While still on paternity leave and in between diaper changes today, my “data disease” got the better of me. I have been digging into the latest QCEW for all states released by BLS this morning and the findings all point to future upward employment revisions. As regular readers know, one sleeps on the QCEW data at their own peril when examining employment trends – see my last state by state look. Working with Mark over email, the following takes a look at these expected revisions across all states.
- The last national benchmark was for March 2012 when the nation – sum of states gap was closed.
- The last state level benchmark was for September 2012, and again the nation – sum of states gap was more or less nonexistent.
- All data from October through today is off benchmark and the lastest QCEW for Q4 indicates there are more private sector jobs than the current estimates show.
Using this quick method of comparing the different data series, the private sector sum of all the individual states is expected to be revised upward by approximately 500,000 (0.5%) for December. However, the topline U.S. employment estimate is currently larger than the sum of states would suggest. As such, we can expect a somewhat smaller upward revisions of around one month’s worth of job gains (200,000) in the national figures.
In terms of how each state is expected to be revised, the following graph shows revisions for December 2012 based on the new QCEW data. As has been the case in recent months, the disparity between employment surveys is particularly pronounced in Oregon. Oregon’s upward revision will likely be around 50% larger than that of the average state.
It remains to be seen what will happen with revisions to public sector job counts, which are notoriously difficult to predict. With the impact of the federal sequester yet to be seen in the jobs data, upcoming government job counts may change more than usual.
Even with large changes to government job counts, it is hard to see how the next national employment benchmark revisions will not be up.
UPDATE: Due to an inquiry, the following tries to estimate the accuracy of this method based on recent years. I am taking the snapshot used in the previous edition of this work – QCEW through June 2012, with the most recent benchmark at that time being June 2011. The blue bars in the graph below show the predicted monthly revisions to the sum of states data available at the time of publication, while the red line shows the actual revisions based on the data available today. As one can see they are quite close. Over the entire period (July 2011 – June 2012) there is about a 90% accuracy.
Another way to look at it is examining the cumulative changes over the year.
Overall the actual revisions were smaller than the predicted revisions by about 10%. At least here in Oregon, I know the reason for this was due to changes in non-covered employment in the state. The QCEW data is, by definition, employment covered by the UI system so therefore cannot be used to gauge trends in non-covered employment. I suspect this is at least part of the explanation for the national divergence as well.