As an emerging research area, application of satellite-based nighttime lights data in the social sciences has increased rapidly in recent years. This study, building on the recent surge in the use of satellite-based lights data, explores whether information provided by such data can be used to address attenuation bias in the estimated coefficient when the regressor variable, Gross Domestic Product (GDP), is measured with large error. Using an example of a study on infant mortality rates (IMRs) in the People’s Republic of China (PRC), this paper compares four models with different indicators of GDP as the regressor of IMR: (1) observed GDP alone, (2) lights variable as a substitute, (3) a synthetic measure based on weighted observed GDP and lights, and (4) GDP with lights as an instrumental variable. The results show that the inclusion of nighttime lights can reduce the bias in coefficient estimates compared with the model using observed GDP. Among the three approaches discussed, the instrumental-variable approach proves to be the best approach in correcting the bias caused by GDP measurement error and estimates the effect of GDP much higher than do the models using observed GDP. The study concludes that beyond the topic of this study, nighttime lights data have great potential to be used in other sociological research areas facing estimation bias problems due to measurement errors in economic indicators. The potential is especially great for those focusing on developing regions or small areas lacking high-quality measures of economic and demographic variables.
Respondent-driven sampling (RDS) is a chain-referral method for sampling members of hidden or hard-to-reach populations, such as sex workers, homeless people, or drug users, via their social networks. Most methodological work on RDS has focused on inference of population means under the assumption that subjects’ network degree determines their probability of being sampled. Criticism of existing estimators is usually focused on missing data: the underlying network is only partially observed, so it is difficult to determine correct sampling probabilities. In this article, the author shows that data collected in ordinary RDS studies contain information about the structure of the respondents’ social network. The author constructs a continuous-time model of RDS recruitment that incorporates the time series of recruitment events, the pattern of coupon use, and the network degrees of sampled subjects. Together, the observed data and the recruitment model place a well-defined probability distribution on the recruitment-induced subgraph of respondents. The author shows that this distribution can be interpreted as an exponential random graph model and develops a computationally efficient method for estimating the hidden graph. The author validates the method using simulated data and applies the technique to an RDS study of injection drug users in St. Petersburg, Russia.
Writings on the optimal length for survey questions are characterized by a variety of perspectives and very little empirical evidence. Where evidence exists, support seems to favor lengthy questions in some cases and shorter ones in others. However, on the basis of theories of the survey response process, the use of an excessive number of words may get in the way of the respondent’s comprehension of the information requested, and because of the cognitive burden of longer questions, there may be increased measurement errors. Results are reported from a study of reliability estimates for 426 (exactly replicated) survey questions in face-to-face interviews in six large-scale panel surveys conducted by the University of Michigan’s Survey Research Center. The findings suggest that, at least with respect to some types of survey questions, there are declining levels of reliability for questions with greater numbers of words and provide further support for the advice given to survey researchers that questions should be as short as possible, within constraints defined by survey objectives. Findings reinforce conclusions of previous studies that verbiage in survey questions—either in the question text or in the introduction to the question—has negative consequences for the quality of measurement, thus supporting the KISS principle ("keep it simple, stupid") concerning simplicity and parsimony.
"Rapport" has been used to refer to a range of positive psychological features of an interaction, including a situated sense of connection or affiliation between interactional partners, comfort, willingness to disclose or share sensitive information, motivation to please, and empathy. Rapport could potentially benefit survey participation and response quality by increasing respondents’ motivation to participate, disclose, or provide accurate information. Rapport could also harm data quality if motivation to ingratiate or affiliate causes respondents to suppress undesirable information. Some previous research suggests that motives elicited when rapport is high conflict with the goals of standardized interviewing. The authors examine rapport as an interactional phenomenon, attending to both the content and structure of talk. Using questions about end-of-life planning in the 2003–2005 wave of the Wisconsin Longitudinal Study, the authors observe that rapport consists of behaviors that can be characterized as dimensions of responsiveness by interviewers and engagement by respondents. The authors identify and describe types of responsiveness and engagement in selected question-answer sequences and then devise a coding scheme to examine their analytic potential with respect to the criterion of future study participation. The analysis suggests that responsive and engaged behaviors vary with respect to the goals of standardization; some behaviors conflict with these goals, whereas others complement them.
This article advances interviewing methods by introducing the authors’ original contribution: the iterated questioning approach (IQA). This interviewing technique augments the interviewer’s methodological arsenal by exploiting insights from symbolic interactionism, particularly Goffman’s concepts of frontstage and backstage. IQA consists of sequenced iterations of a baseline question designed to elicit multiple forms of talk. The approach consists of four distinct steps: (1) establishing the baseline iterated question, (2) eliciting frontstage talk, (3) going backstage, and (4) eliciting backstage talk. To illuminate IQA’s versatility, transcript excerpts are reproduced from interviews with two very different populations: disadvantaged high school students and business professionals. IQA promises to invigorate future interview-based inquiry by offering significant advantages compared with conventional interviewing procedures. IQA’s theoretically informed question design offers a more formalized and structured approach to gather interview data on identity-relevant themes. Capitalizing on Goffman’s dramaturgical framework, IQA produces readily classifiable forms of talk that correspond to frontstage and backstage self-presentations. As a result, IQA ensures replicability and allows interviewers to systematically analyze comparable talk within the same interview as well as across multiple respondents. For these reasons, IQA promises to be an innovative interviewing technique that pushes forward the methodological frontier.
The statistical description of the formation of marriages is hampered by the fact that the intensity of marriage of one sex depends on the available supply of potential spouses from the other. Unlike the situation that occurs in the study of fertility, there is no reason to give a preference to either of the sexes. To address the fundamental problem caused by the two-sex nature of the process, a solution is proposed that considers the sexes jointly. The solution relies on a novel use of generalized averages. The model is formulated in stochastic terms, and it is parametrized in terms of the overall level of nuptiality, the relative propensity of nuptiality by age, and the mutual relative attraction of spouses at different ages. Although national statistics collection relies on data aggregated by age groups, the models are formulated at individual levels to show how estimation could also be carried out in small populations. In particular, examples of how maximum likelihood estimation can be carried out are given for specific parametric models. The methods are illustrated by both monthly and annual nuptiality data from Finland.
Researchers often estimate income inequality by using data that give only the number of cases (e.g., families or households) whose incomes fall in "bins" such as $ 0 to $9,999, $10,000 to $14,999, . . . , ≥$200,000. We find that popular methods for estimating inequality from binned incomes are not robust in small samples, where popular methods can produce infinite, undefined, or arbitrarily large estimates. To solve these and other problems, we develop two improved estimators: a robust Pareto midpoint estimator (RPME) and a multimodel generalized beta estimator (MGBE). In a broad evaluation using U.S. national, state, and county data from 1970 to 2009, we find that both estimators produce very good estimates of the mean and Gini coefficient but less accurate estimates of the Theil index and mean log deviation. Neither estimator is uniformly more accurate, but the RPME is much faster, which may be a consideration when many estimates must be obtained from many data sets. We have made the methods available as the rpme command for Stata and the binequality package for R.
The authors investigate how reporting heterogeneity may bias socioeconomic and demographic disparities in self-rated general health, a widely used health indicator, and how such bias can be adjusted by using new anchoring vignettes designed in the 2012 wave of the China Family Panel Studies (CFPS). The authors find systematic variation by sociodemographic characteristics in thresholds used by respondents in rating their general health status. Such threshold shifts are often nonparallel in that the effect of a certain group characteristic on the shift is stronger at one level than another. The authors find that the resulting bias of measuring group differentials in self-rated health can be too substantial to be ignored. They demonstrate that the CFPS anchoring vignettes prove to be an effective survey instrument in obtaining bias-adjusted estimates of health disparities not only for the CFPS sample but also for an independent sample from the China Health and Retirement Longitudinal Study. Effective adjustment for reporting heterogeneity may require vignette administration only to a small subsample (20 percent to 30 percent of the full sample). Using a single vignette can be as effective as using more in terms of anchoring, but the results are sensitive to the choice of vignette design.
In the context of multilevel latent class models, the goodness-of-fit depends on multiple aspects, among which are two local independence assumptions. However, because of the lack of local fit statistics, the model and any issues relating to model fit can only be inspected jointly through global fit statistics. This hinders the search for model improvements, as it cannot be determined where misfit originates and which of the many model adjustments may improve its fit. Also, when relying solely on global fit statistics, assumption violations may become obscured, leading to wrong substantive results. In this paper, two local fit statistics are proposed to improve the understanding of the model, allow individual testing of the local independence assumptions, and inspect the fit of the higher level of the model. Through an application in which the local fit statistics group-variable residual and paired-case residual are used as guidance, it is shown that they pinpoint misfit, enhance the search for model improvements, provide substantive insight, and lead to a model with different substantive conclusions, which would likely not have been found when relying on global information criteria. Both residuals can be obtained in the user-friendly Latent GOLD 5.0 software package.
In recent decades, cultural consensus theory (CCT) models have been extensively applied across research domains to explore shared cultural knowledge and beliefs. These models are parameterized in terms of person-specific cognitive parameters such as abilities and guessing biases as well as item difficulties. Although psychometric test theory is also formalized in terms of abilities and item difficulties, a quality that clearly sets CCT models apart from other test theory models is their specification to operate on data in which the answer key is latent. In doing so, CCT models specify the answer key as parameters of the model, and also involved with this specification are procedures to verify the integrity of the answer key that is estimated. In this article, the authors develop the following methods to propagate the application of these CCT models in the field of social surveys: (1) by extending the underlying cognitive model to be able to account for uncertainty in decision making ("don’t know" responses), (2) by allowing covariate information to be entered in the analysis, and (3) by deriving statistical inference in the hierarchical Bayesian framework. The proposed model is fit to data describing knowledge on science and on aging to demonstrate the novel findings that can be achieved by the approach.
The standard multilevel regressions that are widely used in neighborhood research typically ignore potential between-neighborhood correlations due to underlying spatial processes, and hence they produce inappropriate inferences about neighborhood effects. In contrast, spatial models make estimations and predictions across areas by explicitly modeling the spatial correlations among observations in different locations. A better understanding of the strengths and limitations of spatial models as compared with the standard multilevel model is needed to improve the research on neighborhood and spatial effects. This research systematically compares model estimations and predictions for binary outcomes between (distance- and lattice-based) spatial and the standard multilevel models in the presence of both within- and between-neighborhood correlations, through simulations. Results from simulation analysis reveal that the standard multilevel and spatial models produce similar estimates of fixed effects but different estimates of random effects variances. Both the standard multilevel and pure spatial models tend to overestimate the corresponding random effects variances compared with hybrid models when both nonspatial within-neighborhood and spatial between-neighborhood effects exist. Spatial models also outperform the standard multilevel model by a narrow margin in case of fully out-of-sample predictions. Distance-based spatial models provide additional spatial information and have stronger predictive power than lattice-based models under certain circumstances. These merits of spatial modeling are exhibited in an empirical analysis of the child mortality data from 1880 Newark, New Jersey.
Change in group size and composition has long been an important area of research in the social sciences. Similarly, interest in interaction dynamics has a long history in sociology and social psychology. However, the effects of endogenous group change on interaction dynamics are a surprisingly understudied area. One way to explore these relationships is through social network models. Network dynamics may be viewed as a process of change in the edge structure of a network, in the vertex set on which edges are defined, or in both simultaneously. Although early studies of such processes were primarily descriptive, recent work on this topic has increasingly turned to formal statistical models. Although showing great promise, many of these modern dynamic models are computationally intensive and scale very poorly in the size of the network under study and/or the number of time points considered. Likewise, currently used models focus on edge dynamics, with little support for endogenously changing vertex sets. Here, the authors show how an existing approach based on logistic network regression can be extended to serve as a highly scalable framework for modeling large networks with dynamic vertex sets. The authors place this approach within a general dynamic exponential family (exponential-family random graph modeling) context, clarifying the assumptions underlying the framework (and providing a clear path for extensions), and they show how model assessment methods for cross-sectional networks can be extended to the dynamic case. Finally, the authors illustrate this approach on a classic data set involving interactions among windsurfers on a California beach.
Accurately measuring attributes in neighborhood environments allows researchers to study the influence of neighborhoods on individual-level outcomes. Researchers working to improve the measurement of neighborhood attributes generally advocate doing so in one of two ways: improving the theoretical relevance of measures and correctly defining the appropriate spatial scale. The data required by the first, "ecometric" neighborhood assessments on a sample of neighborhoods, are generally incompatible with the methods of the second, which tend to rely on population data. In this article, the authors describe how ecometric measures of theoretically relevant attributes observed on a sample of city blocks can be combined with a geostatistical method known as kriging to develop city block–level estimates across a city that can be configured to multiple neighborhood definitions. Using a cross-validation study with data from a 2002 systematic social observation of physical disorder on 1,663 city blocks in Chicago, the authors show that this method creates valid results. They then demonstrate, using neighborhood measures aggregated to three different spatial scales, that residents’ perceptions of both fear and neighborhood disorder vary substantially across different spatial scales.
Extreme response style (ERS) and acquiescence response style (ARS) are among the most encountered problems in attitudinal research. The authors investigate whether the response bias caused by these response styles varies with following three aspects of question format: full versus end labeling, numbering answer categories, and bipolar versus agreement response scales. A questionnaire was distributed to a random sample of 5,351 respondents from the Longitudinal Internet Studies for the Social Sciences household panel, of which a subsample was assigned to one of five conditions. The authors apply a latent class factor model that allows for diagnosing and correcting for ERS and ARS simultaneously. The results show clearly that both response styles are present in the data set, but ARS is less pronounced than ERS. With regard to format effects, the authors find that end labeling evokes more ERS than full labeling and that bipolar scales evoke more ERS than agreement style scales. With full labeling, ERS opposes opting for middle response categories, whereas end labeling distinguishes ERS from all other response categories. ARS did not significantly differ depending on test conditions.