3
$\begingroup$

How does the CPS CEPR store employment information for the non-ORG months?

I am trying to reproduce Mueller (2012). He says

The CPS is the main labor force survey for the U.S., representative of the population aged 15 and older. It has a rotating panel structure, where households are surveyed in four consecutive months, rotated out of the panel for eight months, and then surveyed again for another four consecutive months.

But it also says

Note that the CPS records the labor-force status for each person in the sample each month. Weekly hours and earnings, however, are collected only in the fourth and eighth interview of the survey,

This means that the CEPR has

  • ORG files that contain the wages
  • monthly files that contain labor force status

On the CEPR Monthly site, it says that

The CPS Basic Monthly data variables are incorporated into CEPR CPS ORG data.

Which means that I only have to download the CEPR ORG data. Sounds like good news.

I downloaded 2013 and 2014 from here. Minsamp corresponds to the monthly interview, the variable has unique values 4 and 8 (which are the outgoing interview months). So far, so good. There is an empl variable which appears to encode employment status. However, I cannot find the employment status for other months than the interview months. I can't find neither additional rows, which would carry that information in long form, nor additional columns, which could carry it in wide form.

Here's how some data extract (restricting some columns and rows) looks:

                              age      wage4  year  month  minsamp  empl
hhid            hhid2 lineno                                            
000000113071409 03011 1        67        NaN  2014     12        4     0
000005890210971 02011 1        25        NaN  2014      8        4   NaN
                      2        26        NaN  2014      8        4     0
000005893210371 03011 1        24  15.600000  2014     12        4     1
                      2        25  30.333334  2014     12        4     1
                      3        24  51.000000  2014     12        4     1
000008171510365 02011 1        61  19.225000  2014      8        4     1
                02111 1        37  16.250000  2014      7        4     1
                      2        39        NaN  2014      7        4     0
000010415001537 03011 1        45  18.000000  2014     11        4     1
                      2        42  26.442249  2014     11        4     1
000010666500851 02011 1        49  21.538401  2014      8        4     1
                      2        44  20.000000  2014      8        4     1
                      3        19  14.500000  2014      8        4     1
                      4        19   7.750000  2014      8        4     1
                02111 1        61        NaN  2014      8        4     1
                      2        54  15.000000  2014      8        4     1
                      3        53        NaN  2014      8        4     1
000011141520290 90001 1        40   7.500000  2013      4        8     1
                      2        46        NaN  2013      4        8   NaN
                91001 1        62        NaN  2013     12        8     0
                92001 1        32  30.000000  2013      8        4     1
                      1        33  20.150000  2014      8        8     1
                      2        30        NaN  2013      8        4     0
                      2        32        NaN  2014      8        8     0
                93001 1        52  30.250000  2014      4        4     1
                      2        55  18.025000  2014      4        4     1

As you can see, for every unique household member (every hhid-hhid2-lineno combination), there is only one row. And if its somehow stored in the wide format, I'd expect the name to contain 'emp' somewhere:

... [x for x in df.columns if 'emp' in x]
['empl', 'unempt', 'selfemp', 'pdemp1', 'pdemp2', 'nmemp1', 'nmemp2']

where the first four correspond to the individuals employment status (but not in different months), and the last for correspond to the number of employees linked to the individual (and similar).

Clearly, the information must be there somewhere. So, How does the CPS CEPR store employment information for the non-ORG months?

$\endgroup$
1
$\begingroup$

Dusting off the cobwebs... The microdata "wage file" as I call it and the ORG file as you call it only has information for the month for which the respondent was asked about wages. The CPS is a timeseries only in the form of aggregated data. I do not know the study you are replicating so can't be more helpful. Try adding what your objective is, e.g. I want to measure the wage differential of age groups in the XYZ industry/occupation or whatever. My experience was in estimated the hourly wage distribution of workers in California.

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.