Every month, there is one count that the Bureau of Labor Statistics publishes that everyone looks forward to with much anticipation and hope. We hope that the number would be high and we also hope that all other numbers that matter also would be positive for the nation being related to that number. The metric we attach such a halo around is “new jobs added in non-farm payroll category”. As such, we analyze this key metric deeper.
To begin with, it is well known that the number is provisional when first published for a month. For example, Aug 2012 new jobs created stands at 96,000 for now. Since data flows in over the next months, this number undergoes a revision more than once before being settled as final.
Next, we took some sample data (all original and subsequent data for new jobs created under non-farm payroll category for the months July-Oct 2011) and compiled the correction figures, published over time. Refer the quartet graph above. It is interesting to note that in Aug 2011, what was an estimate of zero, got revised upwards twice to be an impressive 104,000 before dipping further. While that merits some explanation, the figure for September too shows nearly 100% variation from original estimate and that for Oct is 40% higher. In contrast July has major flip-flops.
Why do we think this is important?
The answer is not too difficult to fathom if we try to recall, when the last time the public memory gave importance to final estimates, was. We cannot rewind events. Life moves on and the real number has changed but we retain same emotions as of original figures. The new estimate for this month carries more weight in our minds than the final number of jobs created 3 months back. New incorrect sticks, the old correct fades.
The table below shows comparative data on new jobs created for six-month period ie Jul-Dec 2011 vis-à-vis Jan-Jun 2012. Six of the last eight revisions have been downwards. The average change is 2,000 jobs per revision as against an upward revision of 17,000 jobs in previous six-month period.
Whether it is increase or decrease after revisions, or it is a high or low original estimate, two or three revisions etc.- variations are abound but the point to ponder is, with 10-40% to even nearly a 100% change in either direction in most months, do the initial numbers deserve the importance and hype we give them?