• RSS Feed

The LinkUp Blog The Industry's Best-Kept Secret

March 31, 2014 / Toby Dayton

LinkUp Forecasting Model Scores A 58% In 2013 (That’s A Solid A+ In Prognostication 101)

With the release of the February jobs report, the Bureau of Labor Statistics (BLS) has issued its final, 60-day revision for its December jobs number, putting to bed all 12 months of 2013. As such, I thought it would make sense to look back and assess the effectiveness of our model last year in forecasting job growth in the U.S. By no means scientific, I looked first at our forecast each month in relation to the initial BLS release as well as subsequent BLS revisions. In addition, I looked at our forecast in relation to the consensus forecast and whether or not we were directionally accurate in forecasting job growth that was better or worse than the consensus estimates. I also looked at the longer-term, multi-period trend in our data as compared to the BLS data. And finally, I looked back at the commentary each month to assess how insightful our analysis was (or wasn’t) in relation to how the labor market shaped up during the course of the year.

For each of the 12 periods, I assigned a pass/fail grade based on the overall mix of criteria outlined above (with full recognition, again, that this grade is both unscientific and self-awarded). The results are detailed in the table below, including 7 months with a ‘Pass’ grade and 5 months with a ‘Fail’ grade. That turns into a 58% score for the year, which I’d argue is pretty solid given the daunting challenge of trying to predict the future in a difficult environment where anything consistently better than a coin flip has to be regarded as remarkable.

2013 Forecast Grades

In going through the analysis for the year, there are definitely a few things worth pointing out, as a well as a few items worth highlighting. In general, I would give our model very high marks for the general downward trend we began identifying in Q1 and consistently articulated as the year progressed. Beginning in February, the blended average of new and total job openings in LinkUp’s job search engine declined month-over-month in almost every month during the year, presenting a pretty startling view of the weakening jobs market across the country.

2013 Monthly LinkUp Jobs Data

As the graph below indicates, the BLS data, when shown by quarter, confirms that same deterioration in the labor market in Q2 and Q3, and I’d argue in Q4 as well if one more accurately accounts for seasonal hiring.

2013 BLS data by Qtr

Even ignoring seasonal factors and accepting the BLS data as is, I’d make the case that the horrible number for December, which surprised everyone in the markets except us, was enough in and of itself to stand as a more telling proxy for the alarming weakness of the labor market in Q4. That’s also why I gave us a ‘Pass’ grade for November despite the fact that we were a bit premature in calling the horrific number that came out a month later.

A couple of other highlights for the year are worth noting. In June, we got the June BLS number of a net gain of 195,000 jobs exactly right:

June 2013 LinkUp Forecast on Bloomberg

And in December, as we wrote above, the jaw-dropping jobs number released by the BLS was a shock to almost everyone except us, as indicated by the Bloomberg screenshot below showing that LinkUp contributed the lowest forecast to Bloomberg’s consensus forecast.

December 2013 Jobs Call

In terms of a highlight centered around our commentary, I’d point to our prediction in early August that the Fed would not begin tapering in September. At the time, virtually everyone in and around the markets was wildly secure in their belief that the Fed would unquestionably begin curtailing its Quantitative Easing regimen in September, leaving the size of monthly reduction as the only item open for discussion. Needless to say, we were quite alone with our contrarian call.

And lastly, because I am a huge fan of analogies, I’d nominate as my two favorites for the year the October post comparing Ted Cruz to Walter White (aka, Heinsenberg)…

Lego Meth Lab #1

and the December post that compared me to William Shatner in his infamous Twilight Zone episode…(the monster was on the wing, by the way)


It was a great year for our model and we look forward to similar results in 2014.