My Photo
Note: Jeff does not accept guest blog posts on A Dash of Insight.

For inquiries regarding advertising and republication, contact main@newarc.com

Follow Jeff on Twitter!

Enter your email address:

Delivered by FeedBurner

Certifications

  • Seeking Alpha
    Seeking Alpha Certified
  • AllTopSites
    Alltop, all the top stories
  • iStockAnalyst
Talk Markets
Forexpros Contributor
Disclaimer
Copyright 2005-2014
All Rights Reserved

« Guess the Pundit | Main | Weighing the Week Ahead: More Excuses for Selling? »

April 05, 2012

TrackBack

TrackBack URL for this entry:
http://www.typepad.com/services/trackback/6a00d83451ddb269e2016303be4768970d

Listed below are links to weblogs that reference March Employment Report Preview:

Comments

Angel Martin

Jeff, I agree with looking at many different series rather than just the CES. My particular favourite is the CPS wage and salaried for non-farm. Even thought it has a large sampling error, it does not have the complexities that say Trimtabs has with changes in witholding, or ADP estimating from the changes in employment of a self selected sample of clients, or model builders like Jeff Miller using data like the Michigan Survey, where there are exogenous shocks that change the impact of the data. All of these measures has some value, the trick is to weight them appropriately.

If one had multiple measures of the same target variable, say, change in employment, with no "non-sampling" error, the statistically optimal approach would be some sort of weighted average of all the measures, with the weights inversely proportional to sampling error of the measure.

However, non-sampling error is a major issue for trimtabs, adp, modellers, and, i would argue the CES. I think the various measures could still be combined in a weighted average, but one would want to weight inversely proportional to non-sampling error as well as sampling error.

In my view the CPS has the least "moving parts" of all the estimates so it has the least likelihood of non-sampling error. Thus i weight it highly, even though it has the largest sampling error of any of the measures.

****************************

On the CES, for employment it actually is a ratio as is explained in this BLS technical note.

http://www.bls.gov/web/empsit/cestntab.htm#tab2a

the monthly employment is estimated via ratio from the previous month, and ultimately by a chain of ratios back to the benchmark.

Jeff, I'm looking forward to your future postings on the ECRI and a possible recession. I think some of the recent data, including this release, has significantly increased the possibility of ECRI's recession call being correct, but i'm not sold yet.

Angel Martin

Jeff, with the release date, I guess we will just have to agree to disagree, but if Statcan can get the data out on thursday, how come the BLS can't?

With the high volume ETFs, an individual investor can trade after hours, so they don't have to wait until 9:30.

As to the wisdom of doing this, maybe it is a mistake to attempt to vary the asset allocation depending on where you think you are in the cycle, but many investors do it.


oldprof

Angel -- on the BLS methodology --

I don't really understand your point about ratios. They do a survey of establishments and then a number of adjustments. It is not really a ratio. They make revisions as more businesses reply, as they change seasonal adjustments, and as they get final data from state unemployment offices.

As to measuring the change in employment each month -- they could ask that question, but it would still be a survey. They would still have to deal with firms that went out of business.

I have written very extensively on all of these topics.

This is why I look looking at the various alternative measures along with the BLS data -- something that I have advocated for a long time.

Sorry I can't go better with the answer.

Jeff

oldprof

Angel -- I'll split this into two responses so that we can all discuss separately if there is more to say.

On the timing of the release -- I agree that it is important information. Maybe the NYSE should adjust its holiday schedule to conform to the releases, which are known in advance. You also did not respond to my point that any such appeal could have been made long ago.

As to individual investors, on a normal NFP day they get the number and then stocks open an hour later. There is often a gap, so the investor is confronted with dramatically different prices. I have a specific question for you:

Assuming that someone (unwisely) believes that he or she should sell stocks based upon this data point, what difference does it make whether they do it an hour later or a couple of days later?

I'll elaborate on the "unwisely" and the ECRI in my ongoing recession discussion, but not here.

Jeff

oldprof

cd7 --

Good one:)

It does seem that he was on the right track with his speech -- employment brighter than the rest of the economic data.

Jeff

Angel Martin

Jeff, I disagree with your argument about the release date. Yes the employment data is for everyone, not just financial markets. But financial markets are the ones that are most time sensitive.

Suppose a retail investor was worried about the ECRI recession call. A release like today with weakness in employment, hours, temp help, retail employment etc... might be interpreted as increasing the probability that ECRI is going to be right.

But, most retail investors do not trade in the futures market and definitely not in the bond market, so they are stuck until monday. The big institutions get to move right away.

Canada got their labour force data out on thursday and I don't understand why the BLS couldn't do the same.

******************************

On an unrelated point about CES methodology. Jeff, your description:

"The BLS tries to estimate total employment in one month, total employment in another, and subtract the two to determine the difference."

is a bit misleading about what the CES actually does. As you know (since you have written about benchmark revisions), the monthly level the BLS estimates is just a ratio which is then applied to the benchmark.

Jeff, maybe you have some ideas, but I think that if the BLS were to design a survey to estimate the change in employment every month, it would still look very much like the CES (maybe with less sample rotation)...

Paul Nunes

Well written Jeff! as always a helpful analysis of market moving events. Have a great weekend!

lou

A very good post regarding the BLS and job stats. Thanks.

cd7

Ben is warming up the presses as we speak...

The comments to this entry are closed.