Readers of the Big Picture know what a huge fan I am of both the WSJ and their awesome graphics department. So it was with great disappointment that I reviewed this graphic sent in by an emailer:
I cannot recall when any WSJ chart previously made a move of just 3% — a move from 132.5m down to 130m up to 134m — look so good . . .
Let’s explore how else this data can be presented. Its important to investors to understand what they are looking ast, and what it truly means for the economy and the markets.
Here’s a few other ways to depict this data — same numbers, but the entire data series shown — not just the changes:
Quarterly, with a zero basis
Looks less impressive, no?
But that chart is a worthless to the investor as the WSJ chart above is. Why? It provides no particular insight into what is happening with the economy. All I did was plug the data into Excel, and set the base to nothing; Without the zero 1st month, the charts look about identical — and equally worthless.
So you can either exaggerate change by eliminating all but recent delta, or diminish the change by including all of the data. Both are worthless depictions.
If you really want the data to give you insight, you need to compare it to prior post recession periods. Fortunately, the Cleveland Fed does the heavy lifting for that for us:
Nonfarm Payrolls, Post Recession: 2001-05 vs. Average Recovery
Source: Federal Reserve Bank of Cleveland, (Caveat Forecaster, February 2005)
That chart dates from February; As of recently, the two lines are moving in parallel — much better than the chart depicts, but still far far below prior comparable periods . . . (If anyone comes across an updated version of this, please let me know and I’ll post it here also.
>
UPDATE: AUGUST 9 2005 6:40am
Of course that chart is misleading — its from the Op-Ed page. DUH! Just goes to show you that a picture may be worth a thousand words, but if most of them are crap, so’s the picture . . .
Nevermind.
>
Source:
REVIEW & OUTLOOK
The Great American Jobs Machine
August 8, 2005; Page A10
http://online.wsj.com/article/0,,SB112345310644607010,00.html
I don’t think the chart is so misleading, and apparently neither does the FOMC. The decline in employment, though not dramatic by historical standards, was enough to convince them to bring the federal funds rate down to its lowest level in four decades. The increase in employment, though meager relative to past recoveries, was enough to convince them to raise the rate by a factor of 3. (Okay, I’m misleading a little by citing the proportional increase, but 225 basis points is still a lot.)
Normally I have strongly opposed tax cuts during times of deficit spending, and I am certainly no fan of President Bush, but in 2003, with the Fed running out of ammunition, a tax cut was a sensible thing to do (maybe not the best available option, but that’s another argument). While I would always be cautious about attributing economic results to a single cause, I would say the data support the conclusion that the tax cut worked.
This same trick was used, in reverse, on the front page of the WSJ about 12 years ago, to make a rather different point. At that time, Clinton was talking about hiking taxes, and the WSJ wrote that tax hikes couldn’t, simply couldn’t raise revenues, because revenues were a constant share of GDP. Always. Just look at the chart.
The chart ran from 0% to 100%, with variations in revenue as a share of GDP varying by roughly 4% (if memory serves) during the post-WWII period represented to that date. The range (big) and scale (small) of the chart made it look as though revenues as a share of GDP really didn’t change. To repeat, this was on the front page, rather than the editorial page. How it got there I do not know.
The key seems to be the magnitudes involved. Should a drop of 1.5% or so followed a rise of about 2.5% be graphically represented by such steep lines?
I should note, though (see my previous comment), that I do disagree with most of what the editorial says, including the implied mechanism for the effectiveness of the tax cuts. (The editorial mentions “marginal-rate tax cuts”, but it is implausible that the implied supply-side effect would happen so quickly. Moreover the recovery appears to be driven by consumption rather than investment. Lower marginal rates on investment income should discourage consumption, except for the fact that they went along with lower average rates which put more money in people’s pockets.) The chart effectively makes the point that the recovery was related to the tax cut, but support for its other points is rather weak.
For example, “the percentage of ‘long-term unemployed’ workers is about two percentage points lower than it was in the same stage of the Clinton expansion.” This is almost true, but very misleading. For one thing, two percentage points are not very much when you are talking about “percent of the unemployed”. Moreover, it depends critically on how you define “long-term unemployed”. If you define it as “15 weeks & over” rather than “27 weeks & over”, then the proportion is several “percentage points” higher than it was at the same stage in the Clinton expansion. But most important, this whole measure is a dubious indicator this late in the game. They are trying to argue that the supposedly low long-term unemployment means that workers haven’t been discouraged, but by definition those discouraged workers have already left the labor force and won’t be counted in the duration statistic. What matters is how many long-term unemployed there were a few years ago, not how many there are today.
I have rarely found a Single variable that is capable of predicting anything as compelx as the economy, or for that matter, the Markets.
See this:
Single vs. Multiple Variable Analysis in Market Forecasts
http://bigpicture.typepad.com/comments/2005/05/single_vs_multi.html
but the big tell is thee prior post war post-recession recoveries: By every measure, the present recovery is sub par
royce: I think the steep lines are quite appropriate. The point is that the increase has been steeper than the decline. If you thought the decline was something to be concerned about (and as I remember, most people did), then you should be impressed with the recovery.
Barry: If the recovery is sub par, that is not because the tax cut didn’t work but because the Fed has withdrawn its support, because it believes (wrongly, in my opinion, but that’s another story) that the recovery is strong enough. The chart accurately depicts the fact that something changed dramatically in mid-2003.
As we all learned in Stat 101, correlation is not causation; coincidence in time provides insufficient evidence a relationship exists, either empirically or logically. Given the time delay’s typical in fiscal policy, the relatively close proximity of the two events makes the relationship between the tax cuts and employment even less persuasive although one could possibly argue that the cuts convinced businesses to increase hiring – I’m not aware they actually did, of full-time personnel at least, but one could argue it I suppose.
As always data representation depends upon the question being asked but context is important and the chart from the Cleveland Fed – a definite but historically anemic rise in employment presented at appropriate scale – is clearly more generally informative (or less misleading if one wishes) than the chart from the WSJ – it carries its context with it. But now that we know the context of the WSJ chart, we know it was developed to reinforce an argument and there is certainly no crime in that (even if the argument is unpersuasive); it was the loss of context that created a problem.
When I look at the WSJ chart, the time delay looks just right for a Keynesian response to fiscal policy (not for a supply-side response!). For several months, there is no apparent response, and the subsequent response begins slowly, accelerates, and then slows down again, just as it should. And the conclusion of causation is reasonable in the absence of other convincing explanations.
The Cleveland Fed chart is misleading in its own way (at least as a counterpoint to the WSJ chart) because it dates from the peak rather than the trough. The recovery appears much more anemic than it really is, because the decline was so slow, and the eye is fooled into seeing the decline as part of the recovery. It is also misleading because it measures relative to the original level of employment rather than the growth of the working-age population. A “typical” employment recovery today, now that the baby boom and its aftershock have been fully absorbed, would give the Fed a heart attack.
For a much better chart of the Bush Economic Trough, check out this chart of US “Employment Rate”:
http://research.stlouisfed.org/fred2/series/EMRATIO/11/10yrs
Barry warned against single-variable analysis, but to no avail. It is certainly not merely a Keynesian lift that has induced a recovery in hiring. Low long rates have little or nothing to do with fiscal policy, and have clearly been hard at work driving real-estate based consumption. The Fed has not withdrawn its support, merely moderated it. With the lags involved in monetary policy, even if neutral has now been reached, we should be enjoying stimulus from prior easy policy. There is even some reason to think that slowdowns are self-correcting. Policy efforts often tend to have their impact at just the wrong time. If we look at the US savings rate, trade and current account balances, and the CBO’s judgement that 2004 and 2005 are roughly neutral in cyclical terms, there is reason to think fiscal stimulus which was lucky in its early days (Bush did not anticipate the recession), is now inappropriate.
http://newmarksdoor.typepad.com/mainblog/2005/08/in_one_of_my_co.html
In one of my courses I try to teach the students how to avoid constructing misleading graphs. Barry Ritholtz properly criticizes a poor Wall Street Journal graph and provides me with a nice example for my course.