Comparing survey data and administrative records on gross earnings: nonreporting, misreporting, interviewer presence and earnings inequality