This adjustment process is broad based, and one must be careful when comparing series over time. For example, David Altig at the Atlanta Fed remarks that January's revisions to population estimates created a strong downward bias on the employment figures, rendering household survey (employment numbers) less reliable when comparing across time. The payroll figures (the establishment data) are better used to compare employment across time.
But the establishment data (nonfarm payroll) is also subject to heavy end-of-year revisions.
The chart lists the previously published and end-of-year revisions to the monthly change in total nonfarm payroll employment (source: Table B in the first half of the release, or page 6 of the .pdf). For the year as a whole, the revision process slashed another 385,000 jobs over 2008.
What I find rather striking is what happened in the first half of the year. As previously published, the job loss was an accumulated 528,000 through July, but the revisions added another 396,ooo in job loss. As it turns out, almost 1 million jobs were lost just through July, which went unreported until February 2009. Job loss in the first half of the year didn't appear to be too massive when the data was released, but was actually much worse when the benchmark was corrected.
And this occurs at the state level, too. Later this month, the BLS will release its state employment reports for January. But an early preview indicates that the report will show further contraction in the labor market, at least in California. According to the California Employment Development Department:California’s unemployment rate was 10.1 percent in January, and nonfarm payroll jobs declined by 79,300 during the month, according to data released today by the California Employment Development Department (EDD) from two separate surveys.
In December, the state’s unemployment rate was a revised 8.7 percent, and in January 2008, the unemployment rate was 6.1 percent. The unemployment rate is derived from a federal survey of 5,500 California households.
The numbers are awful, but state-level employment numbers are subject to revisions as well. For example, the December unemployment rate was revised downward 0.6% to 8.7%. Furthermore, revisions mitigated some of the surge over the year: in December, the annual change was revised downward 0.8% to 2.6%. Nevertheless, the January 2009 numbers are horrible.
Employment numbers are subject to heavy revisions on a monthly basis and over the year. But all indicators (i.e., the rise in initial claims over the survey period for Feb. compared to that from Jan.) signal to a further deterioration of the labor market, at least in February.