« All else is equal except that one is more equal than the other | Main | Hat tip to the Times editors »


Feed You can follow this conversation by subscribing to the comment feed for this post.


Terrific post Kaiser. I share your frustration with these reporters who cannot seem to understand the data that they report on.


Only about a half of US college+ graduates who get a job have one that requires a college+ qualification. This is better than not having a college+ education where a large proportion won't get a job at all, but it just shows that people who go to college are more employable irrespective of their degree, which is what you would expect. Even the worst of graduates are reasonable at counting out change, using computer software and being sort of organised.

What I hate in Australia is that the commentators ignore the statistical variability. Because we have a smaller sample our 95% confidence intervals are about plus or minus 0.1%, so the monthly change is almost always not significant.

Andrew Gelman


Can you elaborate on your point in which you said it was "inexplicable" when the reporter discussed the cold and snow? I took it to mean that there was unusually cold weather in Dec but not in Jan; thus cold weather could somewhat explain the poor economy in Dec but not in Jan. But if that's what they were saying, what's "inexplicable" here?


Andrew: It sounded to me like "Fool me once, shame on you. Fool me -- you can't be fooled again." If they believe that snow and cold is a cause of unemployment (which is incorrect given the counting rules), then it should apply in both December and January. If they realized in January that the model is wrong, then they should realize that it was wrong in December too.
If what they meant to say was that there emerged another mitigating factor in January which did not appear in December, then they should tell us what that factor is. In that scenario, it would be inappropriate to claim that snow and cold did not have the effect in January.


Ken: The problem is to take that observation and turn it into a policy that says make more college graduates now and arguing that such a policy would fix the unemployment crisis. College grads are displacing non college grads in many jobs that don't require college education. Anecdotally, you have college grads working in Starbucks.
In addition, those statistics are not cohorted. New college graduates (also law school grads, etc.) are having trouble finding jobs--in the States, they are also burdened with debt, which compounds while they are unemployed.
And yes, the month-to-month changes in the jobs reports are almost never significant but ironically, that's how Wall Street can turn it into a casino.

Adam Schwartz

Generally, I agree with the inadequacy of the NY Times analysis, but it doesn't get much less foggy just by adding labor force participation, no? A few thoughts.

One, by not starting the Y axis at zero, doesn't the FRED chart commit one of the typical sins of displaying data - exaggerating the rate of change in the Y axis? Is a slide from 66% to 63% really that steep? How might we assess this? For example, what if we compare data from the world bank on labor force participation rates in many countries?


I didn't filter any countries out, though I removed various aggregations of regions. One might take it a step further and attempt to compare "like" countries to the US, whatever that means.

The median labor force participation in that data for 2012 is 63.4%. While the median change from 2014-2012 was approx. +0.89%, I'd roughly estimate that the distribution places most of the changes between -3% to +9%. The US change isn't so extreme compared to all other changes, and when you consider it brings US labor force participation close to the median by country, is a 63% participation rate so out of line then? For example, China's labor force participation dropped slightly faster than the US did during the same period.

I guess more importantly, not to absolve the NY Times, but do we expect any newspaper to explore every single potential angle? There'd be no such thing as a short news story any more. :)


Adam: I knew someone would complain about the dual axis plot. If you read my other blog, you know I'm not a fan of it. That's why I specifically said that the only thing you should look at is the the point in time (start of 2008) and look at the trends before and after that point in time.

Also, speaking about the labor force participation rate in particular, the steepness of the decline is not exaggerated. The pre 2008 period can be used to establish the normal fluctuations in that particular metric and it is very stable, barely moving beyond 66 percent. If you are responsible for forecasting this metric, and you are at Q4 2007, you would have predicted the rate to be at 66 percent. There is no doubt about that. The fact that it has been on a unidirectional decline for five years sinking to 63 percent is extremely significant. In absolute terms it may not be. But you have to judge movements based on the background variability.

Regional/international analysis just brings in more noise that does not pertain to the question. I'd not go there at all. There is no need for a trend that is so clear. The fact that the Times keep reporting the U3 rate without talking about U6 or labor participation rate is a black mark on the paper.

The comments to this entry are closed.

Get new posts by email:
Kaiser Fung. Business analytics and data visualization expert. Author and Speaker.
Visit my website. Follow my Twitter. See my articles at Daily Beast, 538, HBR, Wired.

See my Youtube and Flickr.


  • only in Big Data
Numbers Rule Your World:
Amazon - Barnes&Noble

Amazon - Barnes&Noble

Junk Charts Blog

Link to junkcharts

Graphics design by Amanda Lee

Next Events

Jan: 10 NYPL Data Science Careers Talk, New York, NY

Past Events

Aug: 15 NYPL Analytics Resume Review Workshop, New York, NY

Apr: 2 Data Visualization Seminar, Pasadena, CA

Mar: 30 ASA DataFest, New York, NY

See more here

Principal Analytics Prep

Link to Principal Analytics Prep