Abstract The literature on information provision in stated preference surveys studying food safety has virtually ignored potential influences on variances of preference parameters (or heteroskedasticity) and has solely focused on systematic heterogeneity in their means estimates. Recently, unraveling both types of preference heterogeneity underlying consumer choices has become possible with the extension of the mixed logit (ML) model into a heteroskedastic variant (HML). In this paper, we use a HML model to analyze data from a choice experiment studying the effect of providing information related to third-party food safety certification on Lebanese food consumers’ purchasing decisions. To our knowledge, this is the first application of the model to study information provision. Results reveal that besides the expected increases in preference parameters’ mean pertaining to the food safety certificates (ISO 22000 and ServSafe®) among informed respondents, variances also significantly increased. In contrast, ...