Heresy About US Jobs Data

To start with the obvious: Friday’s headline news that the addition to non-farm payrolls slumped to only 126k in March was one of the larger shocks of the last few months, down 138k from the revised 264k added in February. This was the weakest number since December 2013, and was bad enough to cut 10yr treasury yields from 1.91% to 1.8% instantly, before settling at around 1.83%.

The key features were a collapse in hiring in the construction sector (the sector shed 1k jobs in March, having hired 29k in February),  a slump in hiring in the leisure & hospitality sector (13k hired in March vs 70k in February), and slowing hiring in education & health (38k in March vs 57k in February).  In addition, the mining sector continues to shed jobs, with 11k lost in both March and February. These details suggest at least some of the weakness can be attributed to the sectoral slump in the oil industry, and unusually poor weather conditions. In fact, the number of people unable to work owing to bad weather came to 182k (nsa) in March, which is 41k more than usual for March.

But in addition, March’s weak non-farm payrolls data echoes other signals of domestic demand weakness  recently: in February retail sales fell 0.6% mom and wholesales sales fell 3.1% mom and orders for durable goods fell 1.4% (as did orders for capital goods non-def ex-air). In addition,  housing starts fell by 17% mom in February, to the lowest since Jan 2014.  This weaknesses are also linked (both cause and effect) to the mini inventory-cycle which the industrial sector is currently working through.  Finally, one should acknowledge the possibility that the inadvertent tightening of monetary conditions imposed by the strength of the dollar, may also be having a depressing influence on domestic demand.

Having established that the non-farm payrolls number was genuinely weak, and that, although unexpected, it is not totally out of sync with the rest of the economic data,  it is worth exploring an alternative possibility - that the apparent weakness may actually reflect changing labour market behaviour linked to an improvement in labour market morale. Admittedly, such an interpretation seems bizarre, and totally at odds with bond market reaction to the data. 

The case for such an interpretation starts with some of the rest of the data contained in March’s labour market surveys.  First, average hourly wages rose 0.3% mom in March, beating expectations for the second time in the last three months, even though a modest retreat in the average number of hours worked meant that  average weekly earnings actually fell 0.1% mom.  In other words, even though conditions are relatively slack, there is no sign of weakening wage pressure. 

Second,  although the weakness of the Establishment Survey’s count of non-farm payrolls was echoed in the Household Survey’s count of employment - up just 34k - the Household Survey contained strands of information which run counter to a simple ‘bearish’ interpretation. First, although the survey showed those ‘employed’ up only 34k, it also showed those in work up 346k, whilst those unemployed fell by 130k. The difference in the totals is explained by two factors: first, the number self-employed rose by 299k on the month, and the number not in the workforce rose by 227k. Those 227k who fell out of the labour force cut March’s labour participation ratio to 62.7% - a retreat back to the lows seen in Sept 2014 and again in Dec 2014. 

Looking at that 227k rise in the number 'not in the workforce', it turns out that the number wanting a job fell by 169k.

Both these factors, if believed (and the volatility of the Households Survey means there must be a question-mark over its findings), are difficult to reconcile with the straightforward ‘times are tough, so hiring is down’ reading of the data.  That bearish reading is also difficult to reconcile with two other factors: first, the record job openings data, and secondly, the currently very strong consumer confidence readings, which specifically include sharply-improved perceptions of labour market conditions.

Traditionally, such improved perceptions would suggest that wages would need to rise in order to attract new entrants into the market, or to fill the record-high number of openings currently unfilled.  In the absence of such wage-rises one would expect to see labour participation rates not rising in line with the economic cycle, whilst rising self-employment coexisted with falling unemployment. Put bluntly, if people consider that the economy is strengthening and that the likelihood of finding a job if needed has improved, then employers may find themselves needing to raise wages more than they had expected in order to attract new employees. 

Ever since the Great Recession, there have been numerous attempts to guess at what level the US might be fully-employed, in the sense of the cycle provoking accelerating wage rises. As time has gone on, the estimated unemployment rate at which full employment is reached has been revised down and down.  It is consistent with the totality of March’s employment surveys  that that rate is now being reached at just the point at which exogenous factors (oil prices, weather, West Coast port problems) have engendered a sub-cycle of inventory and capex related economic weakness.   

It is perhaps deep heresy, and maybe just plain stupidity to suggest it, but the data overall is consistent not simply with the subcyclical weakness, but also with the underlying improvement and strength of the US jobs market. In which case, the immediate bond-market reaction may turn out to be . . . . wrong.