Anyone who deals in data professionally is sharply aware that any single figure measure is misleading on it's own. And if they live in the US they will more than likely be aware of the broad swath of government statistics that are made available ( for free ) to the public. As such it's quite easy to develop a contemptuous attitude towards lay people who aren't aware of things that are so foundational to people who work in the field.

That the "lie of the official unemployment statistics" keeps on getting rediscovered is a testament to the general lack of numeracy in the population at large; and to the construction of political reality by the media and the political class.

And in a political sense, the unemployment statistics are a lie; since we are supposed to be reassured that the situation is fine when in fact it's not. If the vast majority of people in their prime earning years ( 35-55 ) are drastically un- or under- employed what does that mean a decade out? Are we going to have millions of 60 year olds who haven't held a straight job in a decade? What's going to happen to them? Are they all going to survive by panhandling in the parking lot at Walmart? What?

Indeed,

In the former USSR and other totalitarian states, the government put out a false, rosy picture of the economy which was generally held in contempt.

In the US, the government puts out a "true after these many caveats" picture that most people are fairly doubtful of.

It seems like the present order is simply developing more defensible propaganda.

The government puts out a range of statistical measures and the media promptly ignores 95% of the data. In what way is that propaganda?

To be fair, it's not as though the government treats the wealth of data as equally important and the media just fixates on one measure. The government wants to tout that low unemployment rate as well, and they make sure to do so whenever it fits their purposes.

Still a far cry from saying they're lying to us, but they're not exactly innocent victims either.

Just to plug ALFRED, you can get the numbers as they were released in real time as well:

Unemployment: http://alfred.stlouisfed.org/series?seid=UNRATE

U6: http://alfred.stlouisfed.org/series?seid=U6RATE

One strong reason to prefer U3 is that the series is available much further back in time.

> One strong reason to prefer U3 is that the series is available much further back in time.

Isn't it a bit like saying one should prefer the bible to carbon dating because we've only had carbon dating for the last handful of decades?

It's not that U3 is useless, but one shouldn't assume it properly describes unemployment.

Well, I think it's more like saying we should prefer temperature data based on tree rings or ice cores (or, even, old mercury thermometers).

"Prefer" is probably the wrong word, so you certainly have a point. "Accept, with a note that the data isn't ideal" is much better. I'm OK with using less-than-ideal data when ideal data isn't available; but I'm disappointed with the number of reports that fail to mention any limitations of their. It's entirely acceptable to say "we don't have enough data to establish whether this claim is true."

No, it's not relevant how long we've had the method--what's relevant is how far back the data goes.

It's a bit like saying that having data on 1960 is useful.

Serious question - looking at the graph, it looks like the disparity between U3 and U6 jumped quite a bit in 2009 and is just starting to get back the historical difference today. Do you think it was valid to be concerned about the disparity increasing by so much in 2009? That part does strike me as a little worrying, though it is coming back in line now, which seems good.

The difference looks fairly consistent:

Febuary 1995: U6 is 183% of U3.

Febuary 2000: U6 is 175% of U3.

Febuary 2004: 173%

Febuary 2006: 155%

Febuary 2007: 182%

March 2008: 178%

Febuary 2009: 183%

March 2010: 172%

Febuary 2011: 177%

Febuary 2012: 180%

March 2013: 184%

Febuary 2014: 188%

There quite a bit of variance, but it seems to be seems to have been consistently in the 170-185%, with a few outliers.

Steep changes in the slope can introduce visual distortions when you compare the gap between two series. If you download and plot the difference, I think you'll see less of a change.

That's a nice chart.

I find the periodicity very interesting. We're almost 5 years from the latest unemployment peak. Historically, it looks like we're likely to start trending back towards higher unemployment again, any year now.