I am replicating published economics papers that use CPS data. They use a measure of hours worked per week to code a variable of labor market participation. I assume the variable used is ‘hrswork’. The codebook only states that ‘hrswork’ takes two-digt values. Can ‘hrswork’ take value ‘00’? Having listed the values, there appear to be only positive numbers and ‘NIU’ entries. I expected ‘zero hours’ to be a legitimate response to the number of hours worked, though. Calculating labforce participation based on ‘hrswork’ is much lower than using ‘labforce’, which made me think that it is biased towards zero – in short, are individuals that responded ‘zero hours’ represented by ‘NIU’ values?
A code of ‘00’ for the variable HRSWORK only represents ‘Not In Universe’ respondents. This is because the Universe for HRSWORK only includes individuals who were “at work last week.” By definition this requires all respondents to have at least worked 1 hour last week. This also means that unemployed individuals, persons who were on vacation or on leave the previous week were considered ‘Not in Universe’ and coded as zero. Also, since the universe is limited to civilians, individuals in the armed forces were coded as Not In Universe. Compare this to LABFORCE, which counts unemployed individuals, respondents who were absent from work last week, and persons in the armed forces. This could explain why HRSWORK seems to under-count labor force participation compared to LABFORCE.