Several Ways to Look at US Unemployment Data

The US unemployment rate was down to 4.4% in April, which is the lowest we've seen since before the big recession (about 10 years ago). But a single number seldom tells the whole story, so let's look at unemployment data in several different ways, to get a more complete picture...

My quest for better unemployment graphs started back in 2011, when I saw the following graph in the news. I think it must have been created by an artist (rather than software), because there's no way the 8.6% (which I circled in red) would be lined up with the 9.0%:

I decided to create my own version of the graph using SAS software, so that the values would be accurately represented in their correct locations. Here's what I came up with --my version shows that the 8.6% should be far below the 9.0% when the axes are scaled this way:

But you probably also noticed the big red 'X' through the graph (which is something I do to bad/misleading graphs in my blogs, so people don't mistakenly think they are good examples). The problem with this graph is that axes make it look like there was a huge drop ... when it wasn't really that huge in the grand scheme of things. Here's the same data plotted with the y-axis going to zero, which I think is what the news was trying to show in the first place:

The graph above is a reasonable representation of the 2011 data, but why just show one year of data? Wouldn't it be even more interesting and informative to show lots of years of data -- that way we can compare the current recession to previous ones, etc? I think a graph like the following is much more informative:

In the above graph, we can see that the unemployment rate at the beginning of various recessions was different. Therefore it's difficult to compare the recessions, and how long it took to recover back to the pre-recession unemployment rate. The following graph makes it easier to compare recessions in that regard:

I had heard that the 2008 recession was different from previous recessions, because it took the unemployed people longer to find jobs. The following plot shows the number of people unemployed for 27 or more weeks:

Some claim that the traditional unemployment number is lower than the true number of people who are unemployed. The Bureau of Labor Statistics has some "alternative unemployment rates" that might help with that claim. Looks like the number is indeed much larger if you include the unemployed part-time workers, for example:

Another number you can look at, instead of just unemployment, is the civilian labor force participation rate. Some of the fluctuations in this number can be explained by women & baby boomers entering the work force, and recently the baby boomers starting to retire.

And possibly more important than how many people are unemployed, is where people are unemployed. This wouldn't be such an issue if the unemployment rate was fairly constant across the nation, but there are definitely some areas that have much higher rates than others. Here's a snapshot of the latest data, by county (you can also click this link to see the map animated over time).

What did I use to create all of the above graphs? ... You guessed it -- SAS Software!

I know there are a lot of analysts out there with much more in-depth knowledge about unemployment data than me. Do any of you have any special insight, or favorite ways to graph the data, that you'd like to share? Feel free to leave a comment.

This blog was reposted from the SAS Learning Post. Go there to view the original.

Robert Allison, The Graph Guy!, SAS

Robert Allison has worked at SAS for more than 20 years and is perhaps the foremost expert in creating custom graphs using SAS/GRAPH. His educational background is in computer science, and he holds a BS, MS, and PhD from North Carolina State University. He is the author of several conference papers, has won a few graphic competitions, and has written a book calledSAS/GRAPH: Beyond the Basics.

What Cities are in Hurricane Irma's Path?

Here's an example of using data and visualization to look at weather -- specifically, the possible path of Hurricane Irma. Does your city need to get ready?

Mapping Out the Next Robot Invasion

Where are all the robots today? Here's a look at a better data visualization to represent where in the US all the robots are.

Re: BLS Unemployment Info
  • 7/6/2017 5:29:53 PM

You're right again, Joe. I recall a time or two when I have been rigorous in my biases ;->

Re: BLS Unemployment Info
  • 7/5/2017 10:04:53 AM

@T Sweeney: Rigor, sure, but one can think one is being fully rigorous while still being subconsciously guided by the very subjectivity that data science seeks to discount and eliminate.

So yes, outreach and discipline are key -- and part of that thoroughness is a commitment to some level of collaboration. Otherwise, you're just patting yourself on the back.

Re: BLS Unemployment Info
  • 7/4/2017 10:05:08 PM

I think what we're talking about here, Joe, is rigor -- thoroughness, discipline and outreach to pull apart methods and conclusions. Not every organization has the luxury of devoting tons of time to this, but a commitment to rigor has to make some difference.

Re: BLS Unemployment Info
  • 7/3/2017 5:50:08 AM

@T Sweeney: Dead on. Moreover, a big part of that is checking your work against that of your peers -- and, if your answers are dramatically different, putting your heads together to account for the difference.

Part of this relies on diversity in collaboration. Not just "diversity" in the classic ways we think about it as related to protected and semi-protected classes, but also diversity of opinion and philosophy.

In an earlier comment, you mentioned the polls leading up to the 2016 Presidential Election. The funny thing is that, for all the people on one side who were shocked by the results, people on the other side weren't so much surprised. Two particular polls with the best/most accurate results over time were predicting a Trump win -- and one of the only major news sites bringing attention to these polls was the Drudge Report (a news aggregator site that is famously right-leaning). Many of the relatively few people paying attention to media narratives on both sides of the political spectrum were, accordingly, prepared for either result in what was a genuinely close race.

Re: BLS Unemployment Info
  • 6/28/2017 11:00:09 AM

I wish that all analytics users could scrutinize their own datasets (and conclusions) with as much granularity as we're able to look at unemployment figures. A more critical eye (and some grounding in statistics) can only help improve the outcomes of analytics projects.

Re: Graph Bad
  • 6/27/2017 11:05:26 AM

> and Hyperbole as a Service (Haas).

Absolutely! The sky's the limit!  ;)

Re: Graph Bad
  • 6/26/2017 6:30:20 PM

@Seth: Maybe the course was some kind of meta-test to see which student would be the first to act rudely. ;)

Re: Graph Bad
  • 6/8/2017 7:02:55 PM

Ouch! Dropping the class was the right move. I don't think I would have made it through that class...

Re: BLS Unemployment Info
  • 6/8/2017 3:30:33 PM


Terry writes

Given the data that factors in to unemployment rates, not to mention how the mix of that data has changed over the years, it's best to treat unemployment figures as a general barometer and not a laser-sharp snapshot of the job market. Bit as we know from experience, presidents and Wall St. traders will latch on to any data bits for justification and to score polling points or major $ trading action.

I agree totally. Unemployment numbers are often treated on a par with, say, data figures for the CERN accelerator, whereas in reality they're more in a category of very wobbly or blurry data associated with the social sciences. A 0.1 point drop in unemployment rate might be hailed by the current administration and its cheerleaders as a huge accomplishment. But in reality it could be just a rounding area, or a fluke resulting from more jobseekers becoming hopeless and ceasing to even look for work ...


Re: BLS Unemployment Info
  • 6/8/2017 1:13:43 PM

All ecellent points, Lynson. Given the data that factors in to unemployment rates, not to mention how the mix of that data has changed over the years, it's best to treat unemployment figures as a general barometer and not a laser-sharp snapshot of the job market. Bit as we know from experience, presidents and Wall St. traders will latch on to any data bits for justification and to score polling points or major $ trading action.

Page 1 / 8   >   >>