US Nonfarm Payrolls signalling recession?

The 258k downward revision to the prior two months of payrolls was 4-5x larger than the median absolute revision since 1979. The median average revision is 56k. If we regress revisions to payrolls on expansion or recession, we find that the 258k downward revision implies a 9pp increase in the probability of recession. However, if we add the current month's payroll data as an additional explanatory variable, the statistical significance of revisions evaporates. The message is the signal from the July payroll (+73k) is likely more important than the abnormally large downward revisions.