gallery: job plots

Job data during the Bush presidency can be looked at in a few ways; consider this graph in a simple analysis:
In this image, the green line displays the number of jobs gained or lost in a given month for each month in the past three years (raw data, straight from BLS). The black line represents the trend in jobs created/lost,i.e. how many more/fewer jobs were created each month than in the previous. The red line is a linear-best fit to the trend data; the thin black line running through it is the graph's zero-line. And the blue line is the total number of jobs created/lost since Jan '01.

I want to talk about the black and red lines first, so let's zoom in on that for a moment (I know the image is crappy; bear with me):
The first thing to notice is that the black line is very, very noisy: it jumps by hundreds of thousands more/fewer jobs created each month than the one before. The second thing to notice is that the trend is positive about as frequently as it is negative. There was a several-month region (starting around Nov 01) where it stayed positive, but generally speaking it doesn't stay positive or negative for very long. In any given month saying "the trend is positive" or "the trend is negative" is pretty much meaningless since in the next month or so it's likely to reverse itself wildly.

Look at the linear fit to the black data, the red line. It's hard to see, because it nearly falls right on top of the graph's thin black zero-line. This illustrates my point: overall, the trend isn't doing much. In fact, the red line has a very slight *positive* slope -- which is to say that on average since Jan 01, more jobs have been created (or fewer jobs have been lost) in each month than in the previous month. If you wanted to, you could spin this as "improvement"... although the slope is in the hundreds (note the graph depicts hundreds of thousands) of jobs per month. The standard deviation of the data to this linear fit is huge -- on the order of 150K jobs; which means (assuming Gaussian distributed data, which I may or may not have for only 3 years) that we have a wide, flat bell-curve and wild swings (on the order of +/- hundred of thousands of jobs) are not unlikely.

Let's go back out for a second and look at the blue line: it's the cumulative number of fewer jobs since Jan '01. Pretty striking, isn't it? It illustrates well how huge a jump in jobs Bush need to get back to where we were when he took office. It's not noisy at all, because of the nature of integration (conversely, a derivative of the trend line -- the change in the change in the change in the number of jobs in our economy from month to month -- would be yet spikier); it clearly suggests an economy in the toilet. It also suggests that the economy had some inertia -- there was still some job growth in the beginning of 2001 -- and that jobs took a nose-dive well in advance of 09/2001 and thus can't entirely be blamed on terrorism (these conclusions, too, are based on a handful of months inside a much larger picture).

There's no question that if I were a stats wrangler for the Dems, I'd be using the blue line. There's also no question that if I were a stats wrangler for the Reps I'd be using the red line and a fit to the green line (which isn't in the graph, but is going [slowly] upward since late '01; I'd say that the economy is growing -- fewer jobs are being lost and now jobs are even being created -- and that that growth is, in fact, accellerating... Since a voter's conclusion will be based on which of these two depictions stick with him, the choice of depiction presented by a given economist or journalist is a political one.

Data source: Bureau of Labor Statistics. Trend (differences), cumulative sums, and linear fit done with xmgrace. Caveat: 36 events is unlikely to be enough to see a Gaussian distribution; a real assessment of the variance in the trend should be done with many years of data (which BLS has, but I'm lazy), not just 3.