
Recently, a problem was discovered with the class of worker codes for the longest job for individuals who reported that they were self employed, either incorporated or unincorporated. The source of this problem was an instrument problem that first occurred in March 1996. We attempted to correct this situation with the March 1998 file, but did not completely rectify it. Following is a description of the problem by year and an overview of how we plan to correct the problem. Also given are procedures for receiving the corrected data.
March 1996 and March 1997:
Due to an error in the instrument, all individuals with a class of worker of self employed (either incorporated or unincorporated) in both their longest job and current job and who had different jobs were assigned a class of worker of self employed, incorporated (LJCW = 5). A simple corrective procedure exists for those who wish to use it: for all individuals with a longest job class of worker = 5 and a current job class of worker = 6, assign their longest job class of worker = 6. This is the recommended corrective procedure since over 99 percent of self employed workers in 1994 and 1995 had identical entries in both of these codes. The recodes ERN_SRCE, CLWK, and WECLW should also be corrected to reflect the change in the LJCW code. This is the procedure we will use to correct the internal files. It is our estimate that about 1000 to 1200 self employed individuals have incorrect longest job class of worker codes.
Another problem with the data was caused by this situation. This deals with some of the earnings recodes on the person, family and household records. These are the recodes that contain aggregate information on the amount of earnings an individual, family or household received from wages and salaries, self employment and/or their farm (variables HWSVAL, HSEVAL, HFRVAL, FWSVAL, FSEVAL, FFRVAL, WSAL-VAL, SEMP-VAL AND FRSE-VAL). Correction of these variables requires recalculating the appropriate variables wherever necessary depending upon how individuals had their class of worker variable revised. The Bureau will be releasing extract datasets with the corrected variables for these 2 years during the first quarter of 1999. We also will be correcting the appropriate variables on the FERRET datasets within the same time frame. A notice will be automatically sent to all FERRET users when the datasets are updated.
March 1998:
Since we were aware of this problem prior to releasing these data, we took action to correct the class of worker problem on the 1998 file. The corrective procedure fixed the problem listed above; unfortunately, it also incorrectly revised class of worker entries for other records. In all about 700 individuals have incorrect longest job class of worker codes. These incorrect codes can be of any value, not just the self employed. Unfortunately, there is no simple correction algorithm for these data as was the case for the previous years. The Bureau is in the process of correcting these class of worker codes and the appropriate earnings recodes as listed above. We plan to have these data corrected within the next few weeks.
The new 1998 file will also contain some additional information that will enhance and correct one other problem with the 1998 file. First, the file contains the new reasons for moving information that was added to the instrument this year. Secondly, the file contains replacement information for the variable HEA on an individual's current health condition.
Please refer any questions about the content of this user note to Bonnie Tarsia on (301) 457-3806.
Annual Demographic Survey (March CPS Supplement) 1998 Public Use File Page
CPS Main Page