ESRA 2019 Programme at a Glance
Analysing the Lives of LGBTI People - Survey Approaches to LGBTI Persons, Couples and Families 1
|Session Organisers|| Dr Stephanie Steinmetz (University of Amsterdam)
Ms Mirjam Fischer (University of Amsterdam)
Dr Nancy Bates (U.S. Census Bureau)
|Time||Wednesday 17th July, 09:00 - 10:30|
In recent years, much progress has been made in the US, Europe and beyond with regard to legislation that is supportive and protective of LGBTs (Lesbian, Gay, Bisexual, Trans persons). While these achievements are laudable, it is important to keep evaluating to what extent structural obstacles to equality remain. Compared to research on other minority groups, sexual minorities have been studied quantitatively much less in the social sciences. Yet, scholars have continuously made efforts to overcome the methodological challenges associated with studying this population quantitatively. This is an important development which should be encouraged and continued.
This session invites contributions showcasing research around the challenges, successes, and best practices when collecting data on sexual minorities. In particular contributions addressing one of the four areas are welcome:
1) How to sample LGBT populations? (e.g. what are common strategies for designing sampling frames intended at capturing LGBT populations? Which advantages and disadvantages in terms of data quality can be detected?);
2) How to measure sexual orientation and gender identity (SOGI) in large-scale, general-population surveys and polls? (e.g. Can sexual orientation be collected by proxy in surveys that use a single household informant? How can issues of cross-cultural validity, language and interviewer effects be addressed? Does the addition of SOGI items harm unit response rates in surveys that do not typically collect such items?);
3)How to estimate the size of LGBT populations (e.g. are there besides surveys alternative inventive methods, such as administrative records or internet web-scraping, for producing valid LGBT population estimates? In the absence of large representative demographic survey data, are there ways to extrapolate from non-random, small area, snowball, or convenience samples?;
4) as issues around SOGI are increasingly visible on the agenda of governments and governmental organizations around the globe the session also invites contributions which highlight best practices how the SOGI topic can most efficiently and successfully be approached?
In addition, this session also invites submissions that focus on topical survey results around LGBT populations such as physical and mental health disparities, income inequality, hate crimes, and household and family structures. The session hopes to draw a cross-section of submissions from different countries and different survey experiences.
Keywords: hard-to-reach populations, SOGI, samplingand estimating LGBT populations
Using Social Media to Increase Sexual and Gender Minority Youth While Maintaining Generalizability: Sampling and Weighting Issues and Solutions
Dr Marcus Berzofsky (RTI International) - Presenting Author
Dr Jill Dever (RTI International)
Ms Caroline Scruggs (RTI International)
Sexual and gender minority (SGM; i.e. LGBTQ+) youth (14 – 21 year olds) are often an undersampled population in representative surveys. This is because probability-based surveys of youth usually sample schools which may miss SGM youth who have dropped out or are not at school. This undercoverage of SGM youth is problematic when researching topics such as suicide ideation and attempt which, while critical to all youth, impact SGM youth at an even greater rate. We present a methodology to use survey data collected through social media which increases the proportion of SGM youth respondents to augment probability-based survey data while maintaining the generalizability of the combined data. The methodology covers the two key design areas: (1) the sampling approach to oversample SGM youth and (2) the weighting approach to create generalizable hybrid estimates.
For the sampling approach, the methodology utilizes two key characteristics. First, focus groups are used to identify keywords or terms highly correlated with SGM youth on social media. Second, the survey is administered through the social media’s advertising platform accounting the keywords.
For the weighting approach, we describe an assessment of multiple methods to produce survey weights for the non-probability social media data and the approach to produce hybrid estimates. We demonstrate differences between calibration and propensity scores to produce the survey weights. For the hybrid estimates, we show how using mean square error or unequal weighting effect to composite the probability and non-probability data impact the estimates.
We present the results of a pilot test of the methodology where we survey youth on Twitter and combined the data with the Youth Risk Behavioral Survey (YRBS). We present results on suicide ideation and attempt for SGM and non-SGM youth through each approach evaluated. In addition to the methodology, we will discuss our approach to use a more inclusive definition of
A Comparison of Social Media and Intercept Recruitment Approaches to Building and Maintaining a Panel of LGBT Respondents for a Longitudinal Survey
Dr Kristine Wiant (RTI International) - Presenting Author
Dr Jamie Guillory (RTI International)
Ms Leah Fiacco (RTI International)
Ms Jessica Pikowski (RTI International)
A key challenge in surveying LGBT (sexual and gender minorities including Lesbian, Gay, Bisexual, Transgender) populations is efficiently and effectively recruiting a sample. This presentation provides a case study of effective use of combining social media recruitment and intercept recruitment methods to build a panel of 18 to 24-year-olds from the LGBT community. The resulting panel is used in the Research and Evaluation Survey for the Public Education Campaign on Tobacco (RESPECT) study to evaluate the impact of a tobacco public education campaign among LGBT 18 to 24-year-olds. Data for the RESPECT study are collected via a web survey in 24 cities in the United States. The first six of seven planned waves of data collection were collected between spring of 2016 and fall of 2018. At each wave of data collection, new participants were recruited either through advertisements posted on Facebook or Instagram or through an intercept survey conducted in LGBT social venues. In this presentation, we first review social media and intercept recruitment and discuss operational challenges of each approach. Next, we compare recruitment results from these two approaches in terms of speed and cost of data collection, data quality, and sample retention over time. Finally, we discuss differences in sample composition and key survey measures from the survey data.
Are Sexual Minorities Hard-to-Survey? Insights from the 2020 Census Barriers, Attitudes and Motivators Survey (CBAMS)
Ms Nancy Bates (U.S. Census Bureau) - Presenting Author
Dr Yazmin Argen Garcia Trejo (US Census Bureau)
Ms Monica Vines (US Census Bureau)
In many societies, sexual minorities are considered a stigmatized or marginalized population. Lesbians, gays, and bisexuals may remain “closeted” choosing not to reveal sexual orientation for fear of denial of services, physical harm, social discrimination, and other negative outcomes. As a result, the LGB population is also often considered “hard-to-survey” due to perception that their answers could be used against them. Consequently, they may have a higher propensity to refuse participation in surveys or be less inclined to answer sexual orientation questions truthfully.
In 2018, the Census Bureau conducted a nationally representative survey. The survey collected data about the attitudes, barriers, and motivators for participating in the 2020 U.S. Census. This information was then used to inform the creative platform and targeted messaging for the 2020 Census communications campaign. As part of the demographic battery of questions, sexual orientation was asked as the last item in the survey. Of the 17,283 respondents, 568 self-identified as lesbian, gay or bisexual.
We reduced the 50+ survey items down to eight uncorrelated constructs. Among others, these included: trust in government, confidentiality concerns, civic engagement, and Census knowledge. Factor scores were calculated for each construct and assigned to each respondent. Scores were then entered into a cluster analysis to produce six Census “mindsets” – cohesive groups with similar attitudes and behaviors, yet distinct from one another.
We then examined the mindsets by sexual orientation. We hypothesized that sexual minorities would skew toward mindsets with below-average propensity to self-respond in the 2020 Census. In fact, the opposite was found – sexual minorities were overrepresented in the “Eager Engager” mindset – the mindset with the highest stated intent to respond to the 2020 Census. We discuss these findings in the context of messaging to sexual minorities in the 2020 Census as well
Surveying Persons in Same-Sex Relationships in a Probabilistic Way – An Example from the Netherlands
Dr Stephanie Steinmetz (University of Amsterdam) - Presenting Author
Ms Mirjam Fischer (University of Amsterdam)
The call for improved estimates of lesbians, gay men and bisexual (LGB) populations has grown steadily. Quantitative scholars continuously make efforts to overcome methodological challenges, such as the establishment of a large representative sample of the LGB population by means of probability sampling. The representativeness challenge is one of the most prominent ones in the field of quantitative LGB studies. The aim of this article is twofold. First, we present the sampling strategy of a recently conducted probability-based survey in the Netherlands which targeted persons in mixed-sex and same-sex couples with and without children (Unions in Context study (UNICON)). Secondly, we evaluate the representativeness of the collected data by comparing it to two national benchmark surveys.
With respect to the first question we can conclude that the strategy was indeed successful. Based on an innovative sampling strategy, we were able to collect a probability survey which allows the analyses of 1,373 individual respondents (in 935 households), including 518 persons in mixed-sex couples and 855 persons in same-sex couples. Regarding our second question the answer is less straightforward. As expected, even after the application of basic sample weights selectivity remains. A first challenge might arise from the web-mode of the survey. People who do not have easy access to a computer and the Internet, who are illiterate or do not fully master the Dutch language, are by default excluded from participating in the study. Second, non-response might be related to the way in which we have framed the main purpose of the study. Participation depends on the personal interest of people in the survey topic (diversity in living arrangements) and their motivation to support the research. In particular, LGB populations in the Netherlands are of high interest to researchers and politicians. This can lead to an alertness and sensitization of this group towards