# How is WNTFINL computed?

Although WNTFINL correspond to pwsswgt (in the NBER basic monthy file),. some cases for PWSSWGT have values of -1 and 0, whereas the WNTFINL does not.

Please share steps for such computation?

The NBER files contain non-interviewed records while the IPUMS CPS files don’t include these records. If you drop all of the non-interviewed records from the NBER file there shouldn’t be any PWSSWGT values of -1. The zero values, however, are valid and are included, in some instances, in WTFINL. A value of zero is given to populations not sampled in other monthly samples such as persons in the Hispanic oversample and members of the armed forces in ASEC samples.

Thanks! That does help.

However, when I apply WTFINL as weight to calculate the number of discourged workers, the number is VERY different to the official BLS numbers as well the number I get when I use the raw NBER basic data.

Using IPUMS JUNE BASIC DATA.

compute marginal = \$sysmis.

DO if LABFORCE eq 1.

If WNLOOK ge 1 AND WNLOOK le 5 marginal = 1.

If WNLOOK ge 6 and WNLOOK le 99 marginal = 2.

END if.

Variable labels

marginal ‘marginal workers’.

value labels marginal

1 ‘Discouraged’

2 ‘Other persons marginally attached to the labor force’.

weight by WTFINL.

display DICTIONARY var = marginal.

freq var marginal.

I get something close to 1.7 million – it should be closer to 500,000.

Is there something wrong with the weight?

I’m not entirely sure how the BLS calculates their official statistic. It seems that they are restricting their universe for this calculation over and above the restrictions already baked into WNLOOK. “Discouraged workers” (according to the BLS) are those who are “marginally attached to the labor force” who have not looked for work within the past 4 weeks, while the universe for WNLOOK includes all people who are unemployed who have not looked for work within the past 4 weeks.

Also, when I look at the NBER raw data I’m calculating the same numbers that I find in the IPUMS CPS data. So, this doesn’t seem to be an issue with WTFINL. If this issue persists, perhaps you can send me more information on the calculations you are performing with the NBER data.