Changes in employment, earnings, and work hours are used as standard indicators of economic health in the United States. In spite of their importance, within the decentralized U.S. statistical system these indicators are based, to some extent, on different underlying concepts, and measured in ways that place different cognitive demands on respondents. In this paper, we look specifically at how "employment," "work hours," and "earnings" have been operationalized in self-administered establishment surveys in two major U.S. statistical agencies. We present the results of an expert review comparing questionnaires and collection procedures for the key variables, and of in-depth interviews with respondents who were asked to provide data using alternative questionnaire approaches. We evaluate respondents' question-answering strategies with an eye toward the effects, if any, that might appear in the resulting data. We then compare employment, work hours, and earnings from published estimates in different surveys. We consider whether differences in reported employment, work hours, or earnings could be attributed to respondent behavior with respect to the structure and content of the questionnaires used to measure them. We also address the implications of our research findings for data users, for respondents, and for statistical agency program managers.