Error in Stata code for ATUS and CPS earnings vars?

Hi, I suspect the Stata extraction code for earnings is incorrect. I get the following means for weekly and hourly earnings:


. su earnweek*

    Variable |        Obs        Mean    Std. Dev.       Min        Max
-------------+---------------------------------------------------------
    earnweek |    201,151     45394.9    49308.98          0   99999.99
earnweek_c~8 |    201,151    46023.38    49374.31          0   99999.99

. su hou*

    Variable |        Obs        Mean    Std. Dev.       Min        Max
-------------+---------------------------------------------------------
    hourwage |    201,151    696.5316    454.4572          0     999.99
hourwage_c~8 |    201,151    705.7871    450.5602          0     999.99


I think the divisors in the generated code below should be much larger, e.g. 10000 for weekly earnings rather than 100. Or if they're correct, pl tell me how to get these jobs :). Thanks much, S

===========
clear
quietly infix                    ///
  long    year            1-5    ///
  double  caseid          6-19   ///
  byte    pernum          20-21  ///
  int     lineno          22-24  ///
  double  wt06            25-41  ///
  byte    fullpart        42-43  ///
  int     uhrsworkt       44-47  ///
  double  earnweek        48-54  ///
  byte    paidhour        55-56  ///
  byte    earnrpt         57-58  ///
  double  hourwage        59-63  ///
  byte    cpsupdate       64-65  ///
  byte    fullpart_cps8   66-67  ///
  int     uhrsworkt_cps8  68-71  ///
  int     hrsworkt_cps8   72-75  ///
  double  earnweek_cps8   76-82  ///
  double  hourwage_cps8   83-87  ///
  using `"atus_00009.dat"'

replace earnweek       = earnweek       / 100
replace hourwage       = hourwage       / 100
replace earnweek_cps8  = earnweek_cps8  / 100
replace hourwage_cps8  = hourwage_cps8  / 100

format caseid         %14.0g
format wt06           %17.0g
format earnweek       %7.2f
format hourwage       %5.2f
format earnweek_cps8  %7.2f
format hourwage_cps8  %5.2f

It looks like these summary statistics are being calculated including the special codes for Not in Universe (e.g., NIU ==99999.99). This is pulling the means upwards quite a bit. NIU cases indicate that the respondent is not eligible for the earnings question. Therefore, they should be excluded from your analysis. Once you exclude these cases your means should be much more reasonable.

Hmmm, is there a var to identify the NIU cases, or some other documentation of those values? thanks much, Sanjiv

Yes, the IPUMS online data extract system includes any and all documentation available. The codes tab for each variable (e.g., EARNWEEK and HOURWAGE) identifies special codes, such as NIU. In EARNWEEK NIU==99999.99 and in HOURWAGE NIU==99.99.