Mind Poison

Econopundit strikes again. When I read this, I immediately thought "What are you talking about?" His claim that there hasn't been a monthly decline in job growth since April is simply not true. Here's the payrolls data from January 2003 on:

Date	Payrolls (Thousands)	Change from Preceding Month
Jan-03	130190	94
Feb-03	130031	-159
Mar-03	129921	-110
Apr-03	129901	-20
May-03	129873	-28
Jun-03	129859	-14
Jul-03	129814	-45
Aug-03	129789	-25
Sep-03	129856	67
Oct-03	129944	88
Nov-03	130027	83
Dec-03	130035	8
Jan-04	130194	159
Feb-04	130277	83
Mar-04	130630	353
Apr-04	130954	324
May-04	131162	208
Jun-04	131240	78
Jul-04	131272	32

So clearly there was a decline in payrolls until September 2003. What he's done in his chart is pick the exact way the data seems like it says what he says it says. So instead of showing the absolute change from period to period (what most people look at), he shows the percent change from period to period. Here's what the percent change from period to period looks like:

Date	Percent Change from Preceding Month
Jan-03	0.07%
Feb-03	-0.12%
Mar-03	-0.08%
Apr-03	-0.02%
May-03	-0.02%
Jun-03	-0.01%
Jul-03	-0.03%
Aug-03	-0.02%
Sep-03	0.05%
Oct-03	0.07%
Nov-03	0.06%
Dec-03	0.01%
Jan-04	0.12%
Feb-04	0.06%
Mar-04	0.27%
Apr-04	0.25%
May-04	0.16%
Jun-04	0.06%
Jul-04	0.02%

Right here, you can see why people don't look at the percentage changes. If gangbusters payroll growth is March 2004, and that's only 0.25% percent growth, what is the point? The payrolls seem to only grow by certain amounts each month over time (around 200 to 300 thousand a month in times of expansion), but since jobs accumulate, a "good" payroll increase is a decreasing percentage change over time. In other words, a 250,000 payroll increase 30 years ago is a larger percentage change than the same increase today, but it's still just as good of an increase in employment. It's kind of surprising, actually. This makes the percent changes uninformative.

But ignore that. It still doesn't look like his claim is correct, does it? So he applies a little more fudge, by rounding the growth rates to one digit. Well, then you get his chart.

Date	Percent Change from Preceding Month	Percent Change Rounded to 1 Digit
Jan-03	0.07%	0%
Feb-03	-0.12%	0%
Mar-03	-0.08%	0%
Apr-03	-0.02%	0%
May-03	-0.02%	0%
Jun-03	-0.01%	0%
Jul-03	-0.03%	0%
Aug-03	-0.02%	0%
Sep-03	0.05%	0.1%
Oct-03	0.07%	0.1%
Nov-03	0.06%	0.1%
Dec-03	0.01%	0%
Jan-04	0.12%	0.1%
Feb-04	0.06%	0.1%
Mar-04	0.27%	0.3%
Apr-04	0.25%	0.2%
May-04	0.16%	0.2%
Jun-04	0.06%	0.1%
Jul-04	0.02%	0%

As you can see, not only does this now make it look like there wasn't actually a decline in jobs in those first five months after April 2003, but it also exaggerates the growth for September through December, by almost doubling the growth rate in those months. He's hidden a decline of over 130,000 jobs and called it "zero" because it's broken up into enough periods that the percent changes were rounded away by the arbitrary precision he chose.

But even still, all this isn't really the point. The working population is growing all the time, and so people enter the labor force. That means that simply having zero jobs growth is actually a decline in employment in percentage terms. In fact, economists estimate that you need about 150,000 new jobs a month just to keep up with population growth. So a smaller monthly increase isn't actually anything that people get excited about. From this perspective, only March, April, and May (and November, barely) are consistent with an increase in real employment.