Unemployment - Improving?
May 18th, 2010OK, month after month after month we hear “the recovery is on it’s way; stocks look GRRRREat!; but oh that unemployment…” And month after month we hold our breath in hopes that unemployment is improving - even if it’s a little. It’s almost like the last missing link to recovery.
So I thought it would be interesting to see if we can use the voice of the common (wo)man to model and predict that good old turnaround in unemployment that we’ve been looking for. And there’s a fantastic source of information to do this - Gallup Poll interviews.
Below you’ll see some data taken from the Gallup Poll website, along with some extrapolations that I inserted. The original and updated polling dataset is at the Gallup website. According to the website, data are based off interviewing 1600 working adults and each data point in the chart represents the three day rolling average.
I used this Gallup Poll data, without the extrapolation, and was able to construct a model that generates an unemployment rate prediction (seasonally adjusted), approximately 3 weeks in advance of when the Bureau of Labor Statistics releases their number. The graph below shows you the match between the model and the actual data from the Bureau.
Although not an super tight match, it’s interesting that a reasonable correlation can be achieved with such a narrow data source - a single source of polling numbers. I am currently predicting the May unemployment number to be 9.7% based only on this model. For the curve fitting geeks out there, I will disclose that the model contains only 4 tuning parameters, and of course all the data points generated here are from a single set of parameters (no customization or adjustments by data point).
The model becomes interesting for doing “what if” scenarios based on the polling numbers. For example, I looked at the trend for the polling numbers over the last 5 months and performed a linear fit, as shown by the extrapolation in the first graph. Then, I mapped it to the model to see what it predicts. So, the conclusion, shown in the second graph, says that IF (big IF) the hiring and letting go trends continue as they have been for the last 5 months, unemployment should drop to very roughly 8.5% by November of this year.
I wanted to post that prediction here for fun, just to see how close I am on the upcoming May unemployment number, as well as the longer term extrapolated trend. I’ll be updating these original graphs with the original predictions, and overlaying the real data as it comes.
May 31st, 2011 at 7:55 am
I AGREE.