Our survey requires enumerators to interview both the primary male and female decision-makers within each household. Commonly the Enumerator 1 will ask either the male or female respondent to answer questions in the household questionnaire, which includes the list of household members (household roster). After the household questionnaire is finished, the Enumerator 1 will continue on with an individual questionnaire and Enumerator 2 will begin the individual questionnaire with the male or female respondent that did not answer the household questionnaire.
So both the male and female respondents are being interviewed simultaneously, however Enumerators 1 and 2 need access to the household roster that only Enumerator 1 collected.
Has anyone else come across this issue or have an idea how to address this in the field?
You approach will work if you have reliable internet connectivity in the field. Then, one enumerator collects hh roster and synchronizes with HQ. You can write a script to automatically create a new questionnaire and send it to the second enumerator.
We find that allowing tablets exchange information in the field without synchronization to the server lead to serious problems and data losses. Until we find a robust way to handle these situations in the field, we are not allowing tablet to tablet information exchange.
while such parallelization may sound like a good idea - there are also weak points:
the interviewer 2 is always sitting idle for the first phase of the interview; (and possibly for the second phase if the second respondent is not immediately available)
there is a problem of data inconsistency if the first interviewer modifies the roster after the second interviewer has already got the snapshot copy of it (or the other way around). There is always a possibility that someone was omitted during the initial enumeration and needs to be subsequently added, or that someone was recognized not being a member during the detailed interview, and needs to be removed from the roster. Working with two independent copies will create chaos in your data.
interviewer 2 may also notice and attempt to correct errors in data entry left unnoticed by interviewer 1, further diverging the two copies.
A Labour Force Survey is a classical example where interviews are conducted directly with each respondent, and they are commonly conducted by a single interviewer per household. They collect all the similar information as you describe, but in a linear fashion. I suggest you stick to one interviewer per household, and send the second interviewer to another household instead. I also realize that this is not always possible, particularly if you want to match the gender of the interviewer and respondent, for example.
We did a similar experiment w/ an ag survey. We had both enumerators fill in the holding characteristics, and household roster together. Then, they separated, and conducted simultaneous interviews at different locations on the holding w/ different hh members.
In the end, we took a look to see if the responses recorded by the enumerators were the same for the holding level questions, and household roster. With very few exceptions, they were.