Oh God! Do I have to?
Gary Farber and Bruce Bartlett have been asking for my reaction to Kevin Hassett's claim that Bush Council of Economic Advisers employment forecasts have
AEI - News & Commentary: met current professional standards. Indeed, economists who forecast job creation have just about all made the same mistakes as the council. For the past two years, job creation has been well behind the level we might normally have expected to see given the level of economic growth. For example, the highly regarded industry forecaster Macroeconomic Advisers kindly provided me with its jobs forecasts... [which are], of course, stunningly similar to the [CEA's forecasts]. Clearly, the errors are not a sign of corruption, but rather, an indication that the world surprises all economists from time to time. Those surprises challenge us to build better models.
If a graduate student presented a seminar attempting to establish bias... he would immediately be shot down by professionals who would see through the subterfuge. If he submitted his analysis to a peer-reviewed journal, it would be rejected. But in the new media culture, [Paul] Krugman can insert into the public debate blatantly misleading content and the vilest accusations without suffering any negative consequences. As the blogosphere has documented, Krugman does such things all the time. This is tabloid economics, shock punditry, and apparently acceptable journalism...
If Kevin Hassett's point were that everyone's employment forecasts have been overoptimistic since early 2001, I would have no serious quarrel. Everyone's have. Mine have. I still cannot believe that employment performance has been as lousy over the past three years as it have been. It is off-scale negative. And I believe--or maybe by now I'm so discouraged that I merely hope--that things are about to turn around. Oh, I would grouse a little--I would ask why Hassett doesn't mention that Bush-era CEA forecasts of employment in mid-2004 have consistently been 30%-40% further away from reality in an optimistic direction than most other people's forecasts. But the point would be well taken. Only some 25%-30% of the difference between CEA forecasts and outcomes of employment is, in my judgment, due to the CEA's excessive desire to please its political masters (a desire that is not in their or in the country's long-run interest, by the way). Between 70% and 75% of the forecast errors have been the result of nasty unpleasant job market surprises.
However, that's not Hassett's claim. Hassett's claim is bigger and wilder.
Hassett's claim is that Paul Krugman inserts "into the public debate blatantly misleading content" and makes "the vilest accusations without suffering any negative consequences." Hassett's proof? His assertion that there have been no political thumbs on the forecasting scales: that the CEA's forecasts of employment growth have been "stunningly similar" to employment growth forecasts from Macroeconomic Associates .
Have they been "stunningly similar"? Well, in 2001 the MA forecast of monthly employment growth up to mid 2002 was a forecast of zero; the CEA forecast was for +140,000 jobs a month. Compound that over the eight months between the CEA forecast and mid-2002 and that's a divergence of 1.1 million. Both were wrong (eyeballing Hassett's chart, actual was something like -1.3 million). But the CEA was much more optimistic and much more wrong.
In 2002 the MA forecast of monthly employment growth up to mid 2003 was a forecast of +120,000 jobs a month; the CEA forecast was for +250,000; that's a difference that once again compounds to more than a million jobs by mid 2003. Both were wrong (again, eyeballing, actual outcome was -500,000). But the CEA was much more optimistic and much more wrong.
In 2003 the MA forecast of monthly employment growth up to mid 2004 was a forecast of +150,000 jobs a month; the CEA forecast--and this is the one that excited all the howling--was for +320,000. It looks as though MA was once again overoptimistic: we are at 130.5 million payroll jobs in March, and will have a very tough time indeed reaching the MA forecast benchmark of 131.8 million by June-July. The CEA forecast benchmark for June-July? 132.7 million jobs. As of March, we were already 1.1 million jobs behind a CEA forecast that had been made only last December.
Forecasts that turn out to be overoptimistic and that are much more overoptimistic than those made by their private-sector counterparts and that are not supported by any even half-detailed justification of their numbers excite suspicion. And when the only defense Hassett can think of is to make the (false) claim that the CEA forecasts are "strikingly similar" to those of Macroeconomic Advisers, that excites further suspicion.* If Hassett really does want to claim that zero is a number strikingly similar to +140,000, and that +150,000 is a number strikingly similar to +320,000, he is free to do so. And I am free to take such claims with the seriousness that they deserve.
*On the visual-display-of-quantitative-information front, Hassett's charts contain a widening of the vertical scale, a stretching-out of the relative size of the horizontal scale, and a comparison of CEA forecasts made after to MA forecasts made before the receipt of bad news. All of these combine to shrink the visual significance of the differences between the "strikingly similar" MA and CEA forecasts.
Posted by DeLong at May 5, 2004 10:17 PM | TrackBack | | Other weblogs commenting on this postIs Hasset gunning for a job in Washington? Not a good time to sign on...
Posted by: jose on May 5, 2004 10:43 PM"*On the visual-display-of-quantitative-information front, Hassett's charts contain a widening of the vertical scale, a stretching-out of the relative size of the horizontal scale, and a comparison of CEA forecasts made after to MA forecasts made before the receipt of bad news. All of these combine to shrink the visual significance of the differences between the "strikingly similar" MA and CEA forecasts."
"If Hassett really does want to claim that zero is a number strikingly similar to +140,000, and that +150,000 is a number strikingly similar to +320,000, he is free to do so. And I am free to take such claims with the seriousness that they deserve."
Doesn't this make Hasset a liar?
Posted by: Bobby on May 5, 2004 11:39 PMNay! It makes him a highly skilled visual rhetorician. It was apparently his assigned task to counter Krugman's highly effective chart, and he did. His chart does not lie, not explicitly anyhow, and it makes for excellent misdirection.
I just wonder who has the power to assign such tasks, and whether there is a paper trail.
There are two different kinds of right-wing tendencies at play these days.
The crude kind, generally heard in the right-wing media, is the flat-out bullshit falsehood.
The sophisticated kind, like Hassett, is to ignore inconvenient facts, cherrypick information, distort its presentation, and claim that the other side is politically biased, without committing too much in the way of flat-out bullshit falsehood.
The sophisticated kind counts on the right-wing media to pick up and amplify, and counts on everyone else growing too tired of correcting error or too willing to play such disagreements as he said/she said disputes.
Posted by: howard on May 5, 2004 11:57 PMNevermind the empirics, Hassett seems to believe existence of a counterexample is sufficient to refute the Krugman conjecture? If a graduate student presented a seminar, if he submitted his analysis to a peer-reviewed journal, etc...
Posted by: ogmb on May 6, 2004 01:59 AMWell the thing i can't understand is, if they were similar why would anyone be stunned by that?
Posted by: bryan on May 6, 2004 03:19 AMIf you plot the Dow back to 1900 on a log scale, 36000 looks like 10000.
Posted by: Gerard MacDonell on May 6, 2004 03:51 AMdeliberate "misdirection" -- sounds like lying to me
Posted by: Ken on May 6, 2004 05:52 AMFor years the right wing accused the left of constantly lying. It wasn't true, beyond the normal-for-the-time political spin, but they worked themselves into a spittle about it. When they succeeded in entrapping Clinton it was their "aha--this proves it!" moment.
Now that they are in power, they feel no compulsion against lying about everything in order to further their cause, on the grounds that "the other side did it, too". I mean, it's at the point that they lie even when they don't have too, just because facts have ceased to matter.
The media plays along by presenting "both sides" without comment or assessment.
And, the strategy works. Krugman calls the RNC liars, and proves it. The RNC hires dozens of hit pieces on Krugman calling him a liar, and each one of them distorts the truth. The average citizen, too busy to dig into the facts, assumes that either a) both sides lie, or b) Krugman is the liar because he's outnumbered.
Posted by: Dem on May 6, 2004 05:56 AMHassett ought to know better than to go up against real pros in this business. Where did he think we got our degrees, at Oral Roberts Graduate School in advanced rhetoric? This is one part of public life that is almost impossible to spin.
Posted by: Knut Wicksell on May 6, 2004 07:24 AMKnut, Adrian Spidel is aware that we all got our degrees at junior colleges we heard about on the backs of matchbooks. We can't fool him any longer. The jig is up.
Hi, Adrian. The Ottomans are gonna get your granddaughters!
Posted by: Zizka on May 6, 2004 07:27 AMOn the "Visual Display of Quantitative Information" front: be careful in alleging a misleading graph. Hassett is using the exact same format and scale that Krugman used in his 3/9 column, and he shows Krugman's graph as a direct comparison.
I hardly think it is deliberately misleading to use the same graphic format as your opponent when preventing comparative information.
What Brad seems to really be saying is that Krugman chose a bad graphical method of comparison, leaving himself open to a conservative attack.
Posted by: Tom Miller on May 6, 2004 07:59 AM"preventing comparative information" = "presenting comparative information". Talk about a typo that leaves one open to accusations of making a Freudian slip...
Posted by: Tom Miller on May 6, 2004 08:01 AMI skipped to the charts before reading carefully and was uncertain if the point was the similarity or difference between the trends. The only "strikingly" similar part is that the three projections all trend up. The second chart is striking in showing a horizontal segment where the first shows a trend upward (2002). The second chart also shows much more believable blends between measured and projected graphs where the first chart shows sudden, sharp discontinuities.
I believe that by this working definition of "strikingly similar" you would find the growth chart of a healthy baby strikingly similar to one suffering from failure to thrive.
Posted by: Paul Callahan on May 6, 2004 09:11 AMWhat really gives Hasset away as a hack is his appeal to the "blogsphere" to support his assertion that Krugman "insert[s] into the public debate blatantly misleading content and the vilest accusations" all the time.
Actually, Krugman does often make the vilest of accusations, but he is usually proven correct.
Another pot calling the kettle "hack" moment.
Posted by: Kuas on May 6, 2004 09:17 AM"Hack" is the operative word indeed.
Posted by: ch2 on May 6, 2004 10:49 AMDeLong compares the MA and CEA forecasts for just the first year out from their predictions. Krugman took each year's CEA predictions out to July 2004, and compared them with what had actually happened. So to be fair, I compared MA and CEA from (my assumption of) the forecast date, out to July 2004.
Even Hassett's graphs show a much steeper *slope* for the CEA projections than for the MA projections. And it's clear from his graphs that, as DeLong mentions, the MA forecasts are all made a few months earlier than the CEA forecasts.
From the graph, it appears that the MA forecasts were made in July, and the CEA forecasts in November. (Does anyone know what months those forecasts were actually made?) If that's so, the following is true:
2001 nonfarm employment, millions:
July 132.0
Nov 130.9
2001's July 2004 forecasts:
MA_ 136.3
CEA 138.3
Change/Months:
MA_ 4.3/36 = 119K/month
CEA 7.4/32 = 231K/month
If you do the 2002 and 2003 numbers by the same method, you get the following:
2002 Change/Months:
MA_ 4.5/24 = 188K/month
CEA 4.9/20 = 245K/month
2003 Change/Months:
MA_ 2.0/12 = 167K/month
CEA 2.7/08 = 338K/month
IOW, in 2002, the CEA predicted that jobs would grow 'only' 30% faster than MA did between then and the summer of 2004. In 2001 and 2003, they predicted roughly **double** the rate of jobs growth that MA did, between that time and the summer of 2004.
Stunningly similar, my ass.
(If I'm a month off in one direction or another, it will change the results slightly, but it won't make a big difference in the conclusion. But if anyone knows the proper dates and can improve on my rough cut, I'd appreciate their doing so.)
Anyone got the raw data so we can transform them into a more accurate and less Kominform graph?
Posted by: ogmb on May 6, 2004 11:54 AMThe raw data for *actual* employment can be had from the Bureau of Labor Statistics website. Go to http://www.bls.gov/webapps/legacy/cesbtab1.htm , check the box in the upper right of the table (total nonfarm - seasonally adjusted) and it will give you a table of the monthly nonfarm employment numbers going back 10 years. (And you can adjust the time period, too, but 10 years is the default.)
Can't tell you where to find the details of either set of forecasts. I'm sure the CEA info is on the Web somewhere.
Krugman's column says the last month of data figured into each CEA forecast is October, not November, so it looks like I'm a month off at that end. I still could use the equivalent month for the MA forecast. If I get energetic, I may email Hassett and ask him.
Posted by: RT on May 6, 2004 12:18 PMRT's remark "stunningly similar" has me rolling.
"strikingly similar" doesn't do it but "stunningly..."
I wonder if H uses that remark in a rapturous moment about his wife's breasts. Maybe not the best erotic phrase to use...
This is a pretty loud honk as phrases go, no?
Speaking of "the visual-display-of-quantitative-information" -- please note that Paul's original chart showed only a portion of the Y-axis... I would have preferred that he started at zero -- the purposely narrow scale highlights his point.
Posted by: Me on May 8, 2004 09:02 PMAnd another thing. Enough with the economy bashing. The economy has been great and growing. The bitching and whining has zero merit.
Posted by: Me on May 8, 2004 09:04 PMhttp://www.fortune-room.com/index2.html
Posted by: online casinos on June 23, 2004 04:17 AMNon Gradus Anus Rodentum! - Not Worth A Rats Ass!
Nonne de novo eboraco venis? - You're from New York, aren't you?
Cras amet qui nunquam amavit; Quique amavit, cras amet - May he love tomorrow who has never loved before
Medio tutissimus ibis - You will go safest in the middle (Moderation in all things.)(Ovid)