Keywords: online survey, preference falsification, authoritarian contexts, vpn sampling, sampling method, Iran

Online non-probability surveys face significant challenges regarding randomness and representativeness. However, prior research, including our own studies in GAMAAN, has demonstrated that measurement errors in conventional probability surveys—such as those conducted via phone, email, or in-person interviews—can reach up to 50 percentage points for sensitive questions in authoritarian contexts due to self-censorship and preference falsification. These errors cannot be effectively eliminated. By contrast, sampling errors in online surveys can be substantially reduced using balancing and weighting methods, provided the sample size is sufficiently large and diverse.
To address these challenges and improve the representativeness of online surveys in authoritarian contexts, we explored a method to approximate randomized sampling. In Iran's unique authoritarian environment—where most social media and widely used platforms are filtered—VPNs (Virtual Private Networks) are extensively used for internet access, with 85% of internet users in Iran relying on them, according to several studies. Our findings indicate that VPN usage spans nearly all socio-economic groups who use the internet. This widespread adoption of VPNs creates a unique opportunity to implement randomized and anonymous sampling.
In several national surveys conducted in 2023 and 2024, facilitated through a popular free VPN platform, we employed VPN-based sampling to measure public opinion on both sensitive and non-sensitive topics. The results demonstrated that this approach effectively combines the strengths of randomized sampling with the benefits of anonymous participation. VPN sampling, therefore, offers a reliable and innovative method for obtaining more representative data and genuine responses in authoritarian contexts, where a significant and diverse segment of the population uses these anti-filtering tools.