There are different approaches. 1. When it comes to data that can be deduced from the answers of other people. For example, if the group of respondents is from the same neighborhood. Even if that question has not been answered, it can be deduced without problems. 2. When a response can be attributed from a calculation. For example, assigning the population mean or the group response mean. 3. When the answer can be deduced by checking other answers. For example, the employment situation based on contracts, the place of work or others. 4. When you should interpret what you meant by your non-response. This last case is very doubtful. In any case, it does not seem advisable to replace the non-response, but only to interpret in the report why the respondent did not answer.
I am using Questionnaire 5 Likert Scale. Some of responses they miss to complete one item in their demographic information. I think my data related for first approach which is the data can deduce from other and from same neighborhood.