Analyzing multilevel data with QCA: a straightforward procedure

The social sciences are witnessing a growing body of multilevel theories and debates about the proper methodological tools for the analysis of multilevel data. In a recent contribution to this journal, multilevel Qualitative Comparative Analysis (QCA) was proposed as a new methodological tool for discerning set-relational patterns in multilevel data. I argue that the presentation of multilevel QCA is erroneous in two respects. First, multilevel QCA ignores the fact that equifinal solutions entail diversity and therefore leads one to overestimate the complexity of QCA solutions. Second, the ordinary minimization procedure of truth tables that contain multilevel data yields the same solutions as multilevel QCA, but is much easier to implement. I conclude that the established inventory of QCA does not need to be extended by a special multilevel approach.     

Longitudinal Dyad Models in Family Research

Multilevel modeling allows for the simultaneous analysis of data gathered at more than 1 unit of analysis (e.g., children nested in schools). It is often used to examine the effects of various contexts on individual differences in change. This paper promotes the application of multilevel models to longitudinal dyadic data in family research. By focusing on the dyad as context, researchers can examine within‐dyad change and begin to understand the interactive processes that constitute the relationship between partners. They can then frame questions about interdyad differences in within‐dyad change. We present several longitudinal models that researchers can use to examine the pattern of change within dyads, assess heterogeneity in change across dyads, and investigate cross‐partner effects on change. We comment on the implications of these models for family research.     

Exponential random graph models for multilevel networks

Modern multilevel analysis, whereby outcomes of individuals within groups take into account group membership, has been accompanied by impressive theoretical development (e.g. Kozlowski and Klein, 2000) and sophisticated methodology (e.g. Snijders and Bosker, 2012). But typically the approach assumes that links between groups are non-existent, and interdependence among the individuals derives solely from common group membership. It is not plausible that such groups have no internal structure nor they have no links between each other. Networks provide a more complex representation of interdependence. Drawing on a small but crucial body of existing work, we present a general formulation of a multilevel network structure. We extend exponential random graph models (ERGMs) to multilevel networks, and investigate the properties of the proposed models using simulations which show that even very simple meso effects can create structure at one or both levels. We use an empirical example of a collaboration network about French cancer research elites and their affiliations (0125 and 0120) to demonstrate that a full understanding of the network structure requires the cross-level parameters. We see these as the first steps in a full elaboration for general multilevel network analysis using ERGMs.     

Contextualizing the Education Effect on Women's Employment: A Cross‐National Comparative Analysis

The authors examine how and why the effect of education on women's employment varies cross‐nationally. First, they present a theoretical model that (a) outlines the micro‐level mechanisms underlying education effects on women's employment in the couple context and (b) proposes contextual moderators at the country level. Second, they test the theoretical model against survey data from the United Nations' Generations and Gender Programme for 5 European countries (Austria, France, Germany, Hungary, and Norway). The data comprise 10,048 educationally homogamous heterosexual couples involving a woman age 20–45. The results indicate that more highly educated couples are more likely to have dual‐earner arrangements in each country, yet the strength of education effects varied substantially between countries and across the family life cycle. In contrast to prior work, the authors find that education effects are not generally smaller in countries that are supportive of women's employment. This relation holds only for later child‐rearing phases.     

Modeling Repeatable Events Using Discrete‐Time Data: Predicting Marital Dissolution

I join two methodologies by illustrating the application of multilevel modeling principles to hazard‐rate models with an emphasis on procedures for discrete‐time data that contain repeatable events. I demonstrate this application using data taken from the 1995 National Survey of Family Growth (NSFG) to ascertain the relationship between multiple covariates and risk of subsequent marital dissolution. I consider both fixed‐ and random‐effects versions of the multilevel model, as well as a Generalized Estimating Equation alternative to estimating random effects. I compare results obtained from the various estimators, noting why differences occur, and recommend when to choose the various alternatives. I also provide a set of SAS and STATA programs that can be used to analyze the NSFG data.     

Comparing Group and Individual Level Assessments of Job Characteristics in Testing the Job Demand-Control Model: A Multilevel Approach

This article describes a test of Karasek's JobDemand-Control (JD-C) Model using both group andindividual level assessments of job characteristics. Bygroup assessments, we mean aggregated individual data. A random sample from general hospitals andnursing homes included 16 institutions, 64 units, and1489 health care workers (82% response). Because of thehierarchically nested data structure (i.e.,institutions, units, and individuals) the research questionsand hypothesis were tested in multilevel regressionanalyses (VARCL). The results revealed both group leveland individual level effects with regard topsychological outcomes, and stressed the usefulness ofmultilevel techniques. Karasek's JD-C Model was partlyconfirmed by finding two interaction effects at grouplevel and at individual level with regard to jobsatisfaction and work motivation, respectively. Thediscussion focuses on theoretical, methodological, andpractical implications of multilevel modeling withrespect to the JD-C Model.     

Die Analyse sozialer Kontexte

Individual behaviour is always embedded in social contexts such as networks, companies, neighbourhoods. They constitute a central condition to explain individual behaviour. We discuss major problems of context analysis. We first define context to then explore how contexts affect behaviour—the social mechanisms. We specify a multilevel model to demonstrate theoretical problems linking levels and the associated problems of adequate statistical modelling. An example is the explanation of individual behaviour by country level data.We then discuss cross pressures between contexts; here, we show how similar theoretical approaches of Simmel and Lazarsfeld are. Finally, we turn to two other problems of context analyses: non-linear effects and selection bias. We resume our arguments by suggesting a preliminary theory of contexts and strategies for future research.     

Social Context at Work: A Multilevel Analysis of Job Satisfaction

Analyzing data on classified employees working in 18 departments in a university, this paper uses hierarchical linear models to explore the effects of social context on workers' job satisfaction. Drawing on organizational demographers' claims that satisfying social relations in the workplace have demographic sources and that workers are more satisfied in demographically homogeneous work units, we examine the effects of department-level sex and race heterogeneity on workers' feelings about their jobs. Our results show that satisfaction levels are lower in more sex- and race-heterogeneous departments, as hypothesized. In addition, we found that satisfaction is higher in departments with higher average levels of job tenure, though the individual-level effect of tenure on job satisfaction was not statistically significant. Our results provide support for a social-relational view of work and demonstrate the usefulness of multilevel models as an analytic strategy for examining these issues.     

The Effect of Visas on Migration Processes

The effectiveness of migration policies has been widely contested. However, because of methodological and conceptual limitations, evidence has remained inconclusive. Moreover, prior studies focus on the effects of policies on inflows and fail to assess the simultaneous effect of policies on outflows. This is essential from a theoretical point of view as immigration restrictions may reduce both inflows and outflows and, hence, overall circulation. This renders the effect of immigration restrictions on net migration theoretically ambiguous. To fill this gap, and using unique migration and visa data from the Determinants of International Migration (DEMIG) project, this paper assesses the short‐ and long‐term effects of travel visa policy regimes on bilateral immigration and emigration dynamics. The results suggest that travel visa policies significantly decrease inflows, but this effect is undermined by decreasing outflows of the same migrant groups. This confirms that migration restrictions decrease circulation and tend to encourage long‐term settlement, and thereby sharply reduce the responsiveness of migration to economic fluctuations in destination and origin societies. We also identify asymmetric policy effects with migration flows declining only very gradually after a visa introduction but increasing almost immediately after visa removal.     

Bildungschancen von Grundschülern: Die Bedeutung des Klassen- und Schulkontextes am Übergang auf die Sekundarstufe I

Do context characteristics of class and school have an impact on elementary school pupils’ educational opportunities in addition to the confirmed individual effects of social origin and academic performance? The paper investigates the influence of the contextual performance- and SES-level as well as of the ethnic composition of the context on the probability of (1) obtaining a recommendation for higher secondary schools (Gymnasium) and (2) effectively attending a higher secondary school. To this end, we base our theoretical framework on the distinction between primary and secondary effects, as introduced by Boudon (1974), and extend it by a contextual perspective, as well as the related decisions of parents and teachers. The analyses are based on a survey of all 4^th graders of elementary schools in the city of Wiesbaden (Germany) during the school year 2006/07. The context effects are modelled by using logistic multilevel models. The results show that the affiliation to both school- and class context explains about 10 per cent of the total variance of academic achievement. Additionally, there is evidence that the context effects significantly influence the individual educational opportunities in particular when the individual marks are controlled for (secondary context effects). Furthermore, especially the SES-related context composition influences the educational outcomes when controlling for relevant individual variables. With respect to the accomplished transitions to the Gymnasium, the performance level of the school also has significant impact. Contrary to previous studies, we could not detect any effect of the ethnic composition of the context when SES is controlled for. While there are effects of the class context as well as of the school context with the educational recommendation as dependent variable, only the school context is of relevance concerning the effectively attained school transition.     

FAMILY, COMMUNITY, AND LIFESTYLE: ADOLESCENT VICTIMIZATION IN HELSINKI

The proximate correlates of adolescent victimization are well documented in the literature. The purpose of our research is to examine the more fundamental sources of this outcome. Focusing on two basic social contexts—the family and the residential community—we develop a theoretical model and test it empirically against a multilevel data set on juveniles living in Helsinki. Consistent with our hypothesis, we find that the family context is an important determinant of violent victimization. Much of this effect is mediated by involvement in delinquent lifestyles. We also find support for the assumption that community context matters to the risk of victimization independently of the family context and individual-level risk factors. Our research underscores the need for a multilevel approach in explanations of violent victimization.     

Relationship of Dyadic Closeness With Work‐Related Stress: A Daily Diary Study

We examined the association between work‐related stress of both spouses and daily fluctuations in their affective states and dyadic closeness. Daily diary data from 169 Israeli dual‐earner couples were analyzed using multilevel modeling. The findings indicate that work stress has no direct effect on dyadic closeness but rather is mediated by the spouses’ negative mood. Evidence was found for spillover of stress from work to mood at home, as well as negative crossover among couples with higher marital quality, resulting in greater distance on stressful days. Such increased distance may reflect either a deleterious effect of work stress on marital relationships or a protective mechanism used by couples in times of stress.     

Theoretische und statistische Modellierung von Cross-Pressures in Kontextanalysen

Cross-pressures can be defined as individually perceived social pressure that results from heterogeneity in the social context, mainly due to conflicting social expectations of relevant social actors. In this paper we discuss the theoretical position of “cross-pressures” in the structural-individualist paradigm and present methods for modeling cross-pressures in multilevel regression models. We argue that cross pressures form a specific class of context effects, that is macro-micro links, and sketch a reconstruction as part of the problem of defining the social situation in the Model of Frame Selection. The theoretical model implies that empirical applications of the concept would ideally use multiplicative terms for the specification of cross-level interactions in multilevel models. Modeling random slopes and context fixed-effects the interaction can be identified empirically. As an example we analyze the relation between postmaterialism and environmental concern, moderated by national wealth using WVS data. Our results indicate that postmaterialism and environmental concern form an integrated value cluster in wealthy societies, but are separate constructs in poorer societies.     

Comparing Thistles and Roses: The Application of Governmental–Voluntary Sector Relations Theory to Scotland and England

Typologies of government–voluntary sector relations mostly categorize nation-level situations, and omit consideration of intra-national differences. They are also rarely subject to empirical testing and subsequent theoretical development. This paper reports on research to apply Kuhnle and Selle's (2002) relational typology, which characterizes the experience of “Britain,” to a study of Scottish and English voluntary organizations' relations with government. Contrary to Kuhnle and Selle, the paper demonstrates — through survey data — a divergence between Scotland and England, within the “British” context, and advances the case for further testing of such conceptual typologies. The methodological challenges in moving from a theoretical framework to a research design capable of field application, using survey methods, are illustrated and discussed. Further directions for theoretically led work are suggested.     

Oppositional poverty: the quantitative/qualitative divide and other dichotomies

Despite increased recognition that the dichotomy between qualitative and quantitative approaches (and related dichotomies) may be less than useful, such divisions remain entrenched in the core social science disciplines. Theoretical and methodological attempts to move out of and beyond dichotomous frames of reference have not yet led to reconstruction of the social science enterprise. Application of Goody's analysis of the impact of literacy on human cognition to this question suggests that the mode (as well as the context) of academic discourse has structured and constrained the way we (can) think about what we do. The paper concludes by briefly reflecting on the potential of triangulation for reconstructing the way we think about methodological and theoretical dichotomies.     

The Long‐Term Consequences of Relationship Formation for Subjective Well‐Being

This study examines how relationship transitions affect subjective well‐being (SWB) and how this effect changes over time. We used prospective data containing information about 18 years of young adults' lives (PSIN, N = 5, 514). SWB was measured with the Satisfaction with Life Scale. Within‐person multilevel regression analyses showed that dating, unmarried cohabitation, and marriage had additional well‐being enhancing effects. After entry into a union, well‐being slowly decreased. A large SWB decrease was found after union dissolution, but through adaptation or repartnering well‐being increased again. Well‐being of never‐married and never‐cohabiting young adults decreased slowly over time. These effects were independent of parenthood and employment. Our results confirm expectations from the resources theory but contradict some assumptions of the set‐point theory.     

Marriage,Cohabitation, and Happiness: A Cross‐National Analysis of 27 Countries

The authors investigated how the reported happiness of married and cohabiting individuals varies cross‐nationally with societal gender beliefs and religious context. They used the 2002 International Social Survey Programme data from 27 countries (N = 36,889) and specified hierarchical linear models with macro – micro level interactions in order to examine how the social – institutional context affects happiness at the individual level. Consistent with previous research, they found a happiness gap between married and cohabiting persons. In the case of women, this gap was moderated by the gender climate and the religious context in the country. This suggests that, at least for women, this gap is not intractable but is rather an outcome of the social context. For men, the relationship between marital status and happiness was less variable across the different social contexts studied.     

Religious involvement, religious context, and self-assessed health in Europe

In the present study, the authors examine the extent to which effects of individual religious involvement on self-assessed health are influenced by the religious context (i.e., religious involvement at the country level). The authors test their expectations using individual level data (N = 127,257) on 28 countries from the European Social Surveys (2002-2008). Results of multilevel analyses show that individual religious attendance is positively related to self-assessed health in Europe. Protestants appear to feel healthier than Catholics. Moreover, modeling cross-level interactions demonstrates that religious denominations at the national level are influential: The health advantage of Protestants as compared to Catholics is greater as the percentage of Protestants in a country is higher, yet smaller as countries have a higher percentage of Catholics. The association between religious attendance and self-assessed health does not depend on the national level of religious attendance.     

Beyond interviewer effects in the standardized measurement of ego-centric networks

The effects which interviewers exert on the collection of ego-centric networks have recently come into the focus of methodological considerations. Studies consistently show that the size of networks varies depending on the interviewer. We would like to expand on this research strand by pointing to different aspects which have so far gone unremarked in the discussion. First, size is mainly analysed as a network measure which is influenced during data collection, while other common measures such as network density or composition have not received sufficient consideration. Second, large-scale surveys using face-to-face interviews usually allocate interviewers to a single sampling point. Differences between sampling points (locality effects) are attributed to interviewer effects. Hence, we disentangle the effects of the locality and interviewer. Third, the discussion on interviewer effects often follows an “actor-oriented” consideration of how data collection situations are structured by interviewers. Expanding this approach from a relational perspective, we consider the relationship between the interviewers and respondents and whether this relationship influences the collection of network data. To test our hypotheses about the influence of interviewers, the locality and the interviewer-respondent relationship on different network measures, we use data from the 2010 German General Social Survey (n = 2827 respondents, n = 220 interviewers). The multilevel analyses show that the relationship between the interviewer and the respondent is not very relevant. Furthermore, the analyses show that interviewers have an influence on the network size but not on measures of their composition. However, evidence on the prevalence of locality or interviewer effects is mixed. Finally, homophilous interviewer-respondent relationships have very little effect on network characteristics. We find evidence of training and fatigue effects on network size. However, much of the variation in network size caused by the interviewer still remains unexplained. We draw conclusions on how to organize interview situations in surveys.     

A Primer for the Use of Multilevel Models in Couple and Family Therapy Research

This article serves as a primer for understanding how to use multilevel modeling (MLM) techniques in couple and family therapy outcome and process research. It outlines the use of HLM and Mplus—two of many software programs that can perform multilevel modeling—and provides an applied example of using MLM to investigate the use of postsession feedback on couple satisfaction and change, as well as the influence of trauma on the use of postsession feedback. Two graduate students in COAMFTE accredited PhD programs read the primer and provided questions to the first author that arose. Answers to these questions have been incorporated throughout the explanation of this methodological approach in hopes of addressing common stuck points for those using MLM.