10/26/2017 Toby Dayton

The New Abnormal Is Wreaking Havoc on Job Market Forecasts; LinkUp Predicting Net Gain of 120,000 Jobs In October

Hope Ahead

It’s tough to jump into our October non-farm payroll (NFP) forecast without first revisiting our eye-popping forecast for September. Based on significant job listing gains in our job search engine in August, we forecasted a net gain of 260,000 jobs in September. As we noted in that post, however, our model does NOT account for weather (especially hurricanes) or seasonality, and as a result, we knew full well that our forecast was going to be higher than what the Bureau of Labor Statistics (BLS) released in their official numbers.

What we did not expect, to be sure, was that September’s jobs report would be as horrific as initially reported by the BLS. We’ll see what happens in their 30 and 60-day revisions, but the severity of the impact from the hurricanes on the job market, at least to this point, seems to have been far greater than most people expected, us included. But as we noted in a recent blog post, estimating the impact of a hurricane on the job market is no easy feat and when we publish our October jobs data next week, we will also be providing some additional data and analysis related to what’s happened specifically in Texas and Florida over the past two months. Hopefully that helps clear up a bit of the weather-related uncertainty surrounding the current job market environment.

In the meantime, we are forced to seek solace in our still-decent forecasting track record for 2017, during which time we have accurately forecasted above or below-consensus in 5 of 9 months. Again, we’ll see what happens with both future revisions and Q4 data, but we remain encouraged by our .556 batting average despite last month’s abomination.



And since 2012, we have maintained a consistently solid track-record with our NFP forecasts. Our forecasted average monthly job gains by year are depicted below relative to BLS data. In each of the past 5 years, we accurately forecasted monthly job gains above or below consensus estimates in 7 of 12 months for an historical batting average of .583. (And yes, we freely acknowledge that one wouldn’t be wrong in sensing a decidedly defensive bent to this month’s post – an NFP forecast miss of 293,000 jobs is bound to cause a bit of insecurity).



Unfortunately, despite the reassurance provided by our historical track-record, the abnormal environment we’re in (and for purposes of this post, I am only referring to weather-related abnormalities) will continue to wreak havoc on both the job market in general and our forecast model in particular which is dependent on the prior month’s official, initially reported job gains to determine the following month’s net job gains. And because last month’s job number was such an aberration, it makes it exceedingly difficult to get a good reading for this month’s forecast.

So we’ll start with what we do know with absolute or near-absolute certainty…

  1. Labor demand in the U.S. remains very strong
  2. Job listings increased in August
  3. Job listings declined in September (undoubtedly due to some unknowable extent to hurricanes)

Below is LinkUp’s job market data by state for September.



But of critical importance to the data above is the fact that we only index jobs directly from company websites and the number of companies in our index is growing every month. To elaborate further, we index, every day, roughly 4.2 million jobs (at present) from 30,000 employer websites around the world. As a result, our job market data is always current, there are no expired listings, and we eliminate what we call job board pollution. But because we are always adding new companies to the index, there is an upward bias to the ‘raw’ data in the table above.

To counteract that bias in our NFP forecasting model, we use a paired-month methodology that looks at only employers that were in both months being compared – in this case August and September. Keep in mind that because our data is predictive of future job growth (given the safe assumption that the best indicator of a job being added to the economy is an employer posting an opening on their corporate career portal), October job gains are a function of job listing activity in September relative to August.

In comparing September to August using that paired-month methodology, new and total job listings declined 7% and 2% respectively in September.



So based on LinkUp’s September data, our model would normally indicate that job gains were negative to an even greater extent in October than September in October with a net loss of roughly 100,000 jobs. But given the massive uncertainty around September’s BLS data and the hurricane impact in both September and October, we have serious reservations about using the -33,000 September number in our model. As a result, we are reverting to what we think September would have been were it not for the Hurricanes – our original September forecast of a net gain of 260,000 jobs. That may still be a bit high, so using a September gain of approximately 225,000 jobs results in an October NFP forecast of a net gain of just 120,000 jobs.



And with that, we’ll come to a full stop so as to not dig a deeper hole than we’re already in given last month’s forecast. Hopefully things will return to normal next Friday. It’s about as likely as normalcy returning to Washington anytime soon.


Leave a Reply

Your email address will not be published. Required fields are marked *