# comparability of ASEC OCC10LY over time

I’m looking at occupations across time using annual ASEC data and am seeing strange discrepancies in the data. I am using ASECWT to weight my calculations.

Here are my worker counts for OCC10LY code 4760 (Retail salespersons):

1971 - 3,117,567

1972 - 3,182,313

1973 - 3,099,777

1974 - 3,333,314

1975 - 3,249,241

1976 - 3,319,255

1977 - 3,246,498

1978 - 3,369,858

1979 - 3,443,170

1980 - 3,376,371

1981 - 3,400,294

1982 - 3,419,197

1983 - 1,350,126

1984 - 1,446,670

1985 - 1,573,165

1986 - 1,525,183

1987 - 1,574,499

1988 - 1,662,057

1989 - 1,572,091

1990 - 1,734,984

1991 - 1,723,862

1992 - 1,657,755

1993 - 1,614,129

1994 - 1,588,451

1995 - 1,577,295

1996 - 1,688,568

1997 - 1,745,024

1998 - 1,669,642

1999 - 1,801,920

2000 - 1,819,837

2001 - 1,740,213

2002 - 1,734,266

2003 - 56,741

2004 - 74,956

2005 - 57,521

2006 - 80,057

2007 - 78,578

2008 - 3,895,572

2009 - 3,765,988

2010 - 3,686,700

2011 - 3,417,320

2012 - 3,595,344

2013 - 3,387,781

There’s a huge decline 1982 to 1983, another huge decline from 2002 to 2003, then a big jump from 2007 to 2008. What could account for this difference? I assumed “Retail salespersons” was a fairly stable and easily classified occupation that would result in stable numbers.

These sorts of fluctuations are common even with harmonized versions of occupation codes. The decline between 1982 and 1983 can be attributed to a coding scheme change between those years (see the codes tab of OCCLY for more details). The declines and jumps in the 2000’s are roughly offset by corresponding jumps and declines of the occupation code OCC10LY == 4850 “Sales Representatives, Wholesale and Manufacturing”. You can always collapse all of the Sales and Related occupations (OCC10LY == 4700 - 4950) into a broader category that will be less sensitive to these sorts of coding changes over time.