Steven Hansen has observed for some time that WIUC (Weekly Initial Unemployment Claims) around 400,000 mark the region where job gains and job losses are likely to offset, producing small monthly employment changes, either gains or losses. This general observation has intrigued me so I have taken a look at the data details.
Postulate 1: Higher initial unemployment claims mean that net job losses are more likely to occur.
If this is true (and it is obvious that it should be) then there should be a correlation between the quantity (hires – separations) from the U.S. Dept. of Labor JOLTS report and the number of weekly initial unemployment claims, here after called Claims. In the following graph the number of claims average for each month was estimated by taking the average of the four or five weekly readings of the 4-week moving average of the claims reported (seasonally adjusted). The end of data for each month was the week containing the 12th of the month, since that is the survey week for employment data reported each week in the Dept. of Labor Employment Situation Report.
The following graph shows the expected correlation and the postulate is confirmed. The linear trend line for the data crosses the 0 value (x-axis) for net jobs gain (from JOLTS) at 391,000. This is very close to the approximately 400,000 identified by Steve. However, note that the correlation is not very tight; R2 is approximately 0.5 which is only fair correlation. Detailed list of observations follows the graph.
- The linear trend line shows a fair correlation.
- The linear trend line crosses the vertical axis placed at (hires – separations) = 0 just below the claims value of 400,000 (at 391,000).
- If there was good or excellent correlation more of the data points would be expected in quadrants 1 and 3.
- With the horizontal axis placed at the estimated value 391,000, the populations in each quadrant are: Q1 = 38%; Q = 13%; Q3 = 39%; Q4 = 10%. Quadrants 1 and 3 combined contain 77% of all data points.
- If the horizontal axis is moved above the 392,000 to 400,000 region, the populations of quadrants 1 and 3 combined declines gradually. At 450,000 the total is 69%. Below 392,000 the total for Q1 and Q3 remains little changed until one moves below 350,000 when it declines precipitously. At 350,000 the count for Q1 plus Q3 is 75%.
- There are three data points in quadrant 2 that fall outside the designated interval from the origin of the graph (intersection of horizontal and vertical axes as shown). The interval is the radius that contains all the other data points for quadrants 2 and 4. No particular significance is attributed to these three outliers. If the three points are omitted the correlation in improved only modestly (R2 = 0.58, r= 0.76, correlation still fair).
- There is a clear correlation between higher WIUC and greater job separations compared to hirings. However, the correlation is only fair.
- The center of the range of WIUC values associated with separations becoming greater than hirings (net job losses) is at 392,000 average over a month of the 4-week moving average value.
Postulate 2. The relationship between WIUC and net job losses may be different for times when WIUC is increasing compared to times when it is declining.
The following graph shows how the WIUC data is divided into the two groups:
The following two graphs show the correlation scatter diagrams for the two data sets.
The relationship between WIUC and net jobs gain or loss is significantly different when WIUC is rising and declining:
- When WIUC is rising the center of the crossover region from job creation to job loss is at WIUC about 358,000.
- When WIUC is declining the center of the crossover is about 429,000.
- The correlations are much better than for all the data mixed together.
We are in a declining WIUC era (until proven otherwise) and data for the past two months has been has been close to the center of the crossover region. After eight weeks below 400,000 the 4-week moving average of WIUC has been between 440,000 and 400,000 starting April 23. During the time since the return of WIUC above 400,000 job growth has been low, but has remained positive. This supports the supposition that we are still in a declining WIUC regime and have not reached an inflection point. If that were the case (WIUC reversing to an uptrend), past history would project that we would be seeing job losses rather than job gains each month.
If we see WIUC remain in the 400,000 and above region in the coming months and we continue to see small (or larger) employment gains each month, it is unlikely that WIUC is reversing to an uptrend. However if we see one or two months of job losses the improving trend in unemployment claims may be over. If we see a third month of job losses, it is highly likely that WIUC is headed higher, back above 450,000 or more.
Steve has been reporting that the 400,000 region for WIUC represents a battleground between hiring and firing. This analysis shows just where the troops are arranged on that battleground.
Is There Really a Mystery Why More Jobs Are Not Being Created? by Menzie Chinn
June 2011 JOLTS Shows Little Growth Pressure on Employment by Steven Hansen
Employment Friday Is Like Waking Up Next to an Ugly Woman by Steven Hansen
July 2011 BLS Employment Better – But Still Not Good by Steven Hansen
July 2011 ADP Employment Data Shows Continued Degradation by Steven Hansen
Dissecting the Employment Numbers by Elliott Morss