More is not always better: An experimental individual-level validation of the randomized response technique and the crosswise model

Prevalence estimates for COVID-19-related health behaviors based on the cheating detection triangular model

BACKGROUND

Survey studies in medical and health sciences predominantly apply a conventional direct questioning (DQ) format to gather private and highly personal information. If the topic under investigation is sensitive or even stigmatizing, such as COVID-19-related health behaviors and adherence to non-pharmaceutical interventions in general, DQ surveys can lead to nonresponse and untruthful answers due to the influence of social desirability bias (SDB). These effects seriously threaten the validity of the results obtained, potentially leading to distorted prevalence estimates for behaviors for which the prevalence in the population is unknown. While this issue cannot be completely avoided, indirect questioning techniques (IQTs) offer a means to mitigate the harmful influence of SDB by guaranteeing the confidentiality of individual responses. The present study aims at assessing the validity of a recently proposed IQT, the Cheating Detection Triangular Model (CDTRM), in estimating the prevalence of COVID-19-related health behaviors while accounting for cheaters who disregard the instructions.

METHODS

In an online survey of 1,714 participants in Taiwan, we obtained CDTRM prevalence estimates via an Expectation-Maximization algorithm for three COVID-19-related health behaviors with different levels of sensitivity. The CDTRM estimates were compared to DQ estimates and to available official statistics provided by the Taiwan Centers for Disease Control. Additionally, the CDTRM allowed us to estimate the share of cheaters who disregarded the instructions and adjust the prevalence estimates for the COVID-19-related health behaviors accordingly.

RESULTS

For a behavior with low sensitivity, CDTRM and DQ estimates were expectedly comparable and in line with official statistics. However, for behaviors with medium and high sensitivity, CDTRM estimates were higher and thus presumably more valid than DQ estimates. Analogously, the estimated cheating rate increased with higher sensitivity of the behavior under study.

CONCLUSIONS

Our findings strongly support the assumption that the CDTRM successfully controlled for the validity-threatening influence of SDB in a survey on three COVID-19-related health behaviors. Consequently, the CDTRM appears to be a promising technique to increase estimation validity compared to conventional DQ for health-related behaviors, and sensitive attributes in general, for which a strong influence of SDB is to be expected.

The Crosswise Model for Surveys on Sensitive Topics: A General Framework for Item Selection and Statistical Analysis

When surveys contain direct questions about sensitive topics, participants may not provide their true answers. Indirect question techniques incentivize truthful answers by concealing participants' responses in various ways. The Crosswise Model aims to do this by pairing a sensitive target item with a non-sensitive baseline item, and only asking participants to indicate whether their responses to the two items are the same or different. Selection of the baseline item is crucial to guarantee participants' perceived and actual privacy and to enable reliable estimates of the sensitive trait. This research makes the following contributions. First, it describes an integrated methodology to select the baseline item, based on conceptual and statistical considerations. The resulting methodology distinguishes four statistical models. Second, it proposes novel Bayesian estimation methods to implement these models. Third, it shows that the new models introduced here improve efficiency over common applications of the Crosswise Model and may relax the required statistical assumptions. These three contributions facilitate applying the methodology in a variety of settings. An empirical application on attitudes toward LGBT issues shows the potential of the Crosswise Model. An interactive app, Python and MATLAB codes support broader adoption of the model.

More than random responding: Empirical evidence for the validity of the (Extended) Crosswise Model

The Randomized Response Technique (Warner, Journal of the American Statistical Association, 60, 63-69, 1965) has been developed to control for socially desirable responses in surveys on sensitive attributes. The Crosswise Model (CWM; Yu et al., Metrika, 67, 251-263, 2008) and its extension, the Extended Crosswise Model (ECWM; Heck et al., Behavior Research Methods, 50, 1895-1905, 2018), are advancements of the Randomized Response Technique that have provided promising results in terms of improved validity of the obtained prevalence estimates compared to estimates based on conventional direct questions. However, recent studies have raised the question as to whether these promising results might have been primarily driven by a methodological artifact in terms of random responses rather than a successful control of socially desirable responding. The current study was designed to disentangle the influence of successful control of socially desirable responding and random answer behavior on the validity of (E)CWM estimates. To this end, we orthogonally manipulated the direction of social desirability (undesirable vs. desirable) and the prevalence (high vs. low) of sensitive attributes. Our results generally support the notion that the ECWM successfully controls social desirability bias and is inconsistent with the alternative account that ECWM estimates are distorted by a substantial influence of random responding. The results do not rule out a small proportion of random answers, especially when socially undesirable attributes with high prevalence are studied, or when high randomization probabilities are applied. Our results however do rule out that random responding is a major factor that can account for the findings attesting to the improved validity of (E)CWM as compared with DQ estimates.

Refinement of the extended crosswise model with a number sequence randomizer: Evidence from three different studies in the UK

The Extended Crosswise Model (ECWM) is a randomized response model with neutral response categories, relatively simple instructions, and the availability of a goodness-of-fit test. This paper refines this model with a number sequence randomizer that virtually precludes the possibility to give evasive responses. The motivation for developing this model stems from a strategic priority of WADA (World Anti-Doping Agency) to monitor the prevalence of doping use by elite athletes. For this model we derived a maximum likelihood estimator that allows for binary logistic regression analysis. Three studies were conducted on online platforms with a total of over 6, 000 respondents; two on controlled substance use and one on compliance with COVID-19 regulations in the UK during the first lockdown. The results of these studies are promising. The goodness-of-fit tests showed little to no evidence for response biases, and the ECWM yielded higher prevalence estimates than direct questions for sensitive questions, and similar ones for non-sensitive questions. Furthermore, the randomizer with the shortest number sequences yielded the smallest response error rates on a control question with known prevalence.

Hidden figures: Revisiting doping prevalence estimates previously reported for two major international sport events in the context of further empirical evidence and the extant literature

Background

High levels of admitted doping use (43.6% and 57.1%) were reported for two international sport events in 2011. Because these are frequently referenced in evaluating aspects of anti-doping, having high level of confidence in these estimates is paramount.

Objectives

In this study, we present new prevalence estimates from a concurrently administered method, the Single Sample Count (SSC), and critically review the two sets of estimates in the context of other doping prevalence estimates.

Methods

The survey featuring the SSC model was completed by 1,203 athletes at the 2011 World Championships in Athletics (WCA) (65.3% of all participating athletes) and 954 athletes at the 2011 Pan-Arab Games (PAG) (28.2% of all participating athletes). At WCA, athletes completed both UQM and SSC surveys in randomised order. At PAG, athletes were randomly allocated to one of the two surveys. Doping was defined as "having knowingly violated anti-doping regulations by using a prohibited substance or method."

Results

Estimates with the SSC model for 12-month doping prevalence were 21.2% (95% CI: 9.69-32.7) at WCA and 10.6% (95% CI: 1.76-19.4) at PAG. Estimated herbal, mineral, and/or vitamin supplements use was 8.57% (95% CI: 1.3-16.11) at PAG. Reliability of the estimates were confirmed with re-sampling method (n = 1,000, 80% of the sample). Survey non-compliance (31.90%, 95%CI: 26.28-37.52; p < 0.0001) was detected in the WCA data but occurred to a lesser degree at PAG (9.85%, 95% CI: 4.01-15.69, p = 0.0144 and 11.43%, 95% CI: 5.31-11.55, p = 0.0196, for doping and nutritional supplement use, respectively). A large discrepancy between those previously reported from the UQM and the prevalence rate estimated by the SSC model for the same population is evident.

Conclusion

Caution in interpreting these estimates as bona fide prevalence rates is warranted. Critical appraisal of the obtained prevalence rates and triangulation with other sources are recommended over "the higher rate must be closer to the truth" heuristics. Non-compliance appears to be the Achilles heel of the indirect estimation models thus it should be routinely tested for and minimised. Further research into cognitive and behaviour aspects, including motivation for honesty, is needed to improve the ecological validity of the estimated prevalence rates.

Experimental validation of specialized questioning techniques in conservation

Conservation increasingly relies on social science tools to understand human behavior. Specialized questioning techniques (SQTs) are a suite of methods designed to reduce bias in social surveys and are widely used to collect data on sensitive topics, including compliance with conservation rules. Most SQTs have been developed in Western, industrialized, educated, rich, and democratic countries, meaning their suitability in other contexts may be limited. Whether these techniques perform better than conventional direct questioning is important for those considering their use. We designed an experiment to validate the performance of four SQTs (unmatched count technique, randomized response technique, crosswise model, and bean method) against direct questions when asking about a commonly researched sensitive behavior in conservation, wildlife hunting. We developed fictional characters, and for each method asked respondents to report the answers that each fictional character should give when asked if they hunt wildlife. We collected data from 609 individuals living close to protected areas in two different cultural and socioeconomic contexts (Indonesia and Tanzania) to quantify the extent to which respondents understood and followed SQT instructions and to explore the sociodemographic factors that influenced a correct response. Data were modeled using binomial general linear mixed models. Participants were more likely to refuse to answer questions asked using SQTs compared with direct questions. Model results suggested that SQTs were harder for participants to understand. Demographic factors (e.g., age and education level) significantly influenced response accuracy. When sensitive responses to sensitive questions were required, all SQTs (excluding the bean method) outperformed direct questions, demonstrating that SQTs can successfully reduce sensitivity bias. However, when reviewing each method, most respondents (59-89%) reported they would feel uncomfortable using them to provide information on their own hunting behavior, highlighting the considerable challenge of encouraging truthful reporting on sensitive topics. Our results demonstrate the importance of assessing the suitability of social science methods prior to their implementation in conservation contexts.

Assessing the effectiveness of indirect questioning techniques by detecting liars

In many fields of applied research, mostly in sociological, economic, demographic and medical studies, misreporting due to untruthful responding represents a nonsampling error that frequently occurs especially when survey participants are presented with direct questions about sensitive, highly personal or embarrassing issues. Untruthful responses are likely to affect the overall quality of the collected data and flaw subsequent analyses, including the estimation of salient characteristics of the population under study such as the prevalence of people possessing a sensitive attribute. The problem may be mitigated by adopting indirect questioning techniques which guarantee privacy protection and enhance respondent cooperation. In this paper, making use of direct and indirect questions, we propose a procedure to detect the presence of liars in sensitive surveys which allows researchers to evaluate the impact of untruthful responses on the estimation of the prevalence of a sensitive attribute. We first introduce the theoretical framework, then apply the proposal to the Warner randomized response method, the unrelated question model, the item count technique, the crosswise model and the triangular model. To assess the effectiveness of the procedure, a simulation study is carried out. Finally, the presence and the amount of liars is discussed in two real studies concerning racism and workplace mobbing.

Nothing but the truth? Effects of faking on the validity of the crosswise model

In self-reports, socially desirable responding threatens the validity of prevalence estimates for sensitive personal attitudes and behaviors. Indirect questioning techniques such as the crosswise model attempt to control for the influence of social desirability bias. The crosswise model has repeatedly been found to provide more valid prevalence estimates than direct questions. We investigated whether crosswise model estimates are also less susceptible to deliberate faking than direct questions. To this end, we investigated the effect of "fake good" instructions on responses to direct and crosswise model questions. In a sample of 1,946 university students, 12-month prevalence estimates for a sensitive road traffic behavior were higher and thus presumably more valid in the crosswise model than in a direct question. Moreover, "fake good" instructions severely impaired the validity of the direct questioning estimates, whereas the crosswise model estimates were unaffected by deliberate faking. Participants also reported higher levels of perceived confidentiality and a lower perceived ease of faking in the crosswise model compared to direct questions. Our results corroborate previous studies finding the crosswise model to be an effective tool for counteracting the detrimental effects of positive self-presentation in surveys on sensitive issues.

Asking sensitive questions in conservation using Randomised Response Techniques

Conservation increasingly seeks knowledge of human behaviour. However, securing reliable data can be challenging, particularly if the behaviour is illegal or otherwise sensitive. Specialised questioning methods such as Randomised Response Techniques (RRTs) are increasingly used in conservation to provide greater anonymity, increase response rates, and reduce bias. A rich RRT literature exists, but successfully navigating it can be challenging. To help conservationists access this literature, we summarise the various RRT designs available and conduct a systematic review of empirical applications of RRTs within (n = 32), and beyond conservation (n = 66). Our results show increased application of RRTs in conservation since 2000. We compare the performance of RRTs against known prevalence of the sensitive behaviour and relative to other questioning techniques to assess how successful RRTs are at reducing bias (indicated by securing higher estimates). Findings suggest that RRT applications in conservation were less likely than those in other disciplines to provide prevalence estimates equal to, or higher than those derived from direct questions. Across all disciplines, we found reports of non-compliance with RRT instructions were common, but rarely accounted for in study design or analysis. For the first time, we provide conservationists considering RRTs with evidence on what works, and provide guidance on how to develop robust designs suitable for conservation research contexts. We highlight when alternate methods should be used, how to increase design efficiency and improve compliance with RRT instructions. We conclude RRTs are a useful tool, but their performance depends on careful design and implementation.

Are social desirability scales desirable? A meta-analytic test of the validity of social desirability scales in the context of prosocial behavior

Social desirability (SD) scales have been used for decades in psychology and beyond. These scales are sought to measure individuals' tendencies to present themselves overly positive in self-reports, thus allowing to control for SD biases. However, research increasingly questions the validity of SD scales, proposing that SD scales measure substantive trait characteristics rather than response bias. To provide a large-scale empirical test of the validity of SD scales, we conducted a meta-analysis (k = 41; N = 8980) on the relation between SD scale scores and prosocial behavior in economic games (where acting in a prosocial manner is highly socially desirable). If SD scales measure what they are supposed to (namely, SD bias), they should be negatively linked to prosocial behavior; if SD scales measure socially desirable traits, they should be positively linked to prosocial behavior. Unlike both possibilities, the meta-analytic correlation between SD scores and prosocial behavior was close to zero, suggesting that SD scales neither clearly measure bias nor substantive traits. This conclusion was also supported by moderation analyses considering differences in the implementation of games and the SD scales used. The results further question the validity of SD scales with the implication that scholars and practitioners should refrain from using them.

Functionality of the Crosswise Model for Assessing Sensitive or Transgressive Behavior: A Systematic Review and Meta-Analysis

Tools for reliable assessment of socially sensitive or transgressive behavior warrant constant development. Among them, the Crosswise Model (CM) has gained considerable attention. We systematically reviewed and meta-analyzed empirical applications of CM and addressed a gap for quality assessment of indirect estimation models. Guided by the PRISMA protocol, we identified 45 empirical studies from electronic database and reference searches. Thirty of these were comparative validation studies (CVS) comparing CM and direct question (DQ) estimates. Six prevalence studies exclusively used CM. One was a qualitative study. Behavior investigated were substance use and misuse (k = 13), academic misconduct (k = 8), and corruption, tax evasion, and theft (k = 7) among others. Majority of studies (k = 39) applied the "more is better" hypothesis. Thirty-five studies relied on birthday distribution and 22 of these used P = 0.25 for the non-sensitive item. Overall, 11 studies were assessed as high-, 31 as moderate-, and two as low quality (excluding the qualitative study). The effect of non-compliance was assessed in eight studies. From mixed CVS results, the meta-analysis indicates that CM outperforms DQ on the "more is better" validation criterion, and increasingly so with higher behavior sensitivity. However, little difference was observed between DQ and CM estimates for items with DQ prevalence estimate around 50%. Based on empirical evidence available to date, our study provides support for the superiority of CM to DQ in assessing sensitive/transgressive behavior. Despite some limitations, CM is a valuable and promising tool for population level investigation.

Doping Prevalence in Competitive Sport: Evidence Synthesis with "Best Practice" Recommendations and Reporting Guidelines from the WADA Working Group on Doping Prevalence

BACKGROUND

The prevalence of doping in competitive sport, and the methods for assessing prevalence, remain poorly understood. This reduces the ability of researchers, governments, and sporting organizations to determine the extent of doping behavior and the impacts of anti-doping strategies.

OBJECTIVES

The primary aim of this subject-wide systematic review was to collate and synthesize evidence on doping prevalence from published scientific papers. Secondary aims involved reviewing the reporting accuracy and data quality as evidence for doping behavior to (1) develop quality and bias assessment criteria to facilitate future systematic reviews; and (2) establish recommendations for reporting future research on doping behavior in competitive sports to facilitate better meta-analyses of doping behavior.

METHODS

The Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines were used to identify relevant studies. Articles were included if they contained information on doping prevalence of any kind in competitive sport, regardless of the methodology and without time limit. Through an iterative process, we simultaneously developed a set of assessment criteria; and used these to assess the studies for data quality on doping prevalence, potential bias and reporting.

RESULTS

One-hundred and five studies, published between 1975 and 2019,were included. Doping prevalence rates in competitive sport ranged from 0 to 73% for doping behavior with most falling under 5%. To determine prevalence, 89 studies used self-reported survey data (SRP) and 17 used sample analysis data (SAP) to produce evidence for doping prevalence (one study used both SRP and SAP). In total, studies reporting athletes totaled 102,515 participants, (72.8% men and 27.2% women). Studies surveyed athletes in 35 countries with 26 involving athletes in the United States, while 12 studies examined an international population. Studies also surveyed athletes from most international sport federations and major professional sports and examined international, national, and sub-elite level athletes, including youth, masters, amateur, club, and university level athletes. However, inconsistencies in data reporting prevented meta-analysis for sport, gender, region, or competition level. Qualitative syntheses were possible and provided for study type, gender, and geographical region. The quality assessment of prevalence evidence in the studies identified 20 as "High", 60 as "Moderate", and 25 as "Low." Of the 89 studies using SRP, 17 rated as "High", 52 rated as "Moderate", and 20 rated as "Low." Of the 17 studies using SAP, 3 rated as "High", 9 rated as "Moderate", and 5 rated as "Low." Examining ratings by year suggests that both the quality and quantity of the evidence for doping prevalence in published studies are increasing.

CONCLUSIONS

Current knowledge about doping prevalence in competitive sport relies upon weak and disparate evidence. To address this, we offer a comprehensive set of assessment criteria for studies examining doping behavior data as evidence for doping prevalence. To facilitate future evidence syntheses and meta-analyses, we also put forward "best practice" recommendations and reporting guidelines that will improve evidence quality.

A guilt-free strategy increases self-reported non-compliance with COVID-19 preventive measures: Experimental evidence from 12 countries

Studies of citizens' compliance with COVID-19 preventive measures routinely rely on survey data. While such data are essential, public health restrictions provide clear signals of what is socially desirable in this context, creating a potential source of response bias in self-reported measures of compliance. In this research, we examine whether the results of a guilt-free strategy recently proposed to lessen this constraint are generalizable across twelve countries, and whether the treatment effect varies across subgroups. Our findings show that the guilt-free strategy is a useful tool in every country included, increasing respondents' proclivity to report non-compliance by 9 to 16 percentage points. This effect holds for different subgroups based on gender, age and education. We conclude that the inclusion of this strategy should be the new standard for survey research that aims to provide crucial data on the current pandemic.

Indirect questioning methods for sensitive survey questions: Modelling criminal behaviours among a prison population

Information such as the prevalence and frequency of criminal behaviour is difficult to estimate using standard survey techniques because of the tendency of respondents to withhold or misrepresent information. Social desirability bias is a significant threat to the validity of self-reported data, especially when supplied by persons such as sexual offenders or those convicted of theft or substance abuse. The randomized response approach is an alternative to the standard interview method and offers great potential for researchers in the field of criminal justice. By means of a survey of 792 prison inmates, incorporating both indirect and direct response techniques, we investigate if the prison population also has problems recognizing their participation in criminal acts such as theft, illicit drug use, violence against property, reckless driving and arson. Our research findings suggest that self-reported criminal behaviour among a prison population is affected by social desirability bias and that the behaviour considered is significantly associated with the severity of obsessive-compulsive symptoms. The results also demonstrate the inadequacy of traditional, yet widely used, direct questioning methods, and the great potential for indirect questioning techniques to advance policy formation and evaluation in the field of criminal behaviour.

On the validity of non-randomized response techniques: an experimental comparison of the crosswise model and the triangular model

Non-randomized response techniques (NRRTs) such as the crosswise model and the triangular model (CWM and TRM; Yu et al. Metrika, 67, 251-263, 2008) have been developed to control for socially desirable responding in surveys on sensitive personal attributes. We present the first study to directly compare the validity of the CWM and TRM and contrast their performance with a conventional direct questioning (DQ) approach. In a paper-pencil survey of 1382 students, we obtained prevalence estimates for two sensitive attributes (xenophobia and rejection of further refugee admissions) and one nonsensitive control attribute with a known prevalence (the first letter of respondents' surnames). Both NRRTs yielded descriptively higher prevalence estimates for the sensitive attributes than DQ; however, only the CWM estimates were significantly higher. We attribute the higher prevalence estimates for the CWM to its response symmetry, which is lacking in the TRM. Only the CWM provides symmetric answer options, meaning that there is no "safe" alternative respondents can choose to distance themselves from being carriers of the sensitive attribute. Prevalence estimates for the nonsensitive control attribute with known prevalence confirmed that neither method suffered from method-specific bias towards over- or underestimation. Exploratory moderator analyses further suggested that the sensitive attributes were perceived as more sensitive among politically left-oriented than among politically right-oriented respondents. Based on our results, we recommend using the CWM over the TRM in future studies on sensitive personal attributes.

Drivers of Acceptance of COVID-19 Proximity Tracing Apps in Switzerland: Panel Survey Analysis

BACKGROUND

Digital proximity tracing apps have been released to mitigate the transmission of SARS-CoV-2, the virus known to cause COVID-19. However, it remains unclear how the acceptance and uptake of these apps can be improved.

OBJECTIVE

This study aimed to investigate the coverage of the SwissCovid app and the reasons for its nonuse in Switzerland during a period of increasing incidence of COVID-19 cases.

METHODS

We collected data between September 28 and October 8, 2020, via a nationwide online panel survey (COVID-19 Social Monitor, N=1511). We examined sociodemographic and behavioral factors associated with app use by using multivariable logistic regression, whereas reasons for app nonuse were analyzed descriptively.

RESULTS

Overall, 46.5% (703/1511) of the survey participants reported they used the SwissCovid app, which was an increase from 43.9% (662/1508) reported in the previous study wave conducted in July 2020. A higher monthly household income (ie, income >CHF 10,000 or >US $11,000 vs income ≤CHF 6000 or

CONCLUSIONS

Eliminating technical hurdles and communicating the benefits of digital proximity tracing apps are crucial to promote further uptake and adherence of such apps and, ultimately, enhance their effectiveness to aid pandemic mitigation strategies.

Collapse

Key Words

• COVID-19
• SARS-CoV-2
• acceptance
• communication
• compliance
• digital contact tracing
• digital proximity tracing
• mHealth
• tracing
• uptake
• usability

Collapse

MESH Headings

• Adult
• Aged
• COVID-19/complications
• COVID-19/psychology
• COVID-19/transmission
• Contact Tracing/instrumentation
• Contact Tracing/trends
• Female
• Humans
• Male
• Middle Aged
• Mobile Applications/standards
• Mobile Applications/statistics & numerical data
• Physical Distancing
• Surveys and Questionnaires
• Switzerland

Collapse

Grants Collapse

Affiliation(s)

Viktor von Wyl Epdemiology, Biostatistics & Prevention Institute, University of Zurich, Zürich, Switzerland Institute for Implementation Science in Health Care, University of Zurich, Zürich, Switzerland
Marc Höglinger Winterthur Institute of Health Economics, Zurich University of Applied Sciences, Winterthur, Switzerland
Chloé Sieber Epdemiology, Biostatistics & Prevention Institute, University of Zurich, Zürich, Switzerland
Marco Kaufmann Epdemiology, Biostatistics & Prevention Institute, University of Zurich, Zürich, Switzerland
André Moser Epdemiology, Biostatistics & Prevention Institute, University of Zurich, Zürich, Switzerland
Miquel Serra-Burriel Epdemiology, Biostatistics & Prevention Institute, University of Zurich, Zürich, Switzerland
Tala Ballouz Epdemiology, Biostatistics & Prevention Institute, University of Zurich, Zürich, Switzerland
Dominik Menges Epdemiology, Biostatistics & Prevention Institute, University of Zurich, Zürich, Switzerland
Anja Frei Epdemiology, Biostatistics & Prevention Institute, University of Zurich, Zürich, Switzerland
Milo Alan Puhan Epdemiology, Biostatistics & Prevention Institute, University of Zurich, Zürich, Switzerland

Collapse

Estimating the proportion of non-heterosexuals in Taiwan using Christofides' randomized response model: A comparison of different estimation methods

In social research, investigating sensitive, highly personal or embarrassing issues by means of standard survey techniques based on direct questioning leads to refusals to answer or false responses which, generally, flaw the validity of the analyses and produce incorrect inferences. To correct biases induced by nonresponse or underreporting of sensitive matters, Warner (1965) introduced an indirect questioning approach, known as the randomized response technique, which allows researchers to estimate the proportion of individuals with sensitive attributes or behaviors, while ensuring respondents' privacy protection. In this article, we consider the randomized response model proposed by Christofides (2003) and, through a simulation and an empirical study, compare different estimation methods for the prevalence of a sensitive attribute. Specifically, we discuss how the model has been implemented in a pilot study to collect data and derive maximum likelihood and Bayesian estimates for the proportion of non-heterosexuals aged 20 years or older for the Taiwanese population and for some subgroups of it by sex and age. Our analysis, and in particular the Bayesian approach, seems to meet the expectation of social researchers and experts of sexual behaviors. In fact, the produced estimates are higher than official findings in Taiwan obtained by direct questioning in face-to-face interviews and provide a more reliable picture of sexual identity in the country. Moreover, Bayesian estimates appear more accurate than those produced by the method of moment and the maximum likelihood method.

Controlling social desirability bias: An experimental investigation of the extended crosswise model

Indirect questioning techniques such as the crosswise model aim to control for socially desirable responding in surveys on sensitive personal attributes. Recently, the extended crosswise model has been proposed as an improvement over the original crosswise model. It offers all of the advantages of the original crosswise model while also enabling the detection of systematic response biases. We applied the extended crosswise model to a new sensitive attribute, campus islamophobia, and present the first experimental investigation including an extended crosswise model, and a direct questioning control condition, respectively. In a paper-pencil questionnaire, we surveyed 1,361 German university students using either a direct question or the extended crosswise model. We found that the extended crosswise model provided a good model fit, indicating no systematic response bias and allowing for a pooling of the data of both groups of the extended crosswise model. Moreover, the extended crosswise model yielded significantly higher estimates of campus Islamophobia than a direct question. This result could either indicate that the extended crosswise model was successful in controlling for social desirability, or that response biases such as false positives or careless responding have inflated the estimate, which cannot be decided on the basis of the available data. Our findings highlight the importance of detecting response biases in surveys implementing indirect questioning techniques.

Can detailed instructions and comprehension checks increase the validity of crosswise model estimates?

The crosswise model is an indirect questioning technique designed to control for socially desirable responding. Although the technique has delivered promising results in terms of improved validity in survey studies of sensitive issues, recent studies have indicated that the crosswise model may sometimes produce false positives. Hence, we investigated whether an insufficient understanding of the crosswise model instructions might be responsible for these false positives and whether ensuring a deeper understanding of the model and surveying more highly educated respondents reduces the problem of false positives. To this end, we experimentally manipulated the amount of information respondents received in the crosswise model instructions. We compared a crosswise model condition with only brief instructions and a crosswise model condition with detailed instructions and additional comprehension checks. Additionally, we compared the validity of crosswise model estimates between a higher- and a lower-educated subgroup of respondents. Our results indicate that false positives among highly educated respondents can be reduced when detailed instructions and comprehension checks are employed. Since false positives can also occur in direct questioning, they do not appear to be a specific flaw of the crosswise model, but rather a more general problem of self-reports on sensitive topics. False negatives were found to occur for all questioning techniques, but were less prevalent in the crosswise model than in the direct questioning condition. We highlight the importance of comprehension checks when applying indirect questioning and emphasize the necessity of developing instructions suitable for lower-educated respondents.