In my latest Employment Update for January, I pointed out that the Bureau of Labor Statistics revised the December month-over-month change for Nonfarm Payroll Employment upward from 252K to 329K and November from 353K to 423K, a total of 147K to the upside. However, the BLS's initial estimate for this metric should be taken with the proverbial grain of salt. See this link to historic revisions back to 1979 on the BLS website.
In the chart below, I plot the change since the turn of the century from the first to third estimate for each month through November 2014, the most recent month for which we have three estimates.
During this timeframe there were 109 upward revisions and 67 downward revisions. The absolute mean (average) revision was 45K, which breaks down as 47K for the upward adjustments and 43K for the downward adjustments. The latest third estimate is an upward revision of 102K.
Interestingly enough, the direction of revisions was upward during the brief recession of 2001 but downward during the nasty recession from December 2007 to June 2009.
To reiterate what I've frequently stated: Don't take the initial monthly employment jobs data too seriously.
Note: Jeff Miller of NewArc Investments pointed out in an APViewpoint commentary that "the prime source of revisions over the three reports is the effect of late-reporting firms." Jeff's comment got me to wondering about the potential for monthly seasonality to reporting. Here's a column chart showing the averages for each month since the turn of the century.