Shocking News: Democrats Cherry-Pick Stats to Make Trump Look Bad

CNN and FiveThirtyEight Assist

Not that it matters, but while perusing an article from the Intercept demonstrating Mike Bloomberg's plagiarized campaign literature, I came across:

In the four years prior to the passage of the GOP tax law, the economy added an average of 213,000 jobs each month, according to data from the US Bureau of Labor Statistics. In the nearly two years since the law passed, average job creation has actually declined by an average of 11,000 per month.

In case you're wondering, that's a quote from both Bloomberg and this December 2019 article at CNN from one Michael Linden. Verbatim.

Boy, doesn't it sound like they're claiming that jobs have actually been destroyed at a rate of 11,000 per month? For the past couple years? That would be big news, and it's shocking that it hasn't been trumpeted far and wide…

Well, wait a minute.

The Bureau of Labor Statistics (BLS) provides its data ("CES0000000001, All employees, thousands, total nonfarm, seasonally adjusted") here. Their graph:


As long as that line is going up, jobs are being "created". And you'll note a rather steady increase since (roughly) January 2010.

What Linden/Bloomberg are trying to say is that the upward slope of job creation is decreasing. For you calculus fans, the first derivative. Fortunately, BLS provides that data too so we don't have to calculate it ourselves. Here's the graph of that data over the same time period:

[Job creation rate]

That's some noisy data. If you look at the actual numbers:

  • From December 2017 (when the Trump tax cut was signed) to December 2019 (the last data point BLS reports), jobs went from 147.596 million to 152.383 million, an increase of 4.787 million. The smallest monthly change over that period was 56,000 in February 2019; the largest was 330,000 in February 2018. The average job creation rate was about 198K/month.

  • Bloomberg/Linden want us to compare "four years prior" to the passage of the tax law. Fine. Minimum 13K, maximum 339K, average 210K.

So Linden/Bloomberg are actually talking about the difference between 210K jobs created per month versus 198K jobs created per month.

But is that difference in averages "significant"? That's a pretty standard statistical test…

Geez, I haven't done this in a long time.

Fortunately, I don't have to do it myself. There's a nice little online calculator right here. And it says, no, the differences between the two averages is probably not significant, given the natural noisiness of the data. Linden/Bloomberg shouldn't hype it. (Linden might not know better, but Bloomberg should.)

Another site I ran across discussing the jobs data was FiveThirtyEight, offering A Better Way To Think About This Month’s Jobs Numbers. I think this article is updated automatically when the BLS outputs a new report, and it's very nicely done.

But should you get all misty-eyed about the BLS's announced "unemployment rate":

The BLS has a broader measure of unemployment — the U-6 — which includes people counted in the official jobless rate, those who've tried to find a job in the past year but haven't looked in the past four weeks, and part-time workers who want a full-time job. December’s unadjusted U-6 was 6.7 percent. That’s 3.3 percentage points higher than the unadjusted unemployment rate.

Gasp! That darn Trumpified BLS, not telling us the truth about unemployment!

FiveThirtyEight ain't lying, exactly, but they're not providing anything that would just possibly make Trump look good. BLS has been providing U-6 since 1994, apparently. Here's a graph of the "official" unemployment rate (aka U-3, blue line) and U-6 (red line) over that period (via FRED, U-3, U-6).

[Unemployment 3 and 6]

So, yeah: U-6 > U-3. By definition. But what they're not telling us: BLS is also reporting U-6 at a record low.

Don't get me wrong: I'm not a Trump fan. I think his record on deficit spending is awful, his trade wars are stupid. But disbelieve Democrats by default when they try to tell you the economy sucks. It doesn't.