It's the nature of averages...
By using the Median (half above, half below) some of the problems with "averages" can be minimized. But problems exist none-the-less.
There's a lot of data at the Bureau of Labor Statistics site that can give a better picture of what's going on than the overall number. For example, the [link|http://data.bls.gov/cgi-bin/dsrv?nc|National Compensation Survey] pages let you construct tables of values for the past few years. It shows that top level ("Level 14") mathematicians and computer scientists (occupation 07000) had a drop in average (I assume these are means and not medians, but I can't find a clear statement about that) wages since 2002:
September 1997: $47.58
December 1998: $49.86
September 1999: $51.82
July 2000: $56.34
January 2001: $61.36
July 2002: $67.20
July 2003: $58.13
July 2004: $56.92
yet the average for all 07000 jobs in all levels rose continuously from $26.96 to $35.07 over that same time period.
There can be huge differences in wages over time for some occupations in some areas. E.g. #33438
State : Georgia
Area : Atlanta, GA
Occupation : Food counter, fountain and related occupations
Level : Overall occupation average (no work level)
October 1997: $7.00
No data for 1998
January 1999: $5.83
August 1999: $5.87
January 2000: $6.32
January 2001: $7.43
No data for 2002
January 2003: $8.29
December 2003: $7.10
No data for 2004
January 2005: $6.37
Bottom line: Don't put too much stock in a single national number and attempt to correlate it with what's happening in your area. Also, increasing wages for higher level employees can mask stagnant or falling wages for entry level employees.
There's probably enough data at the BLS to determine whether wages in a particular occupation at a particular skill level have risen or fallen since the late 1990s, but you'll probably have to build the tables yourself.
HTH.
Cheers,
Scott.
