A quiz

by Russ Roberts on February 8, 2009

in Data

Find the error in the following picture (HT: Andrew Sullivan). It is a very frightening picture but it is also frightfully misleading. Do you see the error?


Sullivan sees the error, too, but don't click through till you've taken a shot at it.

None of this (or this) is to suggest that everything is hunky-dory. It's just that accuracy and the facts  matter.

On Monday, around mid-day, I'll post what I think is the right measure of how bad things are. It's not as bad as the above, but it is not a pretty picture.

Be Sociable, Share!



48 comments    Share Share    Print    Email


muirgeo February 8, 2009 at 4:53 pm

Maybe that it is not corrected for in terms of population growth.

vikingvista February 8, 2009 at 5:16 pm

Since when is absolute number of job losses more relevant than jobless rate?

Starting from a 3% unemployment rate would be quite different than starting from a 6% unemployment rate.

I can't even tell if this chart refers to gross or net job losses. If twice as many jobs were created as lost, then the economy would be considered booming.

And why is the "current recession" producing such a smooth plot while the past ones had more noise in them. This tells me the current numbers are cruder estimates and likely computed differently.

Finally, and most important, is the stamp "Office of the Speaker". After all these years, Pelosi's reputation precedes her, and that stamp automatically makes the plot highly suspect.

adina February 8, 2009 at 5:16 pm

Hmm…Are comparisons to "peak month" useful measurements? If all months were equally lousy, then the slope wouldn't be so steep for the green. Unless I'm misinterpreting, it seems that all it's saying is that we had at least one really great month.

Zachary Kurtz February 8, 2009 at 5:31 pm

maybe it's just me, but where does Sullivan find the error?

Mark February 8, 2009 at 5:36 pm

Yes. I do see the error.

It was published by Nancy Pelosi's office.

Sanjiv February 8, 2009 at 5:39 pm

"Relative to peak month" sounds dubious to me.
A lesser value for the peak month will make the above curve look less frightening.

Mezzanine February 8, 2009 at 5:54 pm

How the hell do we know what the real figures are anyways? There is way too much noise in the form of government interference.

Ike February 8, 2009 at 5:57 pm

First, it would appear as though all three of those were starting from the same Peak, when in fact they don't.

Second, the green line might be coming from a point so high, that it would still track above the other lines on an absolute scale.

Ironman February 8, 2009 at 6:09 pm

Ooo – can I win by pointing to Bill Polley's post on the topic of job losses (and his much superior and more detailed chart?)

Ike February 8, 2009 at 6:17 pm

WHOA. Let me add one more observation.

I know that I'm one hell of a lot smarter about economics and global trade than I was 10 and 19 years ago. After all, there is a vibrant and active internet helping share information faster and with more ubiquity than there was then.

Also – the analytical tools available for companies to forecast demand are more sophisticated, as is the adherence to the bottom line.

The real kicker here? So much of commerce and industrial production has shifted to sophisticated inventory tracking and just-in-time delivery. Where we used to have wasteful stacks of things in warehouses acting as a buffer, we now have very thin margins greasing the wheels of commerce. There is little slack in the system.

With far less slack in the current economy, businesses will get the signals to shed jobs sooner, and will act on those signals.

That's my theory, and I'm sticking to it.

MWG February 8, 2009 at 7:47 pm

"Yes. I do see the error.

It was published by Nancy Pelosi's office."

How do you spell 2 drum beats and a cymbal crash?

MnM February 8, 2009 at 8:49 pm

Simple. It shows things as aggregates. There are more job losses in this recession because there are more jobs (that is, the labor force is bigger). They should have used the unemployment rate and it would have given a more accurate picture.

MnM February 8, 2009 at 9:06 pm

I'm worried I didn't make myself clear enough. Let me expand.

For the sake of simplicity, I'll use simple numbers in factors of ten.

Recession A: Labor Force: 100 persons. 10 jobs lost.

Recession B: Labor Force: 1,000 persons. 100 jobs lost.

Recession C: Labor Force: 10,000 persons. 1,000 jobs lost.

The the number of jobs lost increases with each recession, but the percent lost is the same. The percentage is the more telling of the two.

Lee Kelly February 8, 2009 at 10:02 pm

Green is, like, so last year.

James February 8, 2009 at 10:11 pm

I can highly recommend John Huizinga's notes – he has the right figures on page 3:

Peak to Trough Job Loss in the current recession: -1.9%

Peak to Trough Job Loss in the 1957/58 recession: -4.4%

Peak to Trough Job Loss in the 1974/75 recession: -2.8%

Peak to Trough Job Loss in the 1981/82 recession: -3.1%


Sam Grove February 8, 2009 at 11:01 pm

But James, that's not scary enough to keep the people in line with policy goals.

Jacob Oost February 9, 2009 at 2:57 am

MWG, I prefer OO*

cpurick February 9, 2009 at 7:47 am

Maybe Congress should publish a new chart, called "The Number of People with Jobs." Then they could all just declare victory and go home.

John Dewey February 9, 2009 at 9:29 am

The obvious error is in comnparing absolute job losses for periods when absolute employment varies by tens of thousands.

Civilian labor force


The second error, as I think someone above noted, is that peak levels, even when measured in percentage employed, may vary greatly. Percentage job losses from higher than normal employment levels, such as 1999-2000, would likely exceed those from periods of normal employment levels, such as 1989-1990.

vidyohs February 9, 2009 at 9:31 am

The only thing I can see in it that may be an error, and in my mind would be, is that if we take the time line and work it backwards it would mean that, according to the publishers of the chart, the current recession began approximately Oct 2007.

Not too sure that's the case. It seems to me that in Oct 2007 the bubble was still bubbling, credit was easy, and unemployment was very low.

Hammer February 9, 2009 at 9:51 am

MWG: The word you are looking for is "rimshot".

Just don't confuse "shot" with "job". Turns out that will get you banned from some boards…

LowcountryJoe February 9, 2009 at 10:55 am

Just don't confuse "shot" with "job". Turns out that will get you banned from some boards…

What's wrong with rimjob? Come to think of it, what's wrong with wanting to 'mount' some tires?

ChrisF February 9, 2009 at 11:33 am

(1) Uses absolute, not relative numbers
(2) Ignores fact that in 2007, unemployment was at historic lows, esp. compared to where the 1990 and 2001 recessions were.
(3) Selectively chooses two relatively mild recessions as comparisons.

kazoolist February 10, 2009 at 12:07 pm

Related: "Lies, Damn Lies and Nancy Peolsi’s Charts" @ http://www.qando.net/?p=529

Previous post:

Next post: