Sociological Methods & Research

This article addresses the following question: Can speculation be used in social science research or should this not be an option? The secondary literature on speculation, which is minimal, is presented and discussed. It is noted that natural scientists often differentiate between a scientific form of speculation and the old metaphysical form of speculation. Following the lead of Charles Lave and James March in An Introduction to Models in the Social Sciences, close attention is also paid to the kind of speculation that is part of ordinary research, as opposed to the extraordinary type of speculation that can be found in the work of geniuses like Newton or Galileo. An attempt is then made to outline situations in which it is not only helpful but also necessary to speculate in social science. It is concluded that speculation is an important tool for social scientists if used with caution.

An innovative mixed-methods approach to exploratory focus group design is presented using a case study conducted with smallholder rice farmers in Vietnam. Understanding human decision-making under the uncertainties of a complex and changing social and environmental context requires a flexible yet structured and theoretically grounded approach. Using Bayesian belief networks as the architecture of our model allows the study to incorporate both qualitative and quantitative data, the former gathered at this stage in a participatory focus group setting and the latter to be collected in a subsequent survey. This framework further lends itself well to incorporating systematic behavioral approaches to decision-making analysis using Ajzen’s theory of planned behavior framework, a symbiosis that remains underexplored in the literature. The visual nature of the networks makes them easily accessible to participants, and the proposed technical solutions to field implementation are flexible, inexpensive, and shown in practice to mitigate issues of co-moderating discussion across language barriers. The tools and methods described are transparent, reproducible in comparative contexts, and transferable to a range of research topics and questions.

The historical turn in social science has prompted scholars to engage with the work of historians on a large scale. Here, social scientists face two standard problems of selection bias: confirmation bias and convenience sampling. So far, the record of dealing with these problems has been poor, and little has been done to specify how social scientists can sift a broader body of historiography. We present a criterial framework that describes how social scientists can mitigate bias when using historical studies. We term the idea behind this framework the Ulysses Principle because it can be understood as a way of avoiding the siren call of confirmation bias when using historical sources as the primary evidentiary base. The framework includes considerations about conceptual consistency, the theoretical vantage points of historical sources, and updated evidence. The three criteria and the trade-offs between them are illustrated using two recent examples from comparative historical analyses.

Since the early 2000s, the proliferation of cameras, whether in mobile phones or CCTV, led to a sharp increase in visual recordings of human behavior. This vast pool of data enables new approaches to analyzing situational dynamics. Application is both qualitative and quantitative and ranges widely in fields such as sociology, psychology, criminology, and education. Despite the potential and numerous applications of this approach, a consolidated methodological frame does not exist. This article draws on various fields of study to outline such a frame, what we call video data analysis (VDA). We discuss VDA’s research agenda, methodological forebears, and applications, introduce an analytic tool kit, and discuss criteria for validity. We aim to establish VDA as a methodological frame and an interdisciplinary analytic approach, thereby enhancing efficiency and comparability of studies, and communication among disciplines that employ VDA. This article can serve as a point of reference for current and future practitioners, reviewers, and interested readers.

Most research on images in surveys has examined how images affect answers to questions, which are known as visual context effects. Less frequently studied is how respondents process images in surveys and how variation in processing impacts observing visual context effects. Using data from two eye-tracking studies, we found that less than half of the respondents looked at the images in our surveys and that rates of looking at the images and time spent processing them significantly differed by gender. Additionally, we found that those who looked at the images for at least 60 milliseconds had their responses influenced by the images. The visual context effects did not occur for respondents who did not fixate on the images. Our findings suggest researchers must be cautious when placing images in surveys, given the possibility of visual context effects and variation of these effects by whether respondents fixate on the images, which could vary among subgroups.

Advances in computer science and computational linguistics have yielded new, and faster, computational approaches to structuring and analyzing textual data. These approaches perform well on tasks like information extraction, but their ability to identify complex, socially constructed, and unsettled theoretical concepts—a central goal of sociological content analysis—has not been tested. To fill this gap, we compare the results produced by three common computer-assisted approaches—dictionary, supervised machine learning (SML), and unsupervised machine learning—to those produced through a rigorous hand-coding analysis of inequality in the news (N = 1,253 articles). Although we find that SML methods perform best in replicating hand-coded results, we document and clarify the strengths and weaknesses of each approach, including how they can complement one another. We argue that content analysts in the social sciences would do well to keep all these approaches in their toolkit, deploying them purposefully according to the task at hand.

Identifying inattentive respondents in self-administered surveys is a challenging goal for survey researchers. Instructed response items (IRIs) provide a measure for inattentiveness in grid questions that is easy to implement. The present article adds to the sparse research on the use and implementation of attention checks by addressing three research objectives. In a first study, we provide evidence that IRIs identify respondents who show an elevated use of straightlining, speeding, item nonresponse, inconsistent answers, and implausible statements throughout a survey. Excluding inattentive respondents, however, did not alter the results of substantive analyses. Our second study suggests that respondents’ inattentiveness partially changes as the context in which they complete the survey changes. In a third study, we present experimental evidence that a mere exposure to an IRI does not negatively or positively affect response behavior within a survey. A critical discussion on using IRI attention checks concludes this article.

Periodic changes in occupational classifications make it difficult to obtain consistent measures of social class over time, potentially jeopardizing research on class-based trends. The severity of this problem depends, in part, on the measurement strategies used to address those changes. The authors propose that when a sample has been coded partly with one occupational classification and partly with another, Krippendorff’s index α be used to identify the best strategy for measuring class consistently across the two classifications and to assess the reliability of the class measure employed in the final analyses. This index can be computed regardless of the metric of the class variable; it can be used to compare measures based on different class schemes or that use different metrics; and statistical inference is straightforward, even with a complex sampling design. The authors put the index to work in conducting a case study of the effects of the switch from the 1970 to the 1980 U.S. Census Bureau Classification of Occupations on the reliability of Erikson–Goldthorpe–Portocarero class measures. Their findings indicate that measurement strategies that seem a priori equally reasonable vary substantially in terms of their reliability, and that the bulk of this variation is accounted for by the extent to which the strategies rely on subjective judgments about the relationships between occupational and class classifications. Most importantly, as long as the best-performing measurement strategies are used, the switch in occupational classifications appears to be substantially less consequential than has been previously argued. A computer program made available as a companion to the paper makes estimation of Krippendorff’s α, and statistical inference, very simple endeavors for nominal class variables.

Multifactorial survey experiments such as stated choice experiments are used more and more frequently in social science research. In this article, based on an experimental study on ethical and political consumption, we explore the potential of hybrid choice models to explicitly model latent psychological factors such as attitudes, overcoming a possible endogeneity bias and misrepresentation of causality. To this end, we employ a hybrid latent class choice model (HLCCM) in which the latent class structure allocates individuals to classes according to underlying latent attitudes that also influence the answers to attitudinal questions. This allows, in line with sociological action theories, a theory-guided testing of preference segmentation and modification caused by attitudes. We compare the complex HLCCM with less complex models that do not take the latent variable nature of attitudes into account and discuss in which cases less complex models might be more appropriate. However, the HLCCM always has the advantage of providing structure for theory testing and is therefore a useful tool to uncover preference heterogeneity, preference modification, and decision-making processes in sociological and other social science research.

This article is developed from the contents from one of the chapters the researcher has written as part of her PhD thesis. It discusses various methodological challenges the researcher had to face during the phase of data collection in the prisons of Kerala, India, and the strategies adopted to overcome these challenges. This article is intended as a guide for researchers who want to conduct a qualitative enquiry in prisons in the future.

Analysts of discrete data often face the challenge of managing the tendency of inflation on certain values. When treated improperly, such phenomenon may lead to biased estimates and incorrect inferences. This study extends the existing literature on single-value inflated models and develops a general framework to handle variables with more than one inflated value. To assess the performance of the proposed maximum likelihood estimator, we conducted Monte Carlo experiments under several scenarios for different levels of inflated probabilities under multinomial, ordinal, Poisson, and zero-truncated Poisson outcomes with covariates. We found that ignoring the inflations leads to substantial bias and poor inference of the inflations—not only for the intercept(s) of the inflated categories but other coefficients as well. Specifically, higher values of inflated probabilities are associated with larger biases. By contrast, the generalized inflated discrete models (GIDMs) perform well with unbiased estimates and satisfactory coverages even when the number of parameters that need to be estimated is quite large. We showed that model fit criteria, such as Akaike information criterion, could be used in selecting the appropriate specifications of inflated models. Lastly, the GIDM was implemented using large-scale health survey data as a comparison to conventional modeling approaches such as various Poisson and Ordered Logit models. We showed that the GIDM fits the data better in general. The current work provides a practical approach to analyze multimodal data that exists in many fields, such as heaping in self-reported behavioral outcomes, inflated categories of indifference and neutral in attitude surveys, large amounts of zero, and low occurrences of delinquent behaviors.

Markov models are important tools for quantitative social research. In this article, new methods for discrete Markov chains are presented. These methods allow us to calculate the distribution of the occupation time in a subset of the state space, the distribution of the waiting time to first entry into a subset of the state space, and the distribution of the waiting time to final exit from a subset of the state space. To demonstrate the usefulness of these methods, we apply them to working life tables for Spanish males to assess how the recent financial crisis affected the length of working life. The results show that the duration of working life decreased considerably, a pattern that can largely be explained by later entry to and earlier exit from the labor market. The findings also indicate that inequality in the length of working life increased.

We investigate the reliability of data from the Wage Indicator (WI), the largest online survey on earnings and working conditions. Comparing WI to nationally representative data sources for 17 countries reveals that participants of WI are not likely to have been representatively drawn from the respective populations. Previous literature has proposed to utilize weights based on inverse propensity scores, but this procedure was shown to leave reweighted WI samples different from the benchmark nationally representative data. We propose a novel procedure, building on covariate balancing propensity score, which achieves complete reweighting of the WI data, making it able to replicate the structure of nationally representative samples on observable characteristics. While rebalancing assures the match between WI and representative benchmark data sources, we show that the wage schedules remain different for a large group of countries. Using the example of a Mincerian wage regression, we find that in more than a third of the cases, our proposed novel reweighting assures that estimates obtained on WI data are not biased relative to nationally representative data. However, in the remaining 60 percent of the analyzed 95 data sets, systematic differences in the estimated coefficients of the Mincerian wage regression between WI and nationally representative data persist even after reweighting. We provide some intuition about the reasons behind these biases. Notably, objective factors such as access to the Internet or richness appear to matter, but self-selection (on unobservable characteristics) among WI participants appears to constitute an important source of bias.