Posts filed under “Data Analysis”
Today, I want to discuss why the Employment Situation report is the single most over-hyped, over-analyzed, over-emphasized, least-understood economic releases known to mankind. It is also one of least useful economic data points where investors are concerned.
Long-time readers must bear with me, as they have heard this too many times before. But to new readers of mine at Bloomberg View, a few explanatory words are in order:
1) What matters most in the Employment Situation is the overall trend in hiring. Is it continuing what it has been doing for the prior three months? What matters is not whether it is expanding or contracting, but when it reverses that trend.
2) In a labor force of 150 million people, the change in net monthly new workers minus those no longer employed — essentially what the nonfarm payroll report measures — is a rounding error, a miniscule one tenth of one percent (i.e., 150k out of 150m).
3) If you use the correct long-term hiring cycles, you can get a pretty good estimate of likely hiring patterns. Since we are now in a post-credit-crisis recovery period, you should expect a mediocre but slowly improving job creation. That is what has been happening.
4) The model that produces the monthly number is part measurement (establishment surveys), part extrapolation (birth-death adjustment). This is then revised, as more and better data comes in later. The overall NFP number is subject to subsequent revision next month and again the following month. The model gets re-benchmarked, is seasonality adjusted, as one-off weather or other events impact (and that impact then attenuates away) the overall NFP series.
5) Like all models, the Bureau of Labor Statistics model is flawed, but not useless (to paraphrase statistician George Box). That is why looking at the overall trend versus any specific data point is so much more valuable.
6) Last, while hiring is a well-known lagging indicator, we can look at the three components that often provide hints as to future hiring activities. I hesitantly call them leading components, but they do help to shed insight each month into whether we are improving or backsliding. They are: Hours Worked, Wages, Temp Help.
Last, let me over oversimplify this for illustration purposes.
Each month, approximately 4 million people leave their jobs, Retirement, layoffs, career changes, whatever. Another 4 million people start new jobs: recent graduates, re-entries to the work force, job changers. What the monthly Employment Situation report measures — in near real time — is the net changes between those two numbers. Take the total net number of new hires, subtract the job losses, and you get the marginal change in employment.
Since we begin with such a huge number — 150 million-plus — and the monthly net changes are so small (50-200k), the overall change is not especially statistically significant. The net changes amount to about one twentieth and one quarter of a percent. During the height of the financial crisis in 2008-09, the net change was about half a percent (-700k).
Thus, any single 0.1 percent data point needs to be recognized for what it is. It is but one data point in a longer series. And not an especially accurate one initially.
Focus on the longer-term employment trends and be on the lookout for reversals, rather than any given monthly data point.
DATA Has been underutilized in the music business. SoundScan was a revolution, but then the Internet happened and everybody was caught flat-footed and was so flabbergasted at the evisceration of the business model that the only actions taken were rearguard ones, and those were executed late. Don’t confuse the product with the sell. The key…Read More
Billions of dollars, seasonally adjusted, after tax without inventory valuation (IVA) and capital consumption adjustments (CCAdj). Source: FRED I have been trying to explain to some of my more Fed obsessed friends there are factors outside the central bank that matter also. One such factor is corporate profits. As the chart from the Federal Reserve…Read More
click for ginormous charts Source: JP Morgan JP Morgan observes: “Shiller P/E shows the market to be overvalued, but not as extreme if you use the NIPA data.” I’ve never used the NIPA data, so I have no real opinion on it. The two charts look directionally similar, but different in terms of magnitude….Read More
Mike Shedlock takes me to task for our discussion on ACA/Part time Jobs: Did Obamacare Cause an Increase in Part-Time Employment? which led to this response Measuring What Didn’t Happen: Did Obamacare Cause an Increase in Part-Time Jobs? No Says Ritholtz, and Reuters; Yes, Says Mish. I was pushing back against the “ACA caused all…Read More
Excerpt “According to Nielsen, from data provided by managers at Nielsen SoundScan, which collects recorded-music sales information, of the eight million unique digital tracks sold in 2011 (the large majority for $0.99 or $1.29 through the iTunes Store), 94 percent – 7.5 million tracks – sold fewer than one hundred units, and an astonishing 32…Read More