Upon learning that the US economy added a dismal 38,000 jobs in May, the worst since September 2010, we simply said that there was no way to spin the number as anything but atrocious.
Well, as it turns out Barclays has a lot more to say about the May jobs number, and the overall employment growth slowdown in general. In a recent note, Barclays points out that since 1960, an employment growth slowdown has preceded every NBER-defined recession, and when the employment growth falls below the average monthly increase during an expansion, a recession follows in about 15 months time. Given the fact that employment growth has fallen short of the current expansion average for four of the past five months, Barclays says that there is now a 30% probability that the US will enter a recession within 12 months.
As Barclays explains
We have long been of the view that labor market data provide the best real-time indicator of underlying economic activity and are the best leading indicator with respect to turning points in the business cycle (see Slow growth? Labor markets say relax, January 15, 2016). Since 1960, an employment growth slowdown has preceded every NBER-defined recession, with employment growth falling below the average monthly increase during the expansion about 15 months before the onset of the next recession (Figure 1). It has now fallen short of the current expansion average for four of the past five months. Based on this slowing, we see the risk of a recession as elevated, at about 30% within 12 months. Absent data revisions, a weak June payroll number would substantially boost this probability, especially if that weakness were accompanied by deterioration in other data.
Furthermore, JOLTS openings have softened, and horrendous manufacturing indicators point to a weak employment outlook.
Consistent with a slower pace of hiring in the employment report, the number of job openings is moving lower. JOLTS openings data show considerable softening after a late-year surge in 2015 (Figure 3). The composite help wanted index produced by the Conference Board turned negative in February on a y/y basis and has moved deeper into negative territory since.
Signs from the business sector, especially manufacturing and energy, are gloomy. Already showing weakness for several years, the growth rate of orders and shipments moved further into negative territory in early 2016 (Figure 4). The poor performance in these series is consistent with the down step in manufacturing activity and employment that has occurred so far this year. These series likely also reflect the ongoing weakness in the energy sector.
And finally, using an internal forward-looking employment model that correlates well with the actual employment growth, the model is predicting an outright decline in employment growth. According to Barclays, the model on its own calls for a recession in just three months - an estimate on the very short end of the 9-18 month lag between the slowing in the labor market and the onset of a recession that has been seen historically.
Since 1990, in addition to a relatively high r2, the model does a good job of matching the cyclical contour of employment growth (Figure 5). Starting the lag structure of the data at six months is clearly a bit too long. The model tends to turn just a touch later than the actual data. We maintain the structure to avoid having to project the explanatory variables to arrive at an employment forecast – essentially tying our hands on the model fit.
The model is calling for an outright decline in employment growth (Figure 6). Since early 2015 (and keep in mind that the model is out of sample since 2014), it follows the contour of the realized data closely, matching even the slowdown in job growth last summer. Clearly, and looking through any negative effects on the data from the Verizon strike, the model sees firmer underlying job growth in May (107k) and expects a slight uptick in employment growth in June (114k). Thereafter, the model decelerates quickly as the lagged effects of weak Q1 shipments and orders begin to weigh more heavily on the data. In plain language, this model is calling for a recession in three months, an estimate on the very short end of the 9-18 month lag between the slowing in the labor market and the onset of a recession that we have seen historically.
Noting that the original model may be overly influenced by the energy sector, Barclays tweaked the model to exclude orders and shipments. The result was similar, however the recession would likely hit in the middle of 2017 - said otherwise, things are bad, and if we ignore the energy sector, things are still bad.
Of course, the model output is weighed down by the very negative orders and shipments data, which may be overly influenced by the relatively small energy sector. At the expense of some of the historical goodness of fit, we can exclude orders and shipments. Visually, the model results are similar, with the output continuing to follow the cyclical contour of the labor market over time (Figure 7). Looking at the output since the beginning of 2015, the model saw an employment growth peak of 279k in December 2015 and has since shown a trend slowing. The model output has now been below the expansion average in eight of the past nine months. The more gradual slowing would be consistent with a recession in the middle of 2017.
* * *
So there it is, even when the energy sector gets excluded from Barclays forward-looking employment growth estimates, if labor is the indicator used to determine when the US will enter a recession, look for a recession to hit in mid-2017. Then again, maybe May's jobs number will be magically revised up to the average, and NIRP will start generating real economic demand, and all will be right in the world - we'll go with the former.