Identify the bias

[ATFFL]

Just wondering how long it will take people to identify the very subtle bias in this graph.

[Bono]

Just wondering how long it will take people to identify the very subtle bias in this graph.

Bush Clinton Bush?
Also, the Y axis are practically no variation at all.

[Gabu]

Just wondering how long it will take people to identify the very subtle bias in this graph.

The fact that it doesn't point out that employment rates went up under Reagan?

[A18]

You mean the fact that it dropped an entire two points? Yes, it's pretty impressive, and definitely shows Bush is destroying America.

[A18]

Still waiting for someone to congratulate me on getting it right.

[ATFFL]

Bono: You are exceedingly close.  Look a bit closer.

Gabu:  Sort of, but not really.  They did leave the data in.  They could have made it a 15-year graph and cut Reagan out entirely.

A18:  Did it really drop 2%?  Look a bit closer.

Need to run for a bit.  More clues when I get back if no one has it.

[A18]

It dropped 2.2% from 2000 to 2004.

[ATFFL]

It dropped 2.2% from 2000 to 2004.

That it did.  Think about that a bit more.  Something is quite right about that.

I got to run, starting a summer class.  If no one has it when I get back, I will point it out.

[phk]

I think the length of the curve from 2000 to 2004 is the bias.

It overstates the negative movement,  essetially makes it feel like a "20%" (22% as Philip pointed out) drop instead of a "2%" drop, because they start from 80 and go to 90.

Because of it only covering Y-Values from 80-to-90 it overstates a certain movement by a factor of 10.

It would be more accurate if the Y-Values went from 0 to 90 instead of just 80-90.

[Brandon H]

I caught it by looking at it before actually reading phknrocket1k's post. Yes, I agree. (I learned about that in my statistics class in college.)

[ATFFL]

That is a big part of it, but look closer at the years.

We use 1989, the year HW Bush was inaugurated.

We use 1993, the year Clinton was inaugurated.

Then we use 2000, the year W. was elected, not the year he was inaugurated.   

So not only do we have the effect exaggerated, but we make it look like the decline occurred entirely under W.

[Jake]

It dropped more than 1% from 2000 to 2001.

[Citizen James]

Well, looking at the bureau of labor statistics you can get a good glimpse of the data (with or without graphs) of historical trends.    (the little dinosaur icon links to the historical stats pages for diferent indicators.  It will default to 10 years, but you can go all the way back to 1939 if you want.  You just have to remember there were fewer people back then, and because of population growth, there needs to be a certain rate of job growth for unemployment to remain steady.

2000 wasn't a spectacular year, and posted net job losses in three different months, the worst of which was June with a loss of 43 thousand jobs.  That was the 2nd worst loss of the Clinton adminstration (first was March of 93 - 46 million jobs).  Of the seven months Febuary-August of 2001, six of them posted net losses.  The only gains were in March - a Respectable though not spectacular 92 thousand jobs.  The worst of these was April (-297k), followed by June(-158k) and August(-138k).  Of course, after 9/11 things went deep south for several months.  Though not a true comparison, April-May-June of 1995 is the closest we can get for under Clinton (the Oklahoma city bombing +163k,-9k, +227k).   From late 2003 to present there have been decent, though not spectacular, gains.

one unrelated thing I do wonder - what the heck happened in August of 1983?  We went from -308k to +1.114M.

[Alcon]

1993 does not look like 87.2%, but maybe that's just my eyes.

[A18]

That is a big part of it, but look closer at the years.

We use 1989, the year HW Bush was inaugurated.

We use 1993, the year Clinton was inaugurated.

Then we use 2000, the year W. was elected, not the year he was inaugurated.   

So not only do we have the effect exaggerated, but we make it look like the decline occurred entirely under W.

It's a two point decline! Who cares anyway?

[Gabu]

1993 does not look like 87.2%, but maybe that's just my eyes.

I had that too.  I put a ruler up to it and it is, indeed, just over the 87 mark, however.

[○∙◄☻¥tπ[╪AV┼cVê└]

It makes the Bushes look worse if you choose the key years to be 1988, 1992, and 2000.

[Beet]

The 1989 and 2000 arrows are pointing down, whereas the 1993 arrow is pointing up. A devious little visual trick kind of like when you look at a straight line running through a curved line, you swear the straight line is curved as well.

[angus]

That is a big part of it, but look closer at the years.

We use 1989, the year HW Bush was inaugurated.

We use 1993, the year Clinton was inaugurated.

Then we use 2000, the year W. was elected, not the year he was inaugurated.   

So not only do we have the effect exaggerated, but we make it look like the decline occurred entirely under W.

that's not a bias.  it's just bad labelling.  a bias would be counting differently, or comparing apples to oranges.  For example, Vorlon has defined "bias" with respect to polling data several times, in terms of sampling and the nature of a question being asked.  If we extrapolate this understanding to unemployment data presentation, we can say that a bias would be something like defining differently "unemployment rate" differently in different years.  And that's actually done sometimes.  But bias is simply skew or prediliction or prejudice, such as an inclination of temperament or outlook, or a peculiarity in the shape of a bowl that causes it to swerve when rolled on the green, or a voltage applied to an electrode to establish a reference level for operation.  I'd say that the phenomenon to which you refer is more correctly identified as "propaganda" or "poor presentation" rather than "bias."  Which term I'd choose would depend on whether I thought there was an intention to mislead.  Giving the presenter the benefit of the doubt, I'll go with poor presentation rather than spin.  It should have been caught by an editor, but in this "everyone's a publisher" age, such mistakes go through with regularity.  I blame the schools.

Also, I couldn't identify it till you pointed it out.  very clever. 

[??????????]

It makes the Bushes look worse if you choose the key years to be 1988, 1992, and 2000.

It certainly looks like some of the trash statistics you post. Your "stats" were nailed like a 2 dollar whore in that thread.

[ATFFL]

angus:  Which word would be best would depend mainly on whatever article was built around this graph.  I also thought it carried the least baggage with it.  If I had used "poor presentation" I think no one would look past the way the small range exaggerated the changes.  If I used "propaganda" I was afraid people might have spent too much time reading too much into the graph.

[angus]

sorry.  I get uptight and Victorian about some things.  Mostly improper use of language and general inaccuracy.  As I said, I didn't even notice the faulty label till you pointed it out.  good call.  I think you're exactly right about trying to use the least emotionally charged term unless you know the context.  My inclination is to give the presenter the benefit of the doubt.  No kidding, with the proliferation of blogs and on-line publishing and the 24 hour news cycle, honest mistakes are made more than ever before.  One constant complaint I hear from my colleagues is the difficulty teaching students to distinguish between authoritative sources and general bullsh**t, and this doesn't just apply to history and politics.  It's a fairly widespread problem in all fields.