It is not often that I find myself interested in domestic politics, but this morning the Obama administration released the following graphic related to job loss data and it caught my eye.
This appears to be a prime example of the danger in politicized data visualization, which Alex Lundry points out beautifully in his Chart Wars talk given at Ignite DC. I think there are several things wrong with this chart, and am a bit disappointed in the Obama administration for touting this as evidence of their accomplishments.
I am a bit pressed for time this morning, so before I comment I would like to ask: what do you think is wrong with this chart?
Add your thoughts to the comments, or ping me on Twitter. I will update this post later in the day with my thoughts, as well as any of yours.
UPDATE: Thanks to everyone who added their thoughts to the comments. I echo all of the sentiments there, especially with regard to the use of color. I; however, am particularly disappointed with this graphic for its implicit message: the Obama administration is responsible for change in job losses. If we were to strip the colors and administration headings we would clearly see the result of a major exogenous shock—the credit crisis of 2008.
The curve of the data would seem to indicate a clear ARIMA process; whereby, the shock occurs and has diminishing effect as time goes forward. Likewise, if we were truly interested in testing the affect of the Obama’s stimulus we would have to specify a further autoregressive model based on that additional shock. Given the impulse-response nature of the system, a vector autoregressive model may be the bust suited to actually tease out these effects.
This graphic is at best uninformative, but at worst deceptive in its implicit message. One would expect more from an administration based on “openness”.
Automatically Generated Related posts:
It’s not that bad. The zero axis is at zero, the data is not massaged in any way that I can tell. A minor issue is that there should be a line at around +100,000 to indicate the number of jobs needed to accommodate population growth. The main problem with it is that it doesn’t prove anything, it’s just showing the correlation between administration and delta-jobs. To show causation requires a much more sophisticated argument.
Also, everyone knows that Obama’s the socialist, so his bars should be red, not Bush’s!
[Reply]
Drew Conway Reply:
February 17th, 2010 at 5:40 pm
From a purely visual perspective, you are right, I think there is nothing that bad. Also, I thought Obama was an elitist, so shouldn’t his bars be purple?
[Reply]
I assume you are talking about the fact that this is representing change in slope (i.e., jobs are not being lost as fast) as a change in direction (i.e., the Obama side is going up, while the Bush side is going down).
It would be more appropriate to show a line chart of jobs (or, conversely unemployment) leveling out after a steep drop.
[Reply]
Drew Conway Reply:
February 17th, 2010 at 5:40 pm
Indeed, but I think the fact that it is time-series data means we need to do more than look at just slope.
[Reply]
I’ll point out the obvious: The Bush years are a sinister, dark, shaded red, while the Obama years are an airy blue evocative of a beautiful blue sky. Good information graphics aim to present information without insinuations, leaving the task of evaluation to the viewer. This graphic is clearly trying to vilify the Bush administration through visual cues that don’t have a place in an presidential administration trying to depict itself as bipartisan.
[Reply]
Drew Conway Reply:
February 17th, 2010 at 5:42 pm
There is no doubt that colors were done to evoke bias, perhaps the red state/blue state Republican/Democrat?
[Reply]
I thought it was misleading because they are using raw job losses numbers. There are less and less people to loss their jobs as time goes by. I would think dividing the number by actual employed people would show a different picture or maybe just using unemployment percent.
ARIMA went way over my head. Moving on to the basic stats blog now.
[Reply]
The big deal to me is that there is no explanation of what kind of job loss this is… new filings for unemployment insurance? Repeat filings? If it is based on the sum of current cases and new filings for UI, then what amount of the decrease is due to the expiration of unemployment insurance benefits (a periodic decrease that is not impacted by job creation – no matter how many jobs you create, if your UI runs out you are off the rolls which is often falliciously construed to be an improvement in job losses)?
Honestly, it is a typical USA Today chart, minimal data, maximum impact. Without knowing what data they are actually using, it is meaningless blather.
It is pretty, but useless.
[Reply]
Michael, the chart is just a repackaging of these numbers: http://data.bls.gov/PDQ/servlet/SurveyOutputServlet?data_tool=latest_numbers&series_id=CES0000000001&output_view=net_1mth . It’s the one-month change in total non-farm payrolls, seasonally adjusted, and it’s a perfectly reasonable snapshot of the changes in employment in the US economy. What it’s NOT is the results of a causal model of Obama policies vs Bush (or McCain) policies.
[Reply]
Drew Conway Reply:
February 18th, 2010 at 9:37 am
But don’t you think the causal link is implied, which makes it very disingenuous?
[Reply]
Harlan Reply:
February 18th, 2010 at 9:49 am
I think it’s a first step in exploratory data analysis. Hey, look! We had this horrible thing happen in 2007-2008. It seems to have stopped getting worse at around January 2009, when the following 17 things happened: TARP, Obama Inauguration, etc, etc. Now let’s go see if any of those things could possibly have affected the shape of the data. As it turns out, economists who are much smarter than I am pretty much agree that the recovery is faster than it would have been if not for a variety of policies enacted by the Bush and Obama administrations in late-Fall 2008 through Spring 2009.
The most politically honest thing to do would be to identify changes in policy, not administration. And the biggest changes in policy happened in October/November 2008, due to the shock of the credit freeze, not so much after Obama was inaugurated.
My primary change to the graph would be to indicate where significant policy changes occurred, including various programs of the Fed, the Recovery Act, the auto bailout, etc.
[Reply]
Drew Conway Reply:
February 18th, 2010 at 10:15 am
My inclination is that none of those exogenous things had much of an affect, but rather this was just the natural movement of the economic system back into equilibrium.
Alas, we would never know unless we specified a really good model. Perhaps a nice secondary talk for the next R meetup?
Harlan Reply:
February 18th, 2010 at 10:24 am
I’d be fascinated by a talk on macroeconomic modeling (with R). I know very little of it, but from what I’ve learned recently, if I had taken econ in college I might be an economist now!
I think it’s possible you’re right, but I think it’s also possible that if none of those exogenous things had been done, that the credit freeze would have worsened and we would have had 20% unemployment now. That is to say, it may be that the main effect of the various government programs might have been like insurance, to drastically reduce the risk of the worst-case, not so much to improve the average-case or best-case.
Mr. E Reply:
February 20th, 2010 at 5:49 pm
Drew, you’re inclination would be wrong. Look to any reputable economic website that or economist, they all agree that spending helps.
You’ll find the same once you unleash R on the problem.
jd long Reply:
February 24th, 2010 at 2:54 pm
Hey Mr. E: It’s weak to assert that something is wrong and others should go look it up and prove your point. There is no consensus that Keynes was correct. Actually there’s probably a strong tide in the opposite direction saying that excessively low interest rates over a prolonged period caused mis-allocation of capital. While there may be some reason to believe that gov’t spending was beneficial in the very short term there’s no reason to believe that it was the most helpful for long term stability.
I must note, however, that I really like that Keynes invoked the term “animal spirits” which always make me picture John Maynard Keynes in a cat suit. Meeee-ow!
[Reply]
The biggest question for me on first seeing this is, OK, if you want to justify the implicit assumption of a Bush/Obama difference, show us how this graph extends back 5 more years. Then show a similar graph for another equally large perturbation – maybe something under Clinton or H.W. Bush. I’m as pleased as the next person that we’re not following Bush 2′s policies, but if we’re going to use correlative evidence, then I need a better basis of comparison to show that this is a trajectory that is abnormal following a major economic crisis.
[Reply]
This is a terribly political graphic by design, but the worst elements of the graphic are very straight forward… first, it relies on bad data – the unemployment data that doesn’t deal with people who have given up hope.
Second, it doesn’t show how there has been a bottoming out and we are now bouncing along on some sort of floor, at least for the moment. After all, a plane crashing from the sky that hits water will slow its descent as it begins to inexorably sink, trapped passengers drowning. People without jobs will not be comforted to know that for the moment, businesses have cut the growth and slack positions, and now are slow to cut into the muscle and bone.
[Reply]
The symmetry is incredible! However, I feel that assuming the labour market is autoregressive is a rather strange observation.
A problem, as I see it, is that you are assuming the labour market is reaching an equilibrium, based upon two estimations of a single random variable, national absolute employment change (Jan 2008/Nov 2009).
“My inclination is that none of those exogenous things had much of an affect, but rather this was just the natural movement of the economic system back into equilibrium”
The chart itself considers absolute changes in employment, and tells us nothing of the economic system (namely public/private sector output). The chart, in of itself, does not consider labour market volatility – which could actually be increasing since the start of 2008.
[Reply]
But can we assume that this particular shock would have naturally diminished over time, or at least over this time frame, absent stimulus measures?
This, indeed, was the crux of the entire policy debate in 2008-09: whether the credit crunch, and financial sector failures, might instead have a cascading effect that would see the recession accelerate rather than slow.
Of course, one could argue that one would expect to see more lag in the job numbers than the graph indicates, especially given how long it actually takes to spend stimulus funds. On the other hand, the package of initiatives launched in the late Bush and early Obama periods were intended not only as Keynesian spending-led economic stimulus, but also as signals to markets and consumers, and as a way of calming financial sector jitters and reversing excessive risk-aversion in capital markets and bank lending.
In fact, I do agree that the graph takes more credit than the Obama Administration can reasonably claim for countering the recession–although it probably doesn’t overstate the Bush Administration’s responsibility for the situation in the first place.
[Reply]
Drew Conway Reply:
March 2nd, 2010 at 10:22 am
Good point, we have no counter-factual, but such is life.
[Reply]
I don’t know which is more comical – the people lamenting that a single graph tells an incomplete story or the people surprised that something political would come out of the White House.
That said, as far as rhetorical use of numbers go, this is no high-crime. At least this White House isn’t giving us outright laws… at least not yet.
[Reply]
While I agree this infographic has its flaws and mainly is politically motivated (imagine!), I don’t think the ‘red vs. blue’ comments are taking into account that for at least the past 10 years (I can’t recall when it started) red represents the GOP and blue represents the Democratic Party.
Of course, they could be making a statement about how the Bush II admin has blood on their hands, but that’s a different topic altogether, and I can’t speak for them anyway.
[Reply]
I’m amazed by the fact that people consider this as valid criticism of the Obama administration.
Fact 1: There was an economic meltdown weeks BEFORE Obama took office.
Fact 2: The economic meltdown was a result of irresponsible lending in the private sector.
Fact 3: The Bush administration campaigned and worked for less restrictions on financial institutions, including but not limited to lowering what’s called the reserve ratio, which essentially increases the financial institutions’ ability to take risks with other people’s money (legally).
These are the FACTS of the matter, and they are uncontested by anyone that I’ve ever spoken to, regardless of political positions. Either people don’t know about them, or they agree. There is no dispute about this amongst those who paid just a little bit of attention while it was happening.
Saying that Obama is responsible for this development is just about as reasonable as blaming him for the perpetual told-you-so in Iraq. It’s completely bonkers.
[Reply]
There are several issues. One is the use of color which implies that the right hand data are “nicer” than the left. In a technical sort of way, I suppose it is true, but don’t try to tell that to someone whose life is a part of these data. More seriously is the implication that things are nearly as good as when the problem began (that we have returned to the state that previously existed before the shock. A more realistic presentation would show that while the rate of loss has slowed, we are still in a deep hole that is still getting deeper. Plotting the first derivative is misleading here.
Thus the issue of causality becomes more obvious. Yes, there was a shock, and yes, the rate of loss (of jobs) has been less since Obama took office, but there is no way to decide that the reduction in the rate of loss is the result of Obama taking office (and implementing his economic policies). It seems clear that the effect of any systemic shock will be self limiting unless that shock has destroyed the system under consideration. Thus it would seem that while it may be that Obama has had some effect, the implication that his administration had an effect of the magnitude implied by the graphic is clearly questionable (I would like to assert that it is false, even though I voted for him, but that would be hyperbolic). In fact I agree with many economists that his policies were quite ineffective – not to say that they were without effect, but once one takes account of self-limiting aspects of the situation, it appears that Obama’s policies had a small, inadequate effect.
Now this is politics and Obama is going to take whatever he can get. It is clear that he is no FDR or LBJ. In fact he is a rather timid conservative trying to pretend that he is accomplishing more than he is attempting – a dangerous limb to be out on, for sure.
I just found the site and will be back – more obvious intelligence than is found in most other places.
Brad Davis, Ph.D>
[Reply]
Drew:
On the issue of data visualization for the masses, have you seen either of these websites:
http://contexts.org/graphicsociology/
http://www.good.is/departments/transparency/
Lots of fun stuff!
[Reply]
How about his graph instead
http://blog.heritage.org/wp-content/uploads/obamajobsgraph2.jpg
from
http://blog.heritage.org/2010/03/03/in-pictures-the-obama-and-pelosi-job-gaps/
Both the chart in your post and the one in my comment, distort things, by assigning all responsibility to either the president or congress, when of course they both affect things, and even collectively its not like they really control the economy, even though they do have influence.
[Reply]
Here’s another way this chart can be (and is, here) misleading. If you feel there are any shortcomings, imperfections, or misleading aspects to this chart – and clearly you do… You can blame them on the Obama administration, assigning fault to him/them for anything you don’t approve of or agree with. You know, the way you’re accusing the other side of doing to Bush. To be accurate, however, you should be complaining about Nancy Pelosi here. Her office prepared and released this chart, not the Obama administration at all. Yet another example of how someone can use this chart to mislead and mis-assign blame. Nice work!
If you bring up this chart again, try to say “Nancy Pelosi’s office released this chart” rather than “The Obama administration released this chart”. In the interests of fairness and accuracy.
[Reply]
Drew Conway Reply:
April 14th, 2010 at 3:11 pm
Except that the chart was first published on the Obama’s blog: http://my.barackobama.com/page/content/recoveryanniversary/
…in the interest of fairness and accuracy.
[Reply]
Well first off, the independent variable is “job loss”, and most of the data points are negative, so if you take the chart literally, then there was negative job loss so there should be more jobs now than in 2008(which we all know is false).
Secondly, once the red and blue are put in, and the chart is completely separated by whose administration this was, then this is no longer a clear informative chart with unbiased data about job loss over time. This is now a form of propaganda by the Obama admin to tell you that they are better than Bush.
Thirdly, This chart is showing job loss and not total number of jobs, so they are showing that since 2009, the loss in jobs has almost stopped, but that is not an impressive fact since there can’t be job loss forever. That is nearly impossible. So eventually job loss was gonna stop no matter what, whether Bush or Obama or even Flava Flav were runnin the show. A more informative chart would be a chart of total number of jobs, but that would show that jobs have been lost since Obama and wouldn’t relfect well on Obama, so they didn’t make that chart. That is called Propagana.
Fourthly, this chart is implying that the second Obama got in office we started losing less jobs as if he started passing bills the next day to help our economy. That didn’t happen though. Bush’s policies were still in effect when we started losing less jobs, so Obama can’t take credit for that.
I could go on, but you get the point. This chart is Propaganda and B.S.
[Reply]
I think most people are ignoring the fact that it is exactly the credit crisis that the chart is trying to address. You assume that something like the credit crisis will naturally resolve itself and that things will get better on their own. What the world has seen is that “on their own” the problems got worse until intervention was necessary. Is it a little presumptuous for President Obama to take full credit for resolving the crisis? Of course, President Obama is not even a legislator and did not make the laws that addressed these concerns. But he did support them, and the one thing that you absolutely can take from this chart is that, based on actions taken by the Legislature, the problem did not get any worse, which was and should have been a real fear. But as much as it is a problem for the President to take credit for Legislative success (especially fiscal legislation), it is equally a problem with those of us who look at the president and blame/congratulate him (and eventually, possibly her) for fiscal success. This is especially so, when the President’s political party matches the majority in the House and the Senate, when the Presidential veto doesn’t even come in to play (for the most part).
[Reply]