|
US Unemployment Projection |
 |
 |
 |
|
Written by Dr. Willy Gerber
|
|
Friday, 13 February 2009 13:22 |
US Unemployment Projection
Based on historical data we can find a typical pattern how unemployment rises and falls in each US crisis. The following Applet allows you to fit the current unemployment figures to the studied pattern and estimate the probable future evolution:
(the applet will load in about to 5-20 sec depending on the speed of your internet)
Explanation of the Simulator
The simulator is based on the historical data of the 9 unemployment crisis from 1948 to 2005:
Data from the U.S. Bureau of Labor Statistics
Each data set is scaled for comparison and shown on the lower part of the Applet:
The Applet also shows a magenta line, representing the mean value of the 9 scaled event curves. You may see the original data de-selecting the checkbox called "Scaling".
The pattern curve can be now scaled to the actual real unemployment data of the last months. If the fit is made with the restriction that the last data (December 2008) is on the curve, the best fit will look like:
If you prefer to fit without the restriction of the coincidence of the last point, switch from the radiobutton "FP" (Fix Point) to "Ge" (General):
Now the curve will look like:
If you like to fit the curve manually, you can slide the bar
and find the best "looking" fit like the following:
Method Precision
The precision can be evaluated by comparing the projections with a growing data simple approaching the real maximum. At the beginning the method has some problems to estimating the maximum because the gradient didn’t change. At the moment the unemployment rate breaks with the grow trend, the system is capable of fitting the right curve converging on the month and on the unemployment maximum:
| Data including | Peak Date forecasted | Maximum | Increment | | January-09 | April-09 | 8.20% | 3.70% | | February-09 | May-09 | 8.80% | 4.20% | | March-09 | June-09 | 9.20% | 4.60% | | April-09 | July-09 | 9.70% | 5.00% | | May-09 | July-09 | 10.20% | 5.50% | | June-09 | June-09 | 9.50% | 4.90% |
Conclusion
Actually the best fit will project a maximum of 8.4% for Mai 2009. But the method is not foolproof, you may also get rather good fit in a extreme situation like:
with a maximum of 32.1% for March 2010!
This example shows the power but also the limitations of the model. At the end it's a tool for learning about the systems "rules" (like the very typical pattern for the unemployment rises) but also the need of insider knowhow for use and interpretation of the results.
Note: Jave applet was build with the help of Easy Java Simulations (EJS) (http://fem.um.es/Ejs). Mathematical and Physical Methods can be found in gPhysics.net (http://www.gphysics.net).
|
|
Last Updated ( Monday, 07 December 2009 13:45 )
|
US Unemployment Projection