Interpreting the August Employment Report

As I write this on Thursday night, very few people know what the August employment report will show. Only the President, Vice-President, and a few top economists get an advance look on Thursday afternoon. The BLS is in lockdown mode for eight days. Media sources get a look on Friday morning, but are cut off from communications until the embargo ends at 9:30 Eastern Time.

Second best is having the background and context for intelligent interpretation of the data.

What is job growth?

Many (most?) sources call the headline number of the report “job growth” or the number of jobs created. Not so. It is actually the net increase in jobs. A look at the actual data will make the significance clear.

Monthly gains in new jobs are nearly 2.5 million. Job losses are a bit less. The difference between the two is the widely reported net increase. Millions of jobs end and millions of others begin each month. The number of people affected by this is far greater than the net jobs story suggests.

Also dwarfing the net change number is the story behind the dynamic. Currently there are over 6 million open jobs. More than 2 million workers voluntarily quit their jobs in the last reported month. The net job change is small in comparison with both of these numbers.

Misleading Focus on the Net Change

Let us now consider the much smaller net change – the most widely cited number. The table below shows the average change in the establishment survey over the last year, about 180K. It also shows the 90% confidence interval, about 120K. The confidence interval is the result of sampling error, inherent in the survey process. It is not something that is “fixed” by revisions in later months.

The household survey counts workers rather than jobs. This means that someone who works several jobs is only counted once. Since the sample is smaller, the confidence interval is much larger, about 500K.

The key point? The typical monthly change is not great when compared to the confidence interval.

Here is a look at the confidence intervals for the overall payroll numbers, as well as some interesting subgroups.

And the BLS provides a helpful table comparing key characteristics of the two surveys.

While the confidence interval is very large on a monthly basis, it is proportionally smaller when we have a longer period and more data. Once again, this is a statistical result, not the consequence of revisions.

Conclusions

  • Those who focus on the net change — which is nearly everyone – are not looking at the more informative picture. Policy changes impact the gross numbers. The net numbers are a secondary consequence. A mere 5% increase in new jobs is more than 100K per month. This is the difference between a sluggish economy and a thriving one.
  • Too much is made of net changes in the absence of confidence intervals. Few understand the sampling process and the implications. Market observers get a fixation on the “headline” number, despite the known error band. It is not that some of the reports might differ significantly from the truth. We know that they will show these differences, both individually and in the aggregate.
  • The complex reports provide ample opportunity for those who want to make a misleading interpretation.
  • Most of the most frequently cited complaints – seasonal adjustments, the birth/death adjustment, etc. – are small in the overall picture. They have also been demonstrated to improve results.
  • Finally, keep each single report in perspective and context. The ADP private employment report provides another source of evidence with a different methodology.

Sources

Employment Situation Summary Table A. Household data, seasonally adjusted

Employment Situation Summary Table A. Household data, seasonally adjusted . (2017). Bls.gov. Retrieved 31 August 2017, from https://www.bls.gov/news.release/emp

  • Comparing employment from the BLS household and payroll surveys

    Comparing employment from the BLS household and payroll surveys. (2017). Bls.gov. Retrieved 31 August 2017, from https://www.bls.gov/web/empsit/ces_cps_trends.htm#concepts

  • Employment Situation Summary Table A. Household data, seasonally adjusted

    Employment Situation Summary Table A. Household data, seasonally adjusted . (2017). Bls.gov. Retrieved 31 August 2017, from https://www.bls.gov/news.release/empsit.a.htm

  • Employment by industry, monthly changes, with confidence intervals

    Employment by industry, monthly changes, with confidence intervals. (2017). Bls.gov. Retrieved 31 August 2017, from https://www.bls.gov/charts/employment-situation/otm-employment-change-by-industry-confidence-intervals.htm