Job Creation numbers assume "death" of at home job means new at home job

[papau]

Prediction: Birth/death add on to actual payroll job growth in Sept will be 35,000!

It looks like last years number will simply reappear this year - which is confusing as we are told that the justification for assuming increases in at home jobs is the increase in GDP - as in the more GDP growth the larger the number of at home businesses. So a revision down in GDP one would think would cause a negative B/D effect.

But as yone can see from the results and from the DOL explanation, the power of assuming that DEATH of an at home job implies BIRTH of a new at home job with a rate of growth equivalent to actual payroll results - over the past few months causes the B/D number to be very stable and in effect follow the seasonality of the payroll numbers themselves. So last years -83 was this years -91, and last years 124 is this years 120, and we would expect last years 34 to be as you suggest around 35. The effect of GDP change appears minimal.

Almost makes the 2 factor formula they put in place in 03 a way to add 800,000 pretend jobs to pre-election job reports - with no media the wiser! Indeed - Mr Bush - about that new job growth - that turning the corner and right direction - that would not be yet another lie from the mouth of Bush? Credibility and Bush do not belong in the same sentence.

Below is from the DOL. I had not noticed that the April 03 forward numbers reflected a "new" B/D 2 factor estimate approach - meaning an interesting opportunity to wire the numbers for the election.

In any case with detail on the "UI universe micro level database" and how it is used not known - as in "is GDP growth really important to final number?" - all one sees is the seasonality and constant growth in at home jobs because the DOL is "simply not reflecting sample units going out of business" - a process they want to sell to us as a valid way to do this job counting business!

These DOL folks are young and have forgot more math than I ever knew is what I once thought - but now I begin to wonder! :-)

:-)
=================================================================
From the DOL:
In 2004, the CES sample includes about 160,000 businesses and government agencies drawn from a sampling frame of Unemployment Insurance tax accounts which cover approximately 400,000 individual worksites. The active CES sample includes approximately one-third of all nonfarm payroll workers. The sample-based estimates are adjusted each month by a statistical model designed to reduce a primary source of non-sampling error, the inability of the sample to capture on a timely basis, employment growth generated by new business formations.
There is an unavoidable lag between an establishment opening for business and its appearing on the sample frame and being available for sampling. Because new firm births generate a portion of employment growth each month, non-sampling methods must be used to estimate this growth.
Earlier research indicated that while both the business birth and death portions of total employment are generally significant, the net contribution is relatively small and stable. To account for this net birth/death portion of total employment, BLS is implementing an estimation procedure with two components: the first component uses business deaths to impute employment for business births. This is incorporated into the sample-based link relative estimate procedure by simply not reflecting sample units going out of business, but imputing to them the same trend as the other firms in the sample.
The second component is an ARIMA time series model designed to estimate the residual net birth/death employment not accounted for by the imputation. The historical time series used to create and test the ARIMA model was derived from the UI universe micro level database, and reflects the actual residual net of births and deaths over the past five years. The ARIMA model component is updated and reviewed on a quarterly basis.
The net birth/death model component figures are unique to each month and exhibit a seasonal pattern that can result in negative adjustments in some months. These models do not attempt to correct for any other potential error sources in the CES estimates such as sampling error or design limitations.
Note that the the net birth/death figures are not seasonally adjusted, and are applied to not seasonally adjusted monthly employment links to determine the final estimate.
The most significant potential drawback to this or any model-based approach is that time series modeling assumes a predictable continuation of historical patterns and relationships and therefore is likely to have some difficulty producing reliable estimates at economic turning points or during periods when there are sudden changes in trend. BLS will continue researching alternative model-based techniques for the net birth/death component; it is likely to remain as the most problematic part of the estimation process.
The table below shows the net birth/death model adjustment used in the published CES estimates since the establishment of the most recent benchmark level for March 2003.

2003 Net Birth/Death Adjustment (in thousands) Supersector Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec
Natural Resources & Mining
-1 1 1 0 1 1 1 -1 0
Construction
13 35 28 -8 16 9 8 -7 -8
Manufacturing
-15 5 5 -29 6 3 -7 3 1
Trade, Transportation, & Utilities
-4 21 18 -19 17 17 13 17 18
Information
-3 4 0 -4 2 0 -1 3 3
Financial Activities
9 8 6 -11 8 4 14 7 13
Professional & Business Services
61 32 21 -22 31 15 18 10 9
Education & Health Services
32 6 -4 -20 14 12 26 10 7
Leisure & Hospitality
29 72 83 40 24 -29 -27 -14 15
Other Services
7 8 6 -10 5 1 0 2 4
Total
128 192 164 -83 124 33 45 30 62



2004 Net Birth/Death Adjustment (in thousands) Supersector Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec
Natural Resources & Mining
-4 0 1 0 1 1 0 1
Construction
-66 7 27 38 39 31 -7 16
Manufacturing
-38 4 7 3 8 7 -22 4
Trade, Transportation, & Utilities
-61 9 22 19 23 22 -15 21
Information
-5 5 2 2 3 1 -6 3
Financial Activities
-12 10 9 16 7 10 -18 8
Professional & Business Services
-95 27 31 66 26 24 -32 24
Education & Health Services
-6 15 10 37 11 -2 -10 17
Leisure & Hospitality
-24 33 37 80 71 81 30 21
Other Services
-10 5 7 9 6 7 -11 5
Total
-321 115 153 270 195 182 -91 120
Note: There is no net birth/death model adjustment for the government supersector.



Additional information on the CES Birth/Death Model is available under Frequently Asked Questions.

Historical Birth/Death factors and Bias factors are available at www.bls.gov/ces/cesbdhst.htm.

[mhr]

1. Thanks Papau

For those of us living on main street, not wall street, this reflects reality.

Too few jobs for too many unemployed workers.

[jwcomer]

2. Some thoughts on the B/D model

I agree with papau that the seasonal component is currently the strongest. However, I don't think the B/D model is unreasonable and I don't think it is politically motivated (but I'm young and perhaps naive.)

To clarify for those who aren’t closely following the B/D model debate: the B/D model has two factors. The first is a 'virtual' imputed factor. The second is a correction factor and this is the number that is reported in the DOL tables. The first number isn't reported at all.

It is easier to explain by example. Suppose that the establishment survey was over 1000 companies. And let's assume 5 of those companies go out of business. The DOL assumes that another 5 identical businesses are created which do not YET show up in their survey. So the DOL uses the remaining 995 companies to impute the employment situation for those 5 new companies. To use specific numbers, if the 995 companies started the period with 199k employees and end the period with 200k employees, then the DOL assumes the 5 missing companies grew by the same ratio. If those 5 companies had 2000 employees, the DOL assumes the birth companies have 200/199*2000=2010 employees bringing the total number employees up to 202,010.

Eventually the DOL survey catches up. So after several months, the DOL can look back and see how well its estimates match reality. This is the basis for the second factor. This factor is composed of several different correction terms. The largest appears to be a seasonal correction. Let's say that historical the B/D model undershoots by 30k during this period; then the DOL adds 30k into the B/D model. After correcting for the seasonal term, there remain residual errors, so the DOL uses a generic model (known as ARIMA) to further correct for these errors. This ARIMA term has little to no physical meaning and may have only minor predictive power. On a technical note, the seasonal component is part of the ARIMA model, but I think it is better to think about them separately. Finally, it has been reported that the DOL includes a factor to correct for growth in GDP. I suspect that any such factor would amount to a multiplier on the magnitude of the seasonal/ARIMA terms.

If the GDP factor is a multiplier, then it is likely going to be the annualized growth since the last benchmark (March, IIRC.) In this case, the B/D model does not assume any new births or deaths due to changes in GDP, rather the GDP growth is being used as a proxy for employment growth for the purposes of adjusting the magnitude of the B/D correction factor. This seems reasonable enough, but would be a nearly unnoticeable effect. I should point out that I have no knowledge of the GDP factor aside from papau’s anecdotes, nor have I seen anything by the DOL as to its existence.

The DOL does this two factor method for each industry separately. On the whole, this might work well but for individual industries, I'm less certain. Consider the loss of a company like Enron. Imputing a new identical company is absurd and leads to a gross over count. It is true that much of the same work is probably being done elsewhere; but this is almost certainly by a competitor who likely hired former employees. Thus a massive bankruptcy could lead the DOL to the erroneous conclusion of a massive increase in employment; effectively counting the laid off employees twice. I don't see any mechanism by which the DOL handles large business failures; but they employee many bright people and it is inconceivable that they are not aware of this issue, so hopefully they have a good solution.

I believe the reason that the seasonal factor is currently the strongest is that the ARIMA model is probably of low order. Hence as long as the current B/D correction factor is nearly the same this year as last, the ARIMA factor decays leaving only the seasonal component. This is especially true if the AR and MA terms are all much smaller than one.

The seasonal factor does not affect the seasonally adjusted survey numbers because it drops back out; leaving only the ARIMA term, which is currently quite small. So if the DOL is using the B/D model for political purposes, they aren't getting much out of it.