Unemployment , Birth/death And Projections

Bob, The following charts shows the impact that the birth/death model appears to have on the correlation between the alternative unemployment series (U series, i.e U3/U6) and the standard CPS #’s that a re published every month (i.e 700K jobs lost last month, Table B1 data, http://www.bls.gov/news.release/empsit.t14.htm). -D Ratio (D=Δ): the delta ratio shows the correlation between U6 and CPS table B1 data. A D ratio value of 2% would mean that 100,000 jobs added to the market would decrease U6 by 2%.. For the period covering 1994-2009 the median D Ration was 0.07%, the mean was 0.24% and the Stdev was 0.85% The data is not very tight due to the number of very large outliers in the data. However if we remove the outliers greater then 1% the stdev drops to about 0.25% and the mean tightens to about 0.15%. The large outliers tend occur when there is a very small change ± in the CPS numbers and so the actual impact on U6 seems to be limited. Also note that this correlation may be applied to U3 using the correlation U3 = 1.77*U6 16

12.0% D Ratio (Change in U6 per 100K people) 10.0%

14

U6 (SA)

12 10 8

6.0%

U6

D Ratio

8.0%

6

4.0%

4 2.0%

2

Feb-09

Feb-08

Feb-07

Feb-06

Feb-05

Feb-04

Feb-03

Feb-02

Feb-01

Feb-00

Feb-99

Feb-98

Feb-97

Feb-96

Feb-95

0 Feb-94

0.0%

D Emp = Δ Employment, the number of jobs in 100,000’s lost or gained per tableB1 (non famr employees) 12.0%

800 D Ratio (Change in U6 per 100K people) D Emp

600

10.0%

400

6.0%

0 -200

4.0%

-400 2.0%

-600

Feb-09

Feb-08

Feb-07

Feb-06

Feb-05

Feb-04

Feb-03

Feb-02

Feb-01

Feb-00

Feb-99

Feb-98

Feb-97

Feb-96

-800 Feb-95

0.0%

D Emp

200

Feb-94

D Ratio

8.0%

Using the data obtained from building the correlation between CPS B1 data and U series allows a projection to be made for unemployment. Given a number of current and expected future economic factors, I expect to see an average of 500,000 to 700,000 jobs lost per month for 2009. Using the median D Ratio correlation for the low projection and a D Ratio value of 0.5% for the high projection. U3 & U6 Projection 2009 22.0%

U6 Low Estimate U6 High Estimate

20.0%

U3 Low Estimate 18.0%

U3 High Estimate

16.0% 14.0% 12.0% 10.0% 8.0% Mar-09

Apr-09

May-09

Jun-09

Jul-09

Aug-09

Sep-09

Oct-09

Nov-09

Dec-09

From what I understand of unemployment data collection and calculation methods used by the BLS, I would tend to think that a more accurate estimate of monthly job loses might be to go backwards from U3/U6. You could do that by taking the change in U3/U6 and using the median value of 0.07% change in U6 for every 100,000 jobs lost/gained. Example: Jan to Feb 09, U6 went from 13.9% to 14.8%, Δ = 0.9% 0.9% / 0.07 = 1.2 million jobs lost If you want to work backwards, 0.1 seems to be a less noisy number to use then the median of 0.07 this idea needs to be fleshed out a little further. But the whole birth/death methodology seems to add a lot of noise to the data. One last thought, the U3/U6 projections could end up being optimistic if we have a serious downward acceleration such as the European banks seeing a large scale run or a general dislocation scale event.