Toward a taxonomy of lying

So much nekkid dishonesty, so little time! Anyway, sorry about the extended delays in posting,* but trying to summarize all the random venality of recent times would just get us further off track, so let's get to some new material from today.

If you amuse easily, you might have gotten word of the bizarre Twitter message from Jack Welch (the former CEO of GE) today, basically proclaiming that the monthly unemployment numbers were fabricated by -- how did he put it, Interwebs? -- "these Chicago guys" to distract attention from the Kenyan Muslim socialist's shortcomings on the debate stage.** Calmer heads soon prevailed (tut-tut, fellows, let's not go overboard here), but they prevailed in a way that allowed the conspiracy theory to remain on the table. Here's Kevin Hassett, director of economic policy studies for the American Enterprise Institute, describing the difference between the stats reported every month by the Bureau of Labor Statistics:

The report, of course, reveals the results of two surveys, one of households, one of establishments. The professional economists and the press usually emphasize the establishment survey because it is viewed as less volatile. The establishment survey was terrible. ... The household survey, on the other hand, portrays a September that was booming.

A little hyperbolic on both counts (when numbers need that much coaching, you should generally try to get them away from their handlers and see if the stories still match up), but not out of bounds.  But he still has a conspiracy to peddle:

Back when President Bush presided over a jobless recovery, the household survey tended to show better news. At the time, every media organization carefully emphasized the establishment numbers, and warned that the household numbers are suspect. That, of course, is what happens when a Republican is in office. For President Obama, you can expect a household survey lovefest.

Now we've left invisible-pink-elephant territory for the land of testable fact claims: The in-the-tank librul media ("every media organization") report labor statistics differently based on the incumbent president's party affiliation, and they practice this distortion in a way that systematically favors Democrats. That sort of accusation ought to be taken seriously. After all, he's accusing an entire employment sector of peeing on its own professional codes for partisan reasons, and he's doing so based on what should be an easily visible pattern in the evidence.

As content analyses go, it isn't the biggest challenge in the history of the free world. We need to define and plasubly sample the "media," determine which of two numbers released on the first Friday of every month is emphasized, and figure out who was in office or running for office at the time. It'd be important to look at some confounds as well: whether unemployment has been generally rising or falling, for example, or whether the nation is involved in the sort of interstate or substate conflicts that might skew the overall picture of "news" that unemployment fits into. It's usually a good idea to start out with a few ranging shots -- say, how the AP (as a proxy for "every media organization") reported the same event last month:

A dismal new snapshot of jobs in America shadowed the presidential campaign on Friday, testing the voter patience that will save or sink President Barack Obama's re-election bid. Seizing on the timing, Republican Mitt Romney said Obama's convention party had given way to quite a "hangover."

Employers added just 96,000 jobs in August, not nearly enough to seriously dent unemployment, let alone inspire confidence that the economy is getting better. Even the good news the unemployment rate dropped from 8.3 percent to 8.1 percent resulted from many job-hunters just giving up.

That's a pretty simple coding call.*** We haven't decided how to code adjectives ("dismal") or the ordering of political comments and inferences (Romney speaks first), but the number that's emphasized is the establishment number. The household number == in which the improvement is nearly as large as this month's (0.2 vs. 0.3 points) -- hardly looks like a "lovefest." An AP analysis the same day waited for the sixth graf before mentioning the numbers; the household report came first, but the scare quotes around "improvement" belong to the AP writer. Our media conspiracy, in short, is a crock.

One conterexample is enough to knock a hole in fatuous claims about what "every media organization carefully emphasized," but that isn't why we do content analysis. We're looking for patterns that actually suggest something systematic about how people give and get the information that makes a difference in how they arrange their lives. So it's worth asking how the September data looked in, oh, 2008:

Businesses slashed jobs and the nation's unemployment rate hit a five-year high in August, the government reported yesterday, dashing hopes that the economy might stabilize in the second half of the year and showing that trouble has spread far beyond the housing and financial sectors.

... The unemployment rate rose to 6.1 percent, from 5.7 percent in July, according to the data released yesterday, making for the most severe four-month rise in joblessness since 1981.

Harder call, isn't it? The establishment number is first in the lede, but the household number is right behind, and it dominates the third graf.**** I'd code this one "household," except that Bush was president, and the numbers were getting worse rather than better, which might suggest we need a few new hypotheses. Or a look at September 2004:

Employers stepped up hiring in August, expanding payrolls by 144,000 and lowering the unemployment rate marginally to 5.4 percent. While the figures didn't amount to a national job fair, analysts said, they did hold out the promise of stronger growth following the summer lull.

I'd score that as an emphasis on the establishment number, as our expert predicted.  But that's one of the fun things about content analysis: The stuff that's easy to code isn't nearly as interesting as the stuff that's hard to code. Here, we need a set of rules for whether the numbers, taken as a whole, make the president's case look good or bad. I'd call these a "good." That isn't in the initial coding scheme, but it does support a really boring hypothesis: Sometimes, periodic developments in macroeconomic statistics look like good news, and sometimes they don't.

We're left, then, with a sort of scale of dishonesty. Fox phrased the whole thing as a question: "Is the number real?" That's not, in itself, a lie. Lots of news organizations use stupid questions for headlines, and unless they're as abnormally distributed as they appear to be at Fox, they're not a consistent indicator of systematic lying.

How about Jack Welch? Different question. The comment could indicate brain-dead stupidity, moral incompetence, or simple hyperbole. If you want to be sarcastic, it probably helps to be funny, but a bungled attempt to do what the Kool Kids are doing might just mean you're clueless rather than dishonest. No doubt Mr. Welch is laughing all the way to the bank.

Kevin Hassett, on the other hand -- he's just lying. He's taking stuff that's fairly easy to check and, in the guise of an economics expert, flat-out making up his own data. He lies about how people do their jobs, and he lies about what information looks like. In the good old days, journalism had a solution for people like that: Tell them their poetry smells and kick them down the stairs. Our little friends at the National Review -- and anyone else who takes his sort of claim at face value -- would do well to review that principle.

Clearly there's more to discuss here, so do come back and visit. We'll try to start doing better with the backlog.


* Particularly to those of you who have been sending in neat stuff. I hope you will keep it coming.
** "Chicago" is how you say "scary colored people" if you want to imply organized crime and electoral fraud, rather than random dystopian thuggery, in which case you say "Detroit" even if you mean "Bloomfield Hills."
*** We'll be doing intercoder agreement in a couple weeks. You should bring your own calculator if you decide to crash the party, but it can be a pretty cheap one.
**** The second is elided here; if you're that much in need of AP prose, go find it yourself