The BLS records the range of revisions from the 1st to the 2nd edition, the 1st to the 3rd edition, and the 2nd to the 3rd edition of the NFP. hereUsing the average absolute correction value for 2019, the reasonable range of employment growth (not percentage growth) is shown below.
figure 1: Non-agricultural employment change (black) and 2 times the mean absolute correction (MAR) band (grey) are used from 1 to 3 in October, and from 2 to 3 in September. MAR is 2019. Source: BLS Employment Status Released in October, Bureau of Labor Statistics, And the author’s calculations.
Therefore, even if there is this uncertainty about the correction (mansky calls it temporary statistical uncertainty), we know that employment is increasing or decreasing (the average correction is 5 and 8). Now, 2019 is before the pandemic. What if we use the 2020 revision? This produced Figure 2.
figure 2: Non-agricultural employment change (black) and 2 times the mean absolute correction (MAR) band (grey) are used from 1 to 3 in October, and from 2 to 3 in September. MAR is for 2020. Source: BLS Employment Status Released in October, Bureau of Labor Statistics, And the author’s calculations.
Therefore, if we use some kind of spread measurement about data revision (BLS uses MAR, not standard deviation, because the change looks abnormal), we can guess that the decline in growth in September will likely continue until the third revision. And the growth in October will continue to the third revision (this address reader rsmQuery, as long as it can be coherent).
