• RSS Feed

The LinkUp Blog The Industry's Best-Kept Secret

August 1, 2014 / Toby Dayton

LinkUp Forecasting 225,000 Jobs In July and Strong Numbers For Rest Of 3rd Quarter

With the jobs report coming out in 45 minutes, there isn’t much time for in-depth analysis of our jobs data this month. As is the case on a few occasions every year, the 1st of the month falls on a Friday, the same day¬†that the Bureau of Labor Statistics (BLS) releases their Employment Situation Report. Our monthly jobs data is compiled at about 5AM on the 1st, which gives us about 2 hours to run the numbers, work them into our forecasting model, put together some quick analysis, and write what can only be classified as a very rushed blog post. The only solace to be had this morning is that no amount of additional hours of leeway would have clarified the foggy view of our data these days.

In July, on a state by state basis, new job openings in LinkUp’s job search engine by state fell 4% from June, while total job openings fell 3%. But we use a paired month methodology in our model to account for the fact that we are always adding new companies into our search engine along with all their jobs (about 500 new companies and 100,000 jobs each month). As a result, we get 2 data points for every month – the first when we compare a give month to the prior month, and the 2nd when we compare that same month to the following month. And then we use both data points for each month in our model. In any event, there are occasionally periods where the simple comparison of job openings from one month to the next conflicts with the more complicated way in which our model is built. The past few months have been just such a case.

The second muddying factor is whether or not our data is a leading indicator by 30 days or 60 days. We start with the assumption that a job opening posted on a company website is the best indicator of a future job being added to the U.S. economy. (And 100% of LinkUp’s job openings in our search engine are indexed from company websites so there are no old jobs, no duplicates, and none of the job pollution that’s found on so may job sites these days – things like work-at-home scams, fraud, identity theft, etc.). But if a job opening is a leading indicator of a job being filled and added to the economy, is the lead-time 30 days or 60 days? Depending on what is going on in the labor market, it could be one or the other, and it shifts over time. Again, it looks like we are in a period where it is shifting, which makes it really hard to know whether or not, for example, the July numbers should be based on May or June’s LinkUp data.

So with all that pre-amble out of the way, not to mention the fact that July is a highly seasonal month, we are forecasting that the U.S. economy added a net gain of 225,000 jobs in July due to the slight drop (-2.6%) in the blended average of new and total job openings in May. Our numbers are a bit below the consensus forecast for the month.

July 2014 Jobs Forecast

While a net gain of 225,000 is a drop from the solid 288,000 jobs added in June (which will be undoubtedly adjusted¬†in 13 minutes), it would be the 6th straight month of monthly gains above 200,000. Even more encouraging is the fact that in June and now again in July, our forecasting model shows monthly increases in job openings from the prior month which bodes well for job gains in August and September. Unfortunately, our raw data has shown declines in new and total job openings for the past 3 months, making it somewhat difficult to garner any confidence in our forecast. Given our solid track record over the past few years, however, I’m sticking with my model and it’s indicating that we’ll see 225,000 in 8 minutes and strong numbers for the rest of the 3rd quarter.


  1. John A / Aug 1 2014 6:02 pm

    Have you ever thought about making a jobs forecast based on the # of job postings that *disappear* from company websites each month?

    In other words, let’s say a company is looking for an engineer. They post an announcement on their website describing the position they’re looking to fill. The way you do your methodology now, you assume they’ll fill it in 30 or 60 days. But needless to say, that’s not realistic for many positions (I’ve seen particular jobs being advertised for a year). Eventually the company will fill the position. When they do, they’ll *remove* the announcement from their company website. That there tells you the position has (finally) been filled. So what you need to do is, (somehow) track each individual listing, and when it’s *removed* from the company website, assume it’s been filled and count *that* as a new job added.

    You would probably have to assume a certain percentage of announcements removed from company websites are companies either giving up trying to fill a particular position, or some other reason why they decide not to fill it (company business getting worse, etc). You’d probably have to survey a lot of HR department people to determine what a good # to choose for that percentage might be, and you’d probably have to do that survey periodically to see if more, or less, companies are choosing not to fill advertised openings.

    Still, I think that would be a more realistic methodology.

    Another possible thing you could do is, instead of just assuming a 30 or 60-day lag between job posting and job filling for *all* job categories, use a separate number for each category of job; e.g. engineer jobs will take longer to fill than nursing jobs. As demand for certain job types shifts throughout the course of a year, and over the course of a business cycle, by using that methodology you’d be able to more easily capture changes in how long job openings are filled for the job market as a whole.

  2. Toby Dayton / Aug 12 2014 11:36 am


    Thanks for the really thoughtful comment. Your suggestion is an excellent one, and we have looked at jobs that roll off the site, but in a little different context. Just this year, we started looking at a number of different data points to try to get at the question of how long job openings were in our search engine as a proxy for the velocity of hiring and therefore, to some extent, the strength of the labor market. One of the reports we started running is, in any given month, what jobs were no longer on the site at the end of the month that were on LinkUp at the beginning of the month. We then take that data set and count how long those jobs were on the site prior to being removed (from the employer’s site and therefore also from LinkUp) to get a data point that we use to measure job duration.

    Your suggestion of also using that same report to get a data point that might correlate to job growth either that month or the following month is most definitely something we will look into and start analyzing. You are exactly correct in stating that while not every job listing removed from a company’s website is perfectly correlated to a hire being made, the vast majority are, and we can determine the ratio by looking data over an extended period of time. The other factor, again a minor one, is that not every job that gets filled results in a job listing being removed from their corporate career portal. Many companies use ‘evergreen’ listings or a single job opening that covers any number of desired hires. It’s worth noting, however, that this issue also impacts our current methodology, so it’s a factor in both approaches.

    We’ll look at the data and keep you posted as to what it looks like as another indicator of job growth. As to looking at various data points on a sector level, that is something we’ve been working on and we should have a bunch of additional data points by sector later this year.

    Thanks John.


Comments are closed.