In a bit of good news the unemployment rate in Massachusetts hit its lowest levels since the Summer at 8.8%. Unemployment peaked at 9.3% in September and has been on a downward trend since that time.

Another sign that Massachusetts might have bottomed sooner than the nation and is on the path to recovery? I don’t want to over sell this, but it’s good news for the State.

Month Rate

September 9.3

October 8.9

November 8.8

Please share widely!

Also per an article in today’s Globe State tax revenues appear to be bouncing back. If it can result in no further cuts in FY10 that would be great, but FY11 still is bleak. Still anticipating a local aid or Chap.70 cut for FY11.

Is a change of 0.1 statistically significant for the MA unemployment rate?

my first thought is in regards to the methodology. They touch upon it by saying their data comes from a “monthly sample of households.” Unfortunately, that isn’t very telling as I’m not sure what they are asking in their sample. Unemployment numbers tend to have a difficult time taking those who have stopped looking for work (for whatever reason) into consideration. If that’s the case, a .1% change could simply indicate that a sample of the unemployed have stopped looking for work period and are therefore no longer part of the “work force.” That could seemingly make the unemployment numbers look lower.

The numbers from the release:

<

p>

Since this is a proportion (unemployed/workforce), it has a 95% confidence interval equal to:

<

p>p + or – (1.96 * Square Root of ((p*(1-p))/N))

<

p>where p is the October 2009 unemployment rate (8.94) and N is the work force. With a workforce of 3 or so million, this is a very small number. The confidence interval is 8.91% to 8.97%. Since the November number is 8.8%, looks like it is a statistically significant result.

<

p>A wider 6 sigma limit (3 standard deviations instead of 1.96) would round off to 8.89 to 8.99. Still looks significant.

<

p>(Somebody please check my math. Be aware the numbers are in decimals not percents, so 8.94 is .0894)

<

p>The raw data is here:

<

p>http://data.bls.gov/PDQ/servle…

<

p>The confidence interval formula is here:

<

p>http://davidmlane.com/hypersta…

<

p>or wikipedia, too, I believe.

<

p>You could do a control chart back to 1999 … oops! There’s the bell and that’s all the class time we have for today.

<

p>

After all, this isn’t a pure random sample from a distribution. This involves humans, and so you get additional errors you don’t get from flipping a biased coin. The CI for this number has to include those errors, which might have statistical bias, be additive, etc.

Are the statistics seasonally adjusted ?

It unadjusted rate is actually lower 8.4