This lecture attempts to discuss the possibility of causal inference in social sciences. There are four strategies for attempting to derive causal inferences from observational data: (1) cross-tabulation approach, (2) regression approach, (3) methods based on panel data, and (4) counter-factual framework. The lecture discusses these four approaches in detail.The lecture concludes with the discussion of the "heterogeneous causal effect" inherent in the society, the idea introduced by Otis Dudley Duncan, and the "causation as generative process" advanced by John H. Goldthorpe. I argue that these two ideas are complementary, and that the two great sociologists have highlighted important issues which must be confronted by social scientists in the course of causal inference using the observational data.

View full abstract

     This paper demonstrates the importance of panel data analysis that has not yet been employed by several Japanese sociologists. Although we can build a longitudinal type of data from a cross-sectional retrospective survey, it may further reduce the reliability of responses because of incorrect recall and the sample selection bias. Causal inference becomes a significant task for social scientists, and we need to prepare the panel data that includes detailed information about changes in order to resolve this task. While we cannot obtain unbiased estimates from Ordinary Least Squares (OLS) owing to endogeneity, which means that there is a correlation between independent variables and unobserved heterogeneity, we can eliminate the bias from the unobserved heterogeneity if we employ an econometric fixed-effect regression model. In addition, this paper also introduces a hybrid model that applies the idea of group mean centering at multilevel modeling, and it enables us to obtain the within and between estimators simultaneously. Finally, I will introduce the future tasks for panel data analysis.

View full abstract

     The aim of this study was to use 4-year 3point cross-lagged analysis in Japanese life course survey longitudinal data and investigate the following hypotheses for causal relationship: psychosocial work characteristics affect mental health and sense of coherence mediated the causal relations. Male subjects between ages 20 and 40 living in Japan were selected as subjects by two-stage stratified random sampling, and self-administered questionnaires were sent by mail between January and March 2007. Follow-up researches were conducted in the same way from January to March of 2008, 2009, 2010 and 2011. In this study, 1,006 working male in every research point was objected. Also, 3 point data of 2007 (Time1), 2009 (Time2) and 2011(Time3) were used. The analysis was performed using a cross-lagged model under structural equation modeling. The model of mediated effect of sense of coherence on the causal relationship that psychosocial work characteristics affect mental health was accepted (χ2/df=3.866, CFI=0.903, RMSEA=0.055). However, the direct causal relation between work characteristics and mental health was non-significant. These results support the Antonovsky's theory 'Salutogenic Model'. The sense of coherence will be key concept in workplace environmental arrangement for improvement in worker's mental health.

View full abstract

     This article aimed to examine the determinants of subjective social status using panel/longitudinal data analyses. The first purpose was to illustrate the impact of class position on subjective social status in consideration of distinction between “between–subject” effects and “within–subject” effects. The second purpose was to show the analytical example of “hybrid model”, which means an application of multilevel modeling.Japanese Life course Panel Survey (JLPS) datasets drawn from first wave to fourth wave survey were used in empirical analyses. Fixed effect models and hybrid models were applied to those datasets. In addition, the performances of both models were compared by monte-carlo simulation method. The effect of intragenerational class mobility on subjective social status was partly supported in empirical analyses. Besides, some differences between “within–subject” regression coefficient and “between–subject” coefficient were found. According to the results of monte-carlo simulation, the performances of hybrid models were almost equivalent to those of fixed effect models. These findings gave suggestion that hybrid model is the one of useful options for analyzing panel/longitudinal data.

View full abstract

     In this paper, I review some methodological challenges that arise when sociologists collect and manage panel data in Japan, and consider the ideal methods for them. I show three points of argument about data collection. First, you need to utilize retrospective methods even in panel surveys. Second, calendar interviewing facilitates accurate retrieval from memory. Third, computer assisted interview (CAI) is highly efficient for collecting panel data for many reasons. Next, I discuss two problems about management of panel data. First, I insist that data editing of panel data ought to follow the principles shown by Fellegi and Hout. Second, survey designers of panel data are responsible for constructing data with high usability. Finally, I insist that introduction of CAI is the most important point and it is the keystone which enables effective discussions on ideal collection and management of panel data in Japan.

View full abstract

     As the causes of intra-generational mobility, some external structural variables have been the focus of attention in social inequality studies; while in occupational attitudes studies, it has been assumed that some inherent attitudinal variables have an impact on job mobility, which is one side of intra-generational mobility. This paper investigates (1) whether workers' occupational attitudes actually affect the separations from their jobs, and (2) whether they do the mobility from worker's stratifications, which is other side of intra-generational mobility. Using panel data that traced male samples over a long period of 27 years (1979-2006) makes it possible to test the effects of workers' occupational attitudes on their intra-generational mobility, which have not yet been demonstrated. In my analysis, I distinguish between Job Mobility from one's company and Class Mobility from one's occupational stratification, and employ occupational commitment as the occupational attitude. The results of the analysis using a discrete-time logit model show that men's occupational commitment in 1979, which is the start time of that panel-observation, decreased the likelihood of their Class Mobility after that. On the other hand, it does not an explicit effect on their Job Mobility. These findings imply that in a period of post economic growth, men's occupational attitudes played the more important role in fluidity and stability of social mobility rather than of those in the labor market.

View full abstract

     The Gini coefficient is one of the indices most typically used to evaluate inequality amongst samples. However, it is inappropriate for evaluating inequality when negative samples are included in the population. This is because a precondition of the coefficient is that all samples must be positive. In order to enable the coefficient to evaluate inequality in such cases, Chen et al. attempted to extend it. However, their extended coefficient cannot be used to evaluate inequality when the sum of the samples is less than 0. In this paper, we extended the geometric expression of the Gini coefficient in order to enable it to be used to evaluate inequality even if the sum of the samples is less than 0. In addition, we confirmed that the extended coefficient was exactly equal to the Gini coefficient when all of the samples in the population were positive. Furthermore, we considered the algebraic expression of the extended coefficient, and using this, we clarified that the extended coefficient satisfied both Pigou-Dalton's population principle and the extended transfer principle.

View full abstract

     In this paper, I explore the relation between important concepts, Myerson value developed in economics (bargaining theory) and betweenness centrality widely used in sociology. In the communication game in which a value is generated only from a connected pair of players, I give a network-based interpretation for the Myerson value of a player as his average betweenness centrality in a network. I also propose various concepts of betweenness centrality (betweenness centrality, weighted betweenness centrality and generalized weighted betweenness centrality), and show that generalized weighted betweenness centrality converges to weighted betweenness centrality as the discount factor approaches one.

View full abstract

     This study examines correlation between unemployment rates and juvenile homicide rates for 16-17 and 18-19 age groups during 1974-2008 in Japan, using cointegration regression and error correction models. Since the publication of Yuma, Kanazawa, and Yuma (2010), two more yearly data sets become available and we include them in this study. For 16-17 age group, our results are consistent with our previous study. For 18-19 age group, on the other hand, we find its homicide rate significantly dropped in 2000, while the effect of unemployment on homicide-higher unemployment rate is associated with higher homicide rate-have been consistently observed during the research period as in our previous study. These results are discussed in terms of the dramatic rise in Japanese suicide rate since 1998 and the public opinion calling for more severe punishment since the late 90s.

View full abstract