectively collected life course data from Germany’s National Educational Panel Study with administrative regional time series data on the supply of training positions at the level of districts (Landkreise).
Our analyses show that the ‘zone of influence’ comprises of the whole of Germany, not only close-by districts, and that these effects substantially differ between structurally weak and strong regions. In structurally strong regions the probability of finding an apprenticeship increases with an increasing number of training spots in non-local districts, while in structurally weak regions, an increasing number of positions in non-local districts is associated with a decreasing probability. In a nutshell, assuming that only close proximity affects individual outcomes may disregard relevant contextual influences, and for spatial models that require an a priori definition of the weights for spatial units, it may be erroneous to make a decision based on this assumption. Concerning spatial autocorrelation, we found that neglecting local spatial autocorrelation at the context level causes considerable bias to the estimates, especially for districts that are close to the home district. That is, in order to map the actual extent of an individual’s spatial opportunity structure, regression results have to be adjusted for local spatial autocorrelation between the home district and all other districts, respectively.
Spatial and geographical aspects play an important role when analyzing partner market’s influence on demographical processes – like finding a partner and forming a relationship. In the project “The macrostructural context of the marriage market: longitudinal dynamics” we constructed a set of small-scale partner market indicators. In our presentation, we will show how these indicators help to improve the analysis of relationship formation.
Usually, studies about the partner market use simple sex ratios with spatial units based on federal states or whole countries. This is in contrast to empirical evidence showing that the choice of a partner is predominantly restricted to much smaller contexts. Individuals’ oppor-tunities to get into contact with others will be described most accurately when the partner market is defined within spatial limits that are congruent with the daily rounds of an individual, i.e. the spatial range of individuals’ everyday activities.
Therefore, our partner market indicators are based on German local districts (NUTS3 regions). In doing so, we assume that people’s daily rounds are restricted to the borders of these districts. Evidently, this assumption might be too strong, since many people pass district borders daily and thus are part of the partner market in adjacent districts, too. Therefore, we developed a special blending-method using commuter data to generate weights to blend indicators for adjacent local districts.
In contrast to usual simple sex ratios, our indicators account adequately for several theoretic arguments about the relevance of men and women in different age-groups and with different educational attainments and about the availableness on the partner market of those individuals who already have a partner. Because of these specifications an indicator value can be assigned uniquely according to sex, age, local district (and education). There is a low probability to have more than one respondent in a survey with the same combination of these traits. Therefore, there is no need for sophisticated methods to produce unbiased effect estimators like multilevel analyses. In addition, we constructed indicators that measure the transparency of the partner market (using information about the supply of potential partners only), and we constructed indicators that additionally account for the competition about potential partners.
Our presentation will illustrate these advantageous methodic features of the developed indicators for the description of the partner market and for explaining relationship formation and the choice of partners.
This research work aims at understanding how mechanisms that can be located in socio-economic and socio-structural contexts contribute towards explaining social disparities in educational transitions. Existing concepts of contextual settings (e.g., neighborhoods, labor markets, peer influences) are enhanced in terms of a flexible conceptualization of spatial data to gain a better understanding of where to locate socio-spatial contexts.
The paper pursues two major goals: (1) By utilizing flexible context measurements (i.e. travel-time radii), I am directly referring to the modifiable areal unit problem (MAUP) with the aim to identify the spatial extent in which contextual effects that impact educational transitions can be located; (2) Arguing from a life-course perspective, I assume that the size of socio-structural and socio-economic contexts differs between stages of the educational career.
My application examples focus on the impact of local unemployment and local educational infrastructure on three transitions across the educational career: (i) primary to secondary school, (ii) secondary school to vocational and academic training and (iii) vocational and academic training to the labor market. The main aim is to identify the most adequate spatial extensions of contextual effects and to compare them across educational stages.
To analyze these transitions, data from the National Educational Panel Study (NEPS) is utilized. The NEPS-Starting Cohort 6 – Adults (NEPS-SC6) is a retrospective questionnaire providing extensive information on education and occupation along individuals’ life courses. The data is linked to macro-level information. Due to the availability of macro data, the observation period spans from 1986 to 2013. In this paper, special emphasize is given to challenges regarding the linkage between survey and macro data on small levels of administrative aggregation (i.e., municipalities). Particularly changes in territorial units over time are discussed.
To overcome the limitations of contextual information that are aggregated within administrative borders, I make use of a travel-time matrix obtained from the Federal Institute for Research on Building, Urban Affairs and Spatial Development (BBSR). This matrix provides information on travel time by car between all German municipalities. Between these municipalities travel-time radii of 15, 30, 45, 60, 90 and 120 minutes are calculated. I will illustrate how the available macro-level variables on labor-market conditions and educational infrastructure can be re-calculated for each of the six travel-time radii so that the contextual units of analysis are no longer municipalities but areas defined by travel-time centered around the municipality of the respondent’s residence. Each travel-time radius for each year of the observation period is characterized by specific contextual information.
Findings indicate strong differences between travel-time radii; Effects decrease with increasing radii. Moreover, the adequate spatial extensions differ strongly between the three transitions across the educational career. In a final step, I compare these results with conventional methods of aggregation within administrative units (e.g., municipalities and administrative districts). Effects seem to be stronger and more precise for travel-time radii than for administrative units.
The transition to secondary school system is one of the most important ones in Germany. Effects of social origin on the educational decision at that transition are multiple investigated. There is less research done on the importance of educational infrastructure. Although we have some findings, that show positive correlations.
Geipel 1965 and Peisert 1967 showed first that children in rural areas leave schools much more often without a high school degree than children in urban areas. Rural and urban areas were defined by the number of residents in a community. Henz and Maaz (1995) found in individual data of different birth cohorts a negative effect of a home place in a rural city on the probability to visit a high school. Even if this effect diminished with time, it could be still observed in the last cohort. The assumption behind: infrastructure in rural areas is less expanded than in urban areas. Therefore Sixt (2010) did a regression of the proportion of high schools on the probability of high school attendance in a multilevel approach. The expected positive correlation of school infrastructure was found. But the question is if the county level is the adequate level and the measurement of educational infrastructure on the basis of administrative borders is adequate.
In the paper at hand we will discuss this question. We ask again, if there is a correlation between high school attendance at grade 5 and availability of high schools. Theoretically we follow the arguments of the rational choice theory. There availability of high schools is a cost factor when evaluating alternative educational decisions. We use data from two waves of a regional population survey from the project „Educational Landscape of Upper Franconia“. These data allow controlling for several relevant variables measured prospective in the first wave in fall/winter 2015/16 (e. g. social origin, educational aspirations and evaluations of different educational alternatives). The choice of secondary school is surveyed at the moment in the second wave.
To investigate the question above we will use two different measurements for the availability of high schools: First, we use the proportion of high schools in relation to other schools in the county a pupil lives. Calculating a logistic regression with a multilevel approach, we expect a positive effect of the proportion of high schools in the home county. But we argue that administrative borders and infrastructure within those borders are not adequate. We argue that people probably don’t know them. What they probably know is how far away the next high school is. Following we use - second - the distance from pupil’s address to the address of the next high school. Doing a logistic regression we expect a negative correlation: the smaller the distance to the next high school the higher the probability to visit a high school. In a third step we will combine both measurements. We would like to discuss differences and challenges in using those measurements with regard to the used methods and results as well as consequences.