Below is the detailed breakdown of this morning's initial claims data from Joshua Steiner and the Hedgeye Financials team. If you would like to setup a call with Josh or Jonathan or trial their research, please contact
In psychology, cognitive dissonance is the mental stress or discomfort experienced by an individual who holds two or more contradictory beliefs, ideas, or values at the same time, or is confronted by new information that conflicts with existing beliefs, ideas, or values. - Wikipedia
The initial jobless claims data is beginning to give me some cognitive dissonance.
CD 1) 50,000 Feet - Claims are simultaneously at their best and worst level. They're at their best in that they're trending around 300k, which is the historical average of the trough level of claims over the last seven economic cycles. See the first table below for details (courtesy of our Macro team's Christian Drake).
However, they're also at their worst, in that the nature of the trough is that .... it's the trough. It can't get any better. We illustrate this in the second chart below. We've tried to help investors frame up this risk/reward scenario by looking to history for guidance. Specifically, we've compiled data on how long markets have risen after claims have fallen to a specific level.
As the third chart below shows, the claims data has now been sub-330k for 10 months. Looking back historically at the last three cycles, rolling SA claims ran at sub-330k for 24, 45, and 31 months, respectively, before the corresponding market peaks in late 1991, March, 2000 and October, 2007.
CD2) The Energy Paradox - We've written a fair amount recently about the deteriorating labor conditions for energy sector workers brought on by the collapse in crude oil prices since the Fall of last year. Last week's note on the subject can be found here: Concentrated Harm Meets Diffuse Benefit. This week's data shows a further decoupling in initial jobless claims for the US as a whole vs those states with higher energy exposures. We shows this in the two charts below, where we break the 8 energy-heavy states into a basket and index them vs the US since May last year.
However, as we pointed out in last week's note, oil and gas jobs make up just 0.6% of nonfarm employment in the US; if that were cut in half (along the lines of what's happened to the price of oil), it would have only modest repercussions on the labor force at the national level. That said, energy jobs tend to pay better than non-energy jobs, so that needs to be taken into account.
CD3) 50 Feet - This week's print wasn't good. Seasonally adjusted 1-week initial claims rose 25k to 304k, a large jump in absolute terms and a rise that was beyond what consensus expected. We're not aware of any distortions in the data for the most recent week. However, if we look at the rolling data (4-wk rolling averages) then we see a different story. 4-wk rolling SA claims dropped 3k to 290k vs the prior week. Meanwhile, 4-wk rolling NSA claims were lower on a YoY basis by 13.2% this week vs 9.9% in the previous week - the third consecutive week of accelerating improvement in that series.
Our firm places a lot of emphasis on the importance of Rate of Change (RoC). Our overarching gestalt is that what happens on the margin matters most since it determines the next price. The challenge in that approach, however, is which data point to key off of, as the most recent week's number is clearly a miss while the rolling 4-wk average shows ongoing improvement.
When dealing with something as large as the US labor market, we're not going to make an inflection call based on a single week's worth of data, especially not based on a data series as routinely volatile as the weekly jobless claims. That said, anytime a high frequency data series signals potential trouble, our caution flags go up and we place added significance on the next data as we look to see whether the beginning of a new trend is taking hold. For now the trend remains positive, but we've got the caution flag up with this morning's claims number.
CD4) Convergence & Mice - One of the challenges of watching RoC is that you have to have proper context for what you're observing. In Science, mice are often used for experimentation. We'll use them for our purposes as well.
Consider the following thought experiment (again, h/t Christian). Let's say that I have 10 mice in my house in 2012, 5 mice in my house in 2013, no mice in my house in 2014 and no mice in my house in 2015. Here's how RoC looks on that: 2013: 50% improvement, 2014: 100% improvement, 2015: 0% improvement.
No mice is the obvious objective here, the analog is maximum potential employment. RoC would have you believe that 2013 was good (50%) , 2014 was better (100%) and 2015 was bad (0%), but obviously 2014 represented the achievement of the goal (no more mice) so no change to 2015 was as good as could be hoped for since you can't have negative mice in your house.
Claims reach their frictional lower bound around 300k. As a reminder, 300k is the trough average over the last seven economic cycles. As we are now at that trough we should expect that as we begin to lap the 300k level the YoY rate of change will converge towards zero, just as no mice YoY equals 0% RoC improvement. This isn't a bad thing. This simply means we've reached the trough in claims. At this point, the important exercise becomes watching for signs of new mice (a rise in claims).
Prior to revision, initial jobless claims rose 26k to 304k from 278k WoW, as the prior week's number was revised up by 1k to 279k.
The headline (unrevised) number shows claims were higher by 25k WoW. Meanwhile, the 4-week rolling average of seasonally-adjusted claims fell -3.25k WoW to 289.75k.
The 4-week rolling average of NSA claims, another way of evaluating the data, was -13.2% lower YoY, which is a sequential improvement versus the previous week's YoY change of -9.9%
Joshua Steiner, CFA
Jonathan Casteleyn, CFA, CMT