AKArie Kapteyn

Arie Kapteyn

*This is an algorithmically generated author record based on publicly available information.

Current Organizations / Institutes

The organizations/institutes from the scholar's latest article

Previous Organizations / Institutes

The organizations / institutes from the scholar's other articles in the past ten years


Arie Kapteyn has a total of 113 co-authors. Here are the top co-authors by number of shared publications:


Latest Publications

  • Open Access

    Atypical survey response patterns as early indicators of dementia onset: a coordinated analysis across longitudinal aging cohorts

    by Stefan Schneider, Pey-Jiuan Lee, Doerte U Junghaenel, Erik Meijer, Raymond Hernandez, Haomiao Jin, Arthur A Stone, Arie Kapteyn, Bart Orriens, Hongxin Gao, et al.

    Objectives: Atypical survey response patterns, such as inconsistent or implausible responses, have been linked to lower cognitive functioning in later life. This study examined whether these behaviors predict future dementia onset across diverse longitudinal studies of aging. Methods: We conducted a coordinated analysis of data from eight longitudinal studies of aging (total N = 76,350). We derived 5 common types of atypical response pattern indicators from participants’ questionnaire data. Associations with subsequent dementia risk were examined within each cohort using Cox proportional hazards models adjusted for demographic covariates and accounting for death as a competing risk. Results were synthesized using random-effects meta-analysis methods. Results: Across cohorts, significant associations with incident dementia were observed for item-nonresponse (overall hazard ratio [HR] per SD increase = 1.06), random response errors (HR = 1.25), multivariate outlier responses (HR = 1.31), and incompatible response patterns indexed by Guttman errors and person-fit statistics (HRs = 1.22–1.25). Extreme responding was not significantly associated with dementia risk. Secondary analyses suggested stronger associations among individuals younger than 75 years and those with higher levels of education. Discussion: Routine survey responses contain behavioral signals associated with emerging dementia risk. Because these indicators can be derived from existing survey data without additional testing burden, they may provide a scalable complement to traditional approaches for studying cognitive aging in large population-based cohorts.

  • Open Access

    CAN RESPONSE TIME ADJUSTMENTS TO QUESTION DEMANDS TOGETHER WITH AVERAGE RESPONSE TIME DISTINGUISH BETWEEN LEVELS OF SATISFICING? FINDINGS FROM THREE SURVEYS IN A LARGE U.S. PANEL STUDY

    by Raymond Hernandez, Stefan Schneider, Erik Meijer, Titus Galama, Doerte U Junghaenel, Arie Kapteyn, Elizabeth Zelinski, Arthur A Stone

    Satisficing is a behavior that often results in careless responding and has been defined as engagement in simplified question responding approaches to reduce mental effort, often at the cost of response quality. Our objective was to examine the possibility of using a new survey response time (RT) based metric, response time adjustment, in combination with average survey RT, to identify groups with different levels of satisficing behavior. The RT adjustment parameter reflects the extent to which individuals adjust the amount of time they spend on survey items as a function of how demanding an item is, with low adjustment along with a low average survey RT expected to be associated with a greater likelihood of satisficing. We estimated a mixture model with RT adjustment and average RT as inputs using three questionnaires (with sample sizes of 5,321, 1,616, and 4,093, respectively) from the Understanding America Study (UAS), a large US Internet-based longitudinal panel. Weak and non-satisficing groups were identified in all three studies, with the former making up 3 to 9 percent of the samples, but no strong satisficing group was detected. Evidence supporting a weak satisficing group was based on good accuracy on easy attention check items (indicating they likely were not careless) but low accuracy on more demanding attention check items involving carefully reading long blocks of instructional text. Additionally, consistent with prior literature on satisficing, this weak satisficing group generally had lower mean values on cognitive ability and motivation-related variables (e.g., conscientiousness) compared to other groups. In two of the three studies, the satisficing group produced notable bias in study results involving high time intensity items. RT regulation and average RT may be useful for identifying satisficing in surveys with a mix of high and low time intensity items such as some tests and surveys with vignette-based items.

  • Open Access

    Raking for estimation and inference in panel models with nonignorable attrition and refreshment

    by Grigory Franguridi, Jinyong Hahn, Pierre Hoonhout, Arie Kapteyn, Geert Ridder

    In panel data subject to nonignorable attrition, auxiliary (refreshment) sampling may restore full identification under weak assumptions on the attrition process. Despite their generality, these identification strategies have seen limited empirical use, largely because the implied estimation procedure requires solving a functional minimization problem for the target density. We show that this problem can be solved using the iterative proportional fitting (raking) algorithm, which converges rapidly even with continuous and moderately high-dimensional data. This resulting density estimator is then used as input into a parametric moment condition. We establish consistency and convergence rates for both the raking-based density estimator and the resulting moment estimator when the distributions of the observed data are parametric. We also derive a simple recursive procedure for estimating the asymptotic variance. Finally, we demonstrate the satisfactory performance of our estimator in simulations and provide an empirical illustration using data from the Understanding America Study panel.

  • Open Access

    American Life in Realtime: Benchmark, publicly available person-generated health data for equity in precision health

    by Ritika R Chaturvedi, Marco Angrisani, Wendy M Troxel, Monika Jain, Tania Gutsche, Eva Ortega, Adrien Boch, Citina Liang, Shiyang Sima, Aziz Mezlini, et al.

    Person-generated health data (PGHD) from smartphones/wearables are invaluable for precision health, a field promoting health equity through tailored disease prevention, detection, and intervention strategies. However, pervasive convenience sampling in extant PGHD research introduces selection biases that systematically underrepresent disadvantaged groups, limit model generalizability, and risk exacerbating health disparities. Benchmark PGHD (representative, validated, longitudinal, and frequently repeated) are urgently needed to support model equity. To address this fieldwide limitation, we established American Life in Realtime (ALiR), a longitudinal population health study involving PGHD collected from a probability-based, nationally representative cohort using study-provided Fitbits and (as needed) 4G tablets. As a result, ALiR's 1,038 participants are broadly representative across comprehensive sociodemographic, behavioral, and health-related US population norms, overcoming disparities in established convenience samples (e.g. NIH's All of Us; AoU). Only two sources of differential enrollment remained: older age (odds ratio [OR]: 1.27, 99% CI: 1.12–1.45) during consent, lower education (OR: 0.86, 99% CI: 0.79–0.94) during enrollment, though oversampling individuals without bachelor's degrees sufficiently counterbalanced the latter. An illustrative coronavirus disease 2019 classification model—chosen for global significance, known disparities in experience and outcomes, and methodological relevance—trained using ALiR performed equivalently when tested in sample (area under the curve [AUC] = 0.84, 95% CI: 0.79–0.89) and out of sample on AoU (AUC = 0.83, 95% CI: 0.78–0.89) overall, and in historically underserved subgroups (AUC = 0.82–1.0). Conversely, an identically trained classification model using AoU underperformed by 35% out of sample on ALiR (overall AUC = 0.68, 95% CI: 0.61–0.75 vs. AUC = 0.93, 95% CI: 0.91–0.96 in sample), with worse performance in older female and non-White subgroups (by 22–40%). Our results suggest that probability sampling and hardware provisioning enabled cohort inclusivity and generalizable model performance, supporting ALiR's benchmarking potential for equitable recruitment, PGHD collection, and precision health application.


  • Evidence supports the validity and reliability of response times from a brief survey as a digital biomarker for processing speed in a large panel study

    by Raymond Hernandez, Arthur A Stone, Elizabeth Zelinski, Erik Meijer, Titus Galama, Jessica Faul, Arie Kapteyn, Doerte U Junghaenel, Haomiao Jin, Margaret Gatz, et al.

    Survey response times (RTs) have hitherto untapped potential to allow researchers to gain more detailed insights into the cognitive performance of participants in online panel studies. We examined if RTs recorded from a brief online survey could serve as a digital biomarker for processing speed. Data from 9893 adults enrolled in the nationally representative Understanding America Study were used in the analyses. Hypotheses included that people’s average survey RTs would have a large correlation with an established processing speed test, small to moderate correlations with other cognitive tests, and associations with functional impairment. We also hypothesized that survey RTs would have sensitivity to various participant characteristics comparable to the established processing speed test’s sensitivity (eg, similar standardized means by sex). Overall, results support the validity and reliability of people’s average RTs to survey items as a digital biomarker for processing speed. The correlation between survey RTs (reverse scored) and the formal processing speed test was 0.61 (P < .001), and small to moderate associations with most other cognitive and functional status measures were observed. Sensitivity of survey RTs to various participant characteristics was nearly identical to the formal processing speed test. Survey RTs may be useful as proxies for processing speed.


  • Does feedback from physical activity measurement devices influence physical activity? Evidence from a randomized controlled trial

    by Arie Kapteyn, Htay-Wah Saw, Jill Darling

    Wearable electronic devices are increasingly used to collect physical activity (PA) data. Most wearables provide PA feedback to users. The feedback has the potential for biasing measurements as users may increase their PA level in response to the feedback. On the other hand, the feedback can also be a desirable property as it can be employed to promote PA across settings. Studies examining the causal feedback effect independent of other factors have been limited. This study analyzed the causal effect of PA feedback provided by wearables. We implemented 4 field experiments over a period of 8 months. In each experiment, we recruited participants from a probability-based internet panel and asked them to wear continuously: (i) a PA device that does not provide feedback for 7 consecutive days (control condition), (ii) a PA device that provides feedback for another 7 consecutive days (treated condition), (iii) both devices for another 1 or 2 consecutive days. After 6 PM each experimental day, participants completed a short online survey asking them about their experiences participating in the study. Of 120 eligible participants assessed, 81 provided valid and complete data while 39 were lost to follow-up. Participants with valid and complete data (n=81) accumulated 7% more PA on a given day when they wore a PA device providing feedback relative to when they wore a PA device that does not provide feedback (p-value<0.001). The feedback effect was robust to the inclusion of additional factors that might influence PA. Use of research-grade PA devices that provide no feedback is warranted for studies whose primary goals are to collect population PA data with minimal measurement errors, while wearables with feedback are most suited for PA intervention studies. When using PA devices that provide feedback, one needs to be aware of the bias that may result.


  • Open Access

    Comparing Health Survey Data Cost and Quality Between Amazon’s Mechanical Turk and Ipsos’ KnowledgePanel: Observational Study

    by Patricia M Herman, Mary E Slaughter, Nabeel Qureshi, Tarek Azzam, David Cella, Ian D Coulter, Graham DiGuiseppi, Maria Orlando Edelen, Arie Kapteyn, Anthony Rodriguez, et al.

    Background: Researchers have many options for web-based survey data collection, ranging from access to curated probability-based panels, where individuals are selectively invited to join based on their membership in a representative population, to convenience panels, which are open for anyone to join. The mix of respondents available also varies greatly regarding representation of a population of interest and in motivation to provide thoughtful and accurate responses. Despite the additional dataset-building labor required of the researcher, convenience panels are much less expensive than probability-based panels. However, it is important to understand what may be given up regarding data quality for those cost savings. Objective: This study examined the relative costs and data quality of fielding equivalent surveys on Amazon’s Mechanical Turk (MTurk), a convenience panel, and KnowledgePanel, a nationally representative probability-based panel. Methods: We administered the same survey measures to MTurk (in 2021) and KnowledgePanel (in 2022) members. We applied several recommended quality assurance steps to enhance the data quality achieved using MTurk. Ipsos, the owner of KnowledgePanel, followed their usual (industry standard) protocols. The survey was designed to support psychometric analyses and included >60 items from the Patient-Reported Outcomes Measurement Information System (PROMIS), demographics, and a list of health conditions. We used 2 fake conditions (“syndomitis” and “chekalism”) to identify those more likely to be honest respondents. We examined the quality of each platform’s data using several recommended metrics (eg, consistency, reliability, representativeness, missing data, and correlations) including and excluding those respondents who had endorsed a fake condition and examined the impact of weighting on representativeness. Results: We found that prescreening in the MTurk sample (removing those who endorsed a fake health condition) improved data quality but KnowledgePanel data quality generally remained superior. While MTurk’s unweighted point estimates for demographics exhibited the usual mismatch with national averages (younger, better educated, and lower income), weighted MTurk data matched national estimates. KnowledgePanel’s point estimates better matched national benchmarks even before poststratification weighting. Correlations between PROMIS measures and age and income were similar in MTurk and KnowledgePanel; the mean absolute value of the difference between each platform’s 137 correlations was 0.06, and 92% were <0.15. However, correlations between PROMIS measures and educational level were dramatically different; the mean absolute value of the difference across these 17 correlation pairs was 0.15, the largest difference was 0.29, and the direction of more than half of these relationships in the MTurk sample was the opposite from that expected from theory. Therefore, caution is needed if using MTurk for studies where educational level is a key variable. Conclusions: The data quality of our MTurk sample was often inferior to that of the KnowledgePanel sample but possibly not so much as to negate the benefits of its cost savings for some uses. International Registered Report Identifier (IRRID): RR2-10.1186/s12891-020-03696-2


  • Open Access

    Early Identification of Cognitive Impairment in Community Environments Through Modeling Subtle Inconsistencies in Questionnaire Responses: Machine Learning Model Development and Validation

    by Hongxin Gao, Stefan Schneider, Raymond Hernandez, Jenny Harris, Danny Maupin, Doerte U Junghaenel, Arie Kapteyn, Arthur Stone, Elizabeth Zelinski, Erik Meijer, et al.

    Background: The underdiagnosis of cognitive impairment hinders timely intervention of dementia. Health professionals working in the community play a critical role in the early detection of cognitive impairment, yet still face several challenges such as a lack of suitable tools, necessary training, and potential stigmatization. Objective: This study explored a novel application integrating psychometric methods with data science techniques to model subtle inconsistencies in questionnaire response data for early identification of cognitive impairment in community environments. Methods: This study analyzed questionnaire response data from participants aged 50 years and older in the Health and Retirement Study (waves 8-9, n=12,942). Predictors included low-quality response indices generated using the graded response model from four brief questionnaires (optimism, hopelessness, purpose in life, and life satisfaction) assessing aspects of overall well-being, a focus of health professionals in communities. The primary and supplemental predicted outcomes were current cognitive impairment derived from a validated criterion and dementia or mortality in the next ten years. Seven predictive models were trained, and the performance of these models was evaluated and compared. Results: The multilayer perceptron exhibited the best performance in predicting current cognitive impairment. In the selected four questionnaires, the area under curve values for identifying current cognitive impairment ranged from 0.63 to 0.66 and was improved to 0.71 to 0.74 when combining the low-quality response indices with age and gender for prediction. We set the threshold for assessing cognitive impairment risk in the tool based on the ratio of underdiagnosis costs to overdiagnosis costs, and a ratio of 4 was used as the default choice. Furthermore, the tool outperformed the efficiency of age or health-based screening strategies for identifying individuals at high risk for cognitive impairment, particularly in the 50- to 59-year and 60- to 69-year age groups. The tool is available on a portal website for the public to access freely. Conclusions: We developed a novel prediction tool that integrates psychometric methods with data science to facilitate “passive or backend” cognitive impairment assessments in community settings, aiming to promote early cognitive impairment detection. This tool simplifies the cognitive impairment assessment process, making it more adaptable and reducing burdens. Our approach also presents a new perspective for using questionnaire data: leveraging, rather than dismissing, low-quality data.


  • Top Cited Papers
    Open Access

    The Understanding America Study (UAS)

    by Arie Kapteyn, Marco Angrisani, Jill Darling, Tania Gutsche

    Purpose The Understanding America Study (UAS) is a probability-based Internet panel housed at the Center for Economic and Social Research at the University of Southern California (USC). The UAS serves as a social and health sciences infrastructure for collecting data on the daily lives of US families and individuals. The collected information includes survey data, DNA from saliva samples, information from wearables, contextual and administrative linkages, ecological momentary assessments, self-recorded narratives and electronic records of financial transactions. The information collected focuses on a defining challenge of our time—identifying factors explaining racial, ethnic, geographic and socioeconomic disparities over the life course, including racial discrimination, inequalities in access to education and healthcare, differences in physical, economic and social environments, and, more generally, the various opportunities and obstacles one encounters over the life course. The UAS infrastructure aims to optimise engagement with the wider research community both in data dissemination and in soliciting input on content and methods. To encourage input from the research community, we have reserved 100 000 min of survey time per year for outside researchers, who can propose to add survey questions four times a year. Participants The UAS currently comprises about 15 000 US residents (including a 3500-person California oversample) recruited by Address-Based Sampling and provided with Internet-enabled tablets if needed. Surveys are conducted in English and Spanish. Findings to date Since the founding of the UAS in 2014, we have conducted more than 600 surveys, including a sequence of surveys collecting biennial information on health and retirement (the complete Health and Retirement Study instrument), 11 cognitive assessments, personality, knowledge and use of information on Social Security programme rules, work disability and subjective well-being. Several hundreds of papers have been published based on the collected data in the UAS. Studies include documentations of the mental health effects of the COVID-19 pandemic and how this varied across socioeconomic groups; comparisons of physical activity measured with accelerometers and by self-reports showing the dramatic biases in the latter; extensive studies have shown the power of using paradata in gauging cognitive change over time; several messaging experiments have shown the effectiveness of information provision on the quality of decision-making affecting well-being at older ages. Future plans The UAS national sample is planned to grow to 20 000 respondents by 2025, with subsamples of about 2500 African American, 2000 Asian and 3000 Hispanic participants and an oversample of rural areas. An increasing amount of non-interview data (contextual information, data from a suite of wearables and administrative linkages) is continually being added to the data files.


  • Open Access

    Comparing Health Survey Data Cost and Quality Between Amazon’s Mechanical Turk and Ipsos’ KnowledgePanel: Observational Study (Preprint)

    by Patricia M Herman, Mary E Slaughter, Nabeel Qureshi, Tarek Azzam, David Cella, Ian D Coulter, Graham DiGuiseppi, Maria Orlando Edelen, Arie Kapteyn, Anthony Rodriguez, et al.

    BACKGROUND Researchers have many options for web-based survey data collection, ranging from access to curated probability-based panels, where individuals are selectively invited to join based on their membership in a representative population, to convenience panels, which are open for anyone to join. The mix of respondents available also varies greatly regarding representation of a population of interest and in motivation to provide thoughtful and accurate responses. Despite the additional dataset-building labor required of the researcher, convenience panels are much less expensive than probability-based panels. However, it is important to understand what may be given up regarding data quality for those cost savings. OBJECTIVE This study examined the relative costs and data quality of fielding equivalent surveys on Amazon’s Mechanical Turk (MTurk), a convenience panel, and KnowledgePanel, a nationally representative probability-based panel. METHODS We administered the same survey measures to MTurk (in 2021) and KnowledgePanel (in 2022) members. We applied several recommended quality assurance steps to enhance the data quality achieved using MTurk. Ipsos, the owner of KnowledgePanel, followed their usual (industry standard) protocols. The survey was designed to support psychometric analyses and included >60 items from the Patient-Reported Outcomes Measurement Information System (PROMIS), demographics, and a list of health conditions. We used 2 fake conditions (“syndomitis” and “chekalism”) to identify those more likely to be honest respondents. We examined the quality of each platform’s data using several recommended metrics (eg, consistency, reliability, representativeness, missing data, and correlations) including and excluding those respondents who had endorsed a fake condition and examined the impact of weighting on representativeness. RESULTS We found that prescreening in the MTurk sample (removing those who endorsed a fake health condition) improved data quality but KnowledgePanel data quality generally remained superior. While MTurk’s unweighted point estimates for demographics exhibited the usual mismatch with national averages (younger, better educated, and lower income), weighted MTurk data matched national estimates. KnowledgePanel’s point estimates better matched national benchmarks even before poststratification weighting. Correlations between PROMIS measures and age and income were similar in MTurk and KnowledgePanel; the mean absolute value of the difference between each platform’s 137 correlations was 0.06, and 92% were <0.15. However, correlations between PROMIS measures and educational level were dramatically different; the mean absolute value of the difference across these 17 correlation pairs was 0.15, the largest difference was 0.29, and the direction of more than half of these relationships in the MTurk sample was the opposite from that expected from theory. Therefore, caution is needed if using MTurk for studies where educational level is a key variable. CONCLUSIONS The data quality of our MTurk sample was often inferior to that of the KnowledgePanel sample but possibly not so much as to negate the benefits of its cost savings for some uses. INTERNATIONAL REGISTERED REPORT RR2-10.1186/s12891-020-03696-2