Frank Hansen writes:
Columbus Park is on Chicago’s west side, in the Austin neighborhood. The park is a big green area which includes a golf course.
Here is the google satellite view.
Here is the nytimes page. Go to Chicago, and zoom over to the census tract 2521, which is just north of the horizontal gray line (Eisenhower Expressway, aka I290) and just east of Oak Park. The park is labeled on the nytimes map.
The census data have around 50 dots (they say 50 people per dot) in the park which has no residential buildings.
Congressional district is Danny Davis, IL7. Here’s a map of the district.
So, how do we explain the map showing ~50 dots worth of people living in the park. What’s up with the algorithm to place the dots?
I dunno. I leave this one to you, the readers.
Looking at other areas, I think they just randomly distributed the dots for a given census tract.
Looks to me like the placement of dots within the census tract is just random, paying no attention to the distribution of population within the census tract. This is pretty apparent when I look at the tract I live in. The dots are randomly distributed but the population is not.
This is also a lot easier to implement, rather than graphing the actual location of each 50th person.
As usual, it helps to look at an area you are intimately familiar with, rather than jump to conclusions about the homeless on the west side, or ghost voters.
My census tract in Utah contains mostly undeveloped National Forest land with a small corner occupied by houses. The dots for our tract are distributed somewhat uniformly across populated and unpopulated areas.
It's all the algorithm.
Zoom out to the county view for an large county with urban and rural areas, like King County, WA. At a resolution of 2,500 persons per dot, you see the density concentrated in the city. Zoom out to 5,000 persons per dot, you see the population dots spread evenly across the county.
There was a similar problem with Eric Fisher's maps that you linked to in your "how segregated is your city?" post. Eric explains why his maps include some people in the middle of the Potomac:
"The problem you are seeing in the river is that the Census data gives population by block, and, it looks like, the entire Bolling Air Force Base is one block. The Summary File does not actually include the shape of the block, only a point within it and its area, so I plot points within the radius from that point that would yield that area. For most blocks this works pretty well, but for large blocks that are much longer in one direction than the other, they can be pretty far off."
http://www.flickr.com/photos/walkingsf/4981417821…
Both these examples just go to show that data organized for high-level viewing should be taken with a grain of salt at lower levels.
I believe the census bureau does that on purpose, ie fuzzes the exact location, so that data miners cannot discover your age and race by clicking on your dot on the map.
Agree with zbike. A simple spatial operation on the server-side could resolve the issue: intersect unpopulated places (parks, water, etc) to census tracts, removing tract number where intersection.
z is correct, go a bit east and you'll see similar patterns with the tracks that cover the big parks and a few condos. It also places dots in the middle of the rivers.
So the NYTimes headline on that page — "Mapping America: Every City, Every Block" — is a little misleading. Also, they take the trouble to not put dots in Hudson River around NYC, but again there are dots in smaller lakes north of the city.
Although not my current neighborhood, I'm not unfamiliar with that park. When I looked at my current neighborhood it looked like discrepancies could have been due to sampling or "small" vagaries of the drawing algorithm. The park was big enough with enough dots that it seemed different. Regarding my possibly thinking about the political implications, I'll take the 5th.
It looks like the dots are drawn anew each time you load the page.
Hm, if it's not an artifact then you should wander up to Fort Tryon Park (near Dyckman – 190th or 200th on the A line) and try to spot the several thousand people living there.
The Cloisters is a damn fine museum too, but perhaps not at this time of year.
I'm sure the algorithmic answers, above, are correct, but my immediate thought was: This is Chicago, are you sure you don't mean "cemetery" instead of "park" and "voters" instead of "residents"?
The "simple server side" solution suggested earlier will deal with many problems, but the multiyear ACS is a new type of spatio-temporal data with new problems. The ACS data do not represent a point in time, but a moving average. How does one define whether a place is populated or not when a large development occurs during the time period? Already, I have seen many newspapers reporting on the newly released ACS data as if they represent the picture at one point in time, when this is absolutely not the case. My local paper reported the 5-year average of "number of housing units built since 2004." How many people can properly interpret this number when it is derived from a 5 year average?
A simple fix in data display would be to limit the zoom scale, so that users could not zoom in too far and view the data at unintended scales. We need to be careful when we report data online, so that people don't do stupid things with it and accuse us of the problem.