The Math Forum

Search All of the Math Forum:

Views expressed in these public forums are not endorsed by NCTM or The Math Forum.

Math Forum » Discussions » sci.math.* » sci.stat.math

Notice: We are no longer accepting new posts, but the forums will continue to be readable.

Topic: Labor statistics
Replies: 2   Last Post: Jul 11, 2011 5:52 PM

Advanced Search

Back to Topic List Back to Topic List Jump to Tree View Jump to Tree View   Messages: [ Previous | Next ]

Posts: 57
Registered: 8/26/09
Re: Labor statistics
Posted: Jul 11, 2011 5:52 PM
  Click to see the message monospaced in plain text Plain Text   Click to reply to this topic Reply

Rich Ulrich <> wrote:
> On Mon, 11 Jul 2011 15:43:38 +0000 (UTC), root <>
> wrote:

>>Two recent employment reports have differed so widely as
>>to suggest to me that either something grossly improbable
>>has happened or that a serious error has been made.
>>Last Thursday the company ADP, which provides data processing
>>services for over 500,000 companies, published a report
>>that non-farm employment rose by 157,000 jobs in the month
>>of June. The ADP report was based upon a survey of 340,000
>>companies. That report can be seen here:

> - I found the report at the same URL except that PDF was pdf.
> Apparently case sensitive on their server, and changed?
> After a "page error", I stripped off the name down to .com/,
> and found the intro to the report, which offered the link as
> which otherwise looks the same.

I am using the Chrome browser and either .pdf or .PDF got
me to the page. Thanks for going to the effort.

>>The very next day The Bureau Of Labor Statistics published
>>their report that during June the non-farm employment
>>rose by 18,000 jobs. The BLS report was based upon a
>>survey of 140,000 companies. This report can be seen here:
>>The historical agreement over the last 10 years of these
>>two employment estimates can be seen in a graph covering
>>the period of 2005-2010 here:
>>Stock markets around the world were buffeted by the
>>two reports, first up on the ADP report and down after
>>the BLS report.
>>I have tried to make simple models of how ADP and the
>>BLS might have generated these numbers. These models
>>result in a ratio distribution whose parameters are
>>not available, but assuming that the two estimates
>>should be highly correlated and that the width of
>>either distribution is much smaller than the median,
>>I find the discrepancy to be too great to explain by
>>sampling. It seems to me that some serious error has
>>been made.
>>I invite comments by interested parties in this group.

> ADP comments that their own report does not directly
> reflect their own clients, but is weighted to match the BLS
> report. So they are trying to be the same.
> Chart 2 (page 3, in the corrected citation above) shows
> the match between the two sets of reports. ADP seems to
> reflect its larger base of companies (or differences in
> methodology?) by having a curve that usually is less jagged.
> I had trouble figuring out the scaling at first, because the
> totals I could project seemed wrong. After looking at the
> other reports, includind BLS, I figured out that this is,
> indeed "private, non-farm" jobs, where the job-total is about
> 107 million. So a 0.1% change is a 100,000 change in jobs.
> Looking at the monthly deviations, instead of the overall
> agreement, I count at least 2 deviations per year that look
> like over 0.1%, for this 10-year chart -- 100,000 jobs or more.

Strangely, when I first looked at the ADP report I didn't see
what are now represented as Charts 2-4, but instead a large
chart showing data from 2005-2010. What I see now doesn't
have a Chart 1.

You are absolutely correct that the current Chart 2 shows that deviations
greater than 0.1% are somewhat common. It looks like Jan of this year
the BLS report was 0.2% higher than ADP.

> So, I end up comforted that the BLS estimate has a good
> chance of being wrong this time. But it looks to me like variation
> within the typical range.
> ***
> By the way, I was lately wondering, What is the new-hire
> rate for a month? (Does an increase of 100,000 mean that
> 1 million started and 900,000 quit?) Or are there too many
> off/on jobs, like construction, to make this easy to count?

I think it means a net of 100,000 and any way that can come
about. Thanks for your effort on resolving my problems with
the two reports.

Point your RSS reader here for a feed of the latest messages in this topic.

[Privacy Policy] [Terms of Use]

© The Math Forum at NCTM 1994-2018. All Rights Reserved.