Relevance & Research Question

The COVID-19 pandemic affected daily life in unprecedented ways, posing serious challenges for governments and societies. Nonpharmaceutical interventions (NPIs), such as stay-at-home orders, physical distancing measures, and mask mandates, were pivotal in reducing transmission, particularly during the early stages when vaccines were unavailable. Understanding how populations responded to these interventions was crucial for developing effective communication strategies and policies. However, the lack of comprehensive data on behaviors and perceptions during the pandemic posed a significant challenge. This study sought to fill this gap by investigating behavioral responses to COVID-19 across diverse demographic groups and countries, examining the interplay between threat perception, preventive behaviors, and compliance with NPIs.

Methods & Data

To explore these dynamics, we conducted the COVID-19 Health Behavior Survey, a large-scale, cross-national online survey administered across eight countries: Belgium, France, Germany, Italy, the Netherlands, Spain, the United Kingdom, and the United States. Data collection relied on targeted Facebook advertisements, enabling rapid recruitment of participants during the pandemic’s initial wave. The survey, conducted between March 13 and August 12, 2020, yielded over 140,000 responses. It captured detailed information on participants’ health status, behaviors, social contacts, and attitudes toward COVID-19. Statistical techniques were employed to address potential sampling biases and ensure robust insights.

Results

The results highlighted significant demographic and national differences in pandemic responses. Women and older individuals perceived COVID-19 as a greater threat than men and younger groups, leading to higher adoption rates of preventive measures such as mask-wearing and physical distancing. Threat perception was particularly influential among vulnerable populations, including the elderly and those with preexisting conditions. Social contact patterns also changed markedly, with physical distancing guidelines leading to a 48%-85% reduction in social contacts compared to pre-pandemic levels across surveyed countries, often exceeding the impact of lockdown measures.

Added Value

This study provides valuable cross-national insights into behavioral responses during a global health crisis. By leveraging innovative survey methods and timely data collection, it underscores the importance of understanding population behavior to inform public health strategies and enhance preparedness for future pandemics. The findings offer actionable guidance for evidence-based policy-making and effective risk communication.

Relevance & Research Question

Data availability is often limited in developing countries, with timely administrative or survey data especially lacking. To address this, we propose a novel survey recruitment and estimation approach. First, we recruit survey participants through Facebook advertisements. While social media surveys are common in high-income countries, they are less frequently used in contexts like Sub-Saharan Africa, where internet and Facebook penetration are low. We aim to assess the potential of this approach in such settings. Additionally, we explore the feasibility of a network reporting approach to estimate fertility rates.

Methods & Data

We used Meta’s advertising platform to recruit survey respondents, targeting Facebook users aged 18 and over in all 14 regions of Senegal. Data collection occurred over one week in October 2024. Our survey included a network reporting component, where respondents provided information about themselves and three people from their regular social network. This approach captures unique data typically inaccessible in standard Facebook surveys, including socio-demographic information such as age, gender, education level, and number and age of children. Our analysis aims to estimate birth rates, using data from the Demographic Health Survey (DHS) as a benchmark for sample composition and fertility rate accuracy.

Results

Our sample includes 350 respondents, with 44% women. About 24% live in Dakar, Senegal's capital, and 37% live in rural areas. The average respondent age is 33. On average, respondents reported contact with 10 people the previous day and provided detailed information for up to three. The network sample (n=567) is gender-balanced (50% women) with an average age of 30. About 23% of network members reportedly do not use Facebook. Further analysis will focus on fertility rate estimation, comparing our findings with DHS data to assess the reliability of our approach.

Added Value

Our study addresses data gaps in African fertility estimates and introduces a new data collection method using social media. By comparing our results with DHS data, we aim to evaluate the potential of this approach for providing timely fertility estimates in African contexts, thereby enhancing understanding of population trends in the region.

Relevance & Research Question

The digital age has transformed survey recruitment, with social media ads enabling cost-effective and rapid access to diverse and hard-to-reach populations. Despite this promise, this method raises critical challenges - one of these is the tendency for measurement errors in form of problematic response behaviors. These behaviors—including satisficing, low-effort responses, and fraudulent participation—threaten data quality. While some studies have shortly mentioned these issues, most social media-recruited surveys do not address them systematically. This raises the question: Should we worry about problematic responses in social media-recruited surveys? This study examines whether problematic respondents in social media-recruited surveys systematically differ from others and assesses their impact on data quality and model estimates. The study addresses two core questions: (1) Are problematic respondents systematically different in socio-demographics and substantive answers? (2) Do they bias multivariate model estimates?

Methods & Data

The study analyzes data from a web survey on labor market discrimination against women with headscarves in Germany, conducted in 2021/2024. Recruitment via targeted Facebook ads yielded 3,021 completed interviews at an average cost of €1.41 per respondent. Response quality was evaluated using indicators such as item non-response, straight-lining, speeding, and identity misrepresentation. Statistical tests, including chi-square, Fisher’s exact, and Mann-Whitney U-tests, were employed to identify significant differences. Multivariate regression models assessed the impact of problematic behaviors on key outcomes, such as perceived anti-Muslim discrimination.

Results

Preliminary findings reveal that problematic respondents differ significantly in socio-demographic composition and substantive answers. For example, behaviors like straight-lining and speeding are more frequent among younger and less-educated respondents. Multivariate analyses show that problematic responses distort key estimates, particularly on discrimination experiences. Cleaning the dataset improves model fit and the reliability of results, emphasizing the value of robust quality control measures.

Added Value

This study provides a systematic investigation of problematic response behaviors in social media-recruited surveys, shedding light on their prevalence, predictors, and implications for data quality. It underscores the necessity of incorporating quality control measures in future surveys, offering practical recommendations for researchers leveraging social media recruitment strategies.

Relevance & Research Question

With social media platforms increasingly utilized for participant recruitment, it’s critical to assess their effectiveness in building balanced non-probabilistic panels. This study investigates whether investing in targeted social media ads on platforms like Meta (Facebook and Instagram) can effectively balance recruitment costs and sample composition quality. We explore how different targeting criteria (such as age and gender) impact sample composition and evaluate the accuracy of platform-provided socio-demographic estimates. The study aims to understand the trade-offs between advertising budget allocation and sample representativeness, addressing the overarching question: do high-cost ad strategies improve recruitment outcomes?

Methods & Data
Our data stems from the GESIS Panel Digital Behavioral Data Sample (GESIS Panel.dbd) recruitment campaign, conducted on Meta platforms in May 2023. We employed a structured ad campaign with four distinct targeting groups based on gender and age, allowing for comparison of recruitment efficacy across various demographics. The recruitment process included a click-through ad link leading to a welcome survey, followed by a registration step to join the panel. Incentives were provided to participants who completed registration. Data analyses focus on the socio-demographic composition, recruitment costs, and response rates across ad groups, as well as the accuracy of Meta’s demographic targeting information.
Results

Our findings reveal that targeted social media recruitment can enhance demographic balance in certain respects but does not fully eliminate sampling biases. For instance, Meta’s age and gender targeting improved representation among older individuals but showed limitations for younger demographics. Additionally, while the socio-demographic estimates provided by Meta are generally reliable, slight misclassifications (around 5%) were observed. Cost analysis revealed that lower recruitment budgets yielded the most cost-effective samples, contradicting the notion that higher spending guarantees improved sample composition. Higher ad expenditures increased reach but also raised cost per participant, suggesting a strategic budget allocation is essential for optimal sample utilization.

Added Value

allocation is essential for optimal sample utilization.

Added Value:
This study offers practical insights for researchers leveraging social media platforms in recruitment, particularly in non-probabilistic settings. By illuminating the cost-structure dynamics and demographic accuracy of Meta’s ad-targeting tools, we provide guidelines for optimizing recruitment budgets without compromising sample quality.