Labor market statistics are critical for assessing and understanding how the economy works. They are also paramount for identifying sources of income, especially in developing countries and particularly for poor people who often have labor as their only asset. In practice, widespread variation exists in how labor statistics are measured at the micro level. Labor modules in multi-topic household surveys, rapid welfare assessments and labor force surveys vary in the recall period used, question sequencing, the use of probing questions, and the respondent interviewed. Little is known whether these differences have an effect on the labor market statistics that they produce. This paper analyses these effects by implementing a randomized intervention in Tanzania varying two key dimensions of the labor module: the type of respondent – self-reporting versus proxy respondent- and the level of detail of the questions – a detailed longer version versus a rapid assessment. Estimating how this affects different labor statistics like those for labor force participation, labor supply, earnings and the distribution of activity, we find significant differences across questionnaire designs and respondent types. Labor force participation rates, for example, vary by as much as 10 percentage points, depending on what module is used. Response by proxy, for instance, has a substantial and statistically significant effect, both on labor force participation and weekly hours, as well as on reported female earnings. Using a shortened questionnaire instead of a detailed one especially affects the type of job or employment status, leading to lower reports of wage employment for men, and lower wage and self employment for women.