Tuesday, October 25, 2016

I hope everyone is expecting huge job reports for October, November and December

Seasonally Adjusted Non-Farm Jobs Change 2013-Present:  Comparison between Griz Method/Simple YoY and BLS Monthly Reports
Source Data:  BLS
Total Nonfarm 2016-10. xlsx

For awhile now I've been showing how wild the monthly BLS seasonally adjusted jobs reports are.  I've said the timing of bad reports has done a great job of holding down interest rates, and the inevitable makeup reports have done an equally good job of pumping up stocks.  The latest cases occurred this summer and fall, with good reports pumping stocks, and then bad reports tabling rate hikes until December after the election.  How convenient!

The chart above shows the official monthly job gain was very low the last two months.  So it is apparent make up reports are on the way just by looking at it.   However, some math and deeper analysis confirms this.

First some basics.  The Griz method/YOY method of seasonal adjustment, moving averages, and moving averages from the seasonally adjusted data set all point to annual job gains pace of 2.4 million or 200k/month.

I decided to look at the BLS seasonally adjusted data set a bit closer to see if there was a way to predict coming reports a little better.  So what I decided to do a 12 month summation of the reported monthly gains.  This really does the same thing as a 12 month moving average, but it provides a table that provided additional insight into how this mathematically works.  What I found is that each month, is that the sum of the most recent 12 month gains lines up relatively closely with they YoY change from the nonadjusted data set.  Makes sense.  However, the summation can vary up to +/-5% from the YoY number.  5% error is about 120k on 2.4 million, large but not so large that it is always obvious without specifically looking.  It could mean BLS reports 190k monthly gains instead of 200k.  What is also glaring is this 5% error is far less than the wild variations that take the reports into the low to mid 100k range and near 300k some months, when it is pretty clear the numbers have really been in the 200k-240k range.

So What is Coming  

Using the table of 12 month summations shown below it becomes somewhat trivial to predict the coming months reports assuming recent hiring trends remain basically intact.  At this point based on weekly jobless claims that seems to be the case.


Table of 12 Month Job Gain Summations using BLS Monthly Reported Seasonally Adjusted Gains (In 1000s of jobs)  
Data Source:  BLS
Total non farm, adj 2016-10.xlsx

Looking at the table we can see the 12 month summation for October 2016 is 2.152 million which includes 0 gains for October which hasn't been reported yet.  So if we assume the real trend is around 2.4 million like it has been for the last 5 months, that means BLS needs to report a 291k gain to get the numbers to line up exactly with reality/the Griz method/make the long term averages right.  If we factor in potential skewing and other random factors from the BLS, the likely range for October is 250k-300k.  Far above expectations with August and September reported around 160k.

Then if we apply a 291k gain for October, the expected reported gain for November should the trend hold is about 287k.  Including 287k for November then predicts a 271k reported gain for December should the trend hold.

My calculations show that currently about 300k jobs are "banked" due to under reporting in early 2016.  Analysis going into 2015 shows when this "banked" number is close to 300k, a huge report is in store.  These huge reports quickly eat into the bank.  A 300k report would reduce the bank by 80k-100k.  History also shows that as the bank comes down the likelihood of a huge report the next month drops.  So it is really hard to know whether the December report will end up being really big, but it is very likely both the October and November reports will be in the 250k-290k range.

My investment plans are to short treasuries with ETFs and with futures.  I also expect that this will trigger a massive stock correction or even crash to end 2016 and/or start 2017 when rate hikes become apparent and actually are implemented no later than December.  A November rate hike is not out of the question, but I don't think the Fed will do anything to spook markets ahead of the election.