If you don’t know already from some of my VizWorld posts, I’m a Flowing Data fangirl. Nathan Yau is the younger, hipper, nerdier Edward Tufte, and one who likes to share his sources and techniques. Understandably, Tufte has his trade secrets, but it was like pulling teeth to get him to share what tools and design methods he uses to make his graphics. Something about Adobe Illustrator and a cadre of assistants is all I got.
Last night, I made a 2009 United States county-specific unemployment map using Flowing Data’s How to Make a US County Thematic Map Using Free Tools tutorial. All you need is a Python installation, the BeautifulSoup XML parser, a good text editor and some patience to debug. (Another reason I like Nathan: He codes in Python, the best, most intuitive programming language out there!)
These are the results, admittedly without a legend (bad Maitri!), which I will work on in Photoshop. So you know what you’re looking at here, the lightest color is 0% unemployment and steps up from there in 2% increments, with the darkest color denoting 10+% unemployment. This data was downloaded from the Bureau of Labor Statistics.
1. The Flowing Data original reproduced:
2. Diverging colors (blue=low; red=high)
3. Sequential colors (white=low; orange=high; black=+10%). The darker the hues, the more trouble folks have telling them apart. Black shows the worst hit spots and provides a backdrop with which to differentiate between the other colors
Check out the original Unemployment, 2004 To Present to see how bad things have become just in the last two years. This isn’t news, but just as well when you look at it in a county-by-county color graphic. The nation is indeed bleeding. Let’s make more casinos at home and start more land wars in Asia!
Cool maps.
The only problem with using counties as the unit of analysis is that sparser the county, the larger it is. And larger blocks of color give an impression that a larger portion of the country is unemployed when in fact the opposite may be true (I’m sure the unemployment rate is high but you get the point, right?)
Patrix, I was thinking about how to further normalize this data. For now, I like the breakdown – that this map gives folks (and local chambers of commerce) an idea how their county’s unemployment number compares with others, and not a whole state’s data lumped together.
Relating to population density somehow would allow you to see if stimulus funds were going in the right direction – if you believe in stimulus funds. Woe is Michigan.
The same data subdivided by major ethnic groups would also be interesting.
I have county-specific census information, presumably from the 2000 census. How does one then accurately normalize percent unemployment based on population density? Per capita? Ideas?
Seems like you could calculate the % of total national unemployment in each county. Would that show anything, or is it circular?
How about this?
Percent Unemployed = Number Unemployed / Number in Workforce
Now I have to find out if that file from BLS bases unemployment percent on total county population or population of county in workforce. If latter, we’re done. Right?