One objective of a survey can be to understand the proportion male vs. female users. To the extent a specific gender correlates to particular use cases of a technology, product design, product/feature prioritization and marketing can be differently managed.
A survey of existing users will inherently have error bars based on sample size, frequency of sampling, etc. As I understand it, when asking the question 'are you male or female?' as part of a demographic survey where all fields are required, an additional selection bias is introduced into the survey.
Specifically, if a survey is sent to 100 users, of which 50 respond, the proportion of responders that are male may not represent the proportion of 100 customers that were initially surveyed.
Assuming there is not a existing sample set of known males and known females whom to survey in order to estimate the responder bias, what are some approaches to prevent or to correct for this error?