Survey instructions bias perceptions of environmental health risks

A basic function of government is to manage environmental risks. This requires measurement of how the public perceive different risks, but it is hard to measure risk perceptions in a neutral, unbiased way. In a pre-registered experiment with a nationally representative sample (N = 800), we tested the effects of survey instructions on perceived risk from common environmental health hazards. Participants were randomised to read instructions that made salient a relatively unfamiliar hazard (electromagnetic fields; EMFs), a familiar hazard (carbon monoxide) or no hazard (i.e., that the study was about environmental health risks in general) before rating perceived risk of a series of hazards. Results showed an asymmetric salience effect. Instructions that highlighted EMFs elicited higher levels of perceived risk from EMFs, but perceptions of other hazards were unaffected. Instructions that made carbon monoxide salient did not affect perceptions of carbon monoxide, but diminished perceived risk from other hazards. Effects were observed on widely used rating scales and on a novel policy budget allocation task. In exploratory analyses, we further sought to test the relationship between perceived risk as elicited by rating scales and revealed risk perceptions. Results showed that how often respondents report thinking about a hazard day-to-day best predicted choices in a hypothetical budget task and self-reported mitigative behaviour, even when other dimensions of perceived risk, such as perceived severity of the consequences of exposure, were controlled for. The results have implications for designing surveys to measure perceived risk of environmental health hazards.