Relevance & Research Question: Researchers increasingly use digital trace data sources such as Google Trends as an alternative or complement to survey data. However, besides technical limitations and issues of external and internal validity, several researchers have noticed issues with Google Trends’ reliability. The data are based on an unknown sample of all Google searches. Downloading Google Trends data for the exact same parameters (i.e., search term, region, time) but at different points in time can therefore produce unreliable values on Google Trends’ search index, especially for queries with low search volume. In this paper, we extend the research on Google Trends’ reliability beyond the retrieval date by examining the effect of the download location on inconsistencies across samples: Do we get different values from Google Trends depending on where we download the data?

Methods & Data: We retrieved Google Trends data for the same regions, time periods, and terms from four different countries on three continents (Austria, Germany, the U.S., and Australia). We then compared the search index values retrieved from each respective country to those downloaded in the other countries, keeping all parameters of the query constant.

Results: Our results show that values from Google Trends differ across download locations depending on the download day and the query’s total search volume. Researchers can minimize these inconsistencies by averaging samples from several days for high search volume queries. Nevertheless, our results point to an additional limitation regarding the reliability and replicability of Google Trends data for its usage in social science research.

Added Value: Our findings help researchers working with Google Trends data in making their research better replicable by averaging samples from several days for high search volume queries. Our results also serve as a tail of caution for research relying on APIs that provide samples of their digital trace data as the download location might impact the findings.

Relevance & Research Question: Online labour markets (OLMs) reflect the globalisation of the past three decades, combined with accelerating digitisation, and are poised to reshape the future of work. For highly educated workers in developing and emerging economies, OLMs offer significant income opportunities. However, existing literature highlights issues such as insufficient regulation, lack of transparency, and inadequate policy focus. Recently, emerging frameworks like the German Act on Due Diligence in Supply Chains (Lieferkettengesetz, LkSG) have introduced legal mechanisms to address human rights violations in global value chains. These frameworks could also help regulate OLMs by requiring clients to exercise due diligence. This obligation, however, depends on the ability to identify clients and assign them corresponding responsibilities.

This study addresses two key research questions:

1. How can labour rights principles, such as the duty of care outlined in emerging supply chain regulations like the LkSG, be applied to OLMs?
2. How can digital trace data from OLMs be used to identify clients, assess their outsourcing behaviour, and evaluate compliance with these regulations?

Methods & Data: We build on digital trace data from Braesemann et al. (2022) and the Online Labour Index (Stephany et al., 2021) who compile data on freelance projects from platforms like UpWork and Fiverr. Our analysis focuses on project histories to examine client outsourcing behaviour. Metrics such as wages, working hours, and gender imbalances are also assessed.

Results: Our study demonstrates the feasibility of identifying clients through project-ID matching algorithms, using a sample of 250 project IDs. Results show that small companies dominate outsourcing activities on OLMs. Wage distributions across case studies in Serbia, Egypt, and Bangladesh reveal that average freelance wages often exceed local minimum wages. However, significant variations exist across occupations and genders, underscoring the need for targeted policy interventions to ensure fair pay and gender equity.

Added Value: This study highlights the potential of supply chain regulations to address regulatory gaps in OLMs by enforcing minimum wage standards and addressing gender disparities. It also advances methods for identifying and analysing clients on OLMs, providing actionable insights for policymakers and researchers.

Relevance & Research Question

Current research on online behavior heavily relies on self-reported data, which, if flawed, can lead to inaccurate inference in subsequent analyses. Researchers examining online behavior require detailed measures beyond "time spent on a platform" to explore, for example, well-being, social media use, or online privacy, particularly to differentiate between active and passive social media use.

This study investigates the extent of misreporting in questions about fine-grained Instagram behavior by comparing them to objective measures collected via data donation. We also explore to what extent the accuracy of self-reports is dependent on the response format (rating scale vs. open text field) and the reference period ("last week" vs. "typical week").

Methods & Data

We collected survey data from over 500 Instagram users in a German probability-based online panel regarding 25 distinct behaviors, including posting, liking, and commenting. Participants first complete survey questions on these behaviors. As part of the survey, we conduct a 2x2 experiment that randomly varies the reference period and, for a subset of behaviors, the response format. Respondents are then asked and, if they agree, instructed to download their Instagram usage data for the last three months and donate them to our research. We analyze correlation coefficients between behavioral self-reports and donated data to assess the accuracy of self-reports in general and for specific behaviors.

Results

Our study’s data collection phase ended on 12 November, and we cannot present any results yet. We have successfully collected self-reported and donated behavior data from 122 respondents. We will update this abstract with the respective findings before 1 March 2025.

Added Value

This study contributes in three ways. First, we inform the field of questionnaire design by offering insights into how to accurately inquire about specific online behaviors, which is particularly interesting for researchers who may not utilize data donation methods. Second, we examine the accuracy of self-reported data on individual Instagram behavior, helping researchers assess the validity of surveying self-reported online behaviors. Third, we illustrate the potential of data donation to gather detailed, fine-grained data on individual behaviors, which participants might be unable to report accurately.