Google Maps + Census Data

September 14, 2005 1:03 PM

Google Maps and Census Data. Navigate to a place via Google Maps. Click and you get Census demographic data such as population, median income and housing.
posted by caddis (18 comments total)

Interesting. Of course, I hope the 2000 data was collected better than the data in 1980. :)
posted by ?! at 1:19 PM on September 14, 2005

[note]US specific[/note]
posted by blue_beetle at 1:25 PM on September 14, 2005

is there anything that can't be combined with google maps?
posted by mystyk at 1:59 PM on September 14, 2005

I managed to find a place of average income of ~$146,000 just south of Laguna Beach. Can anyone get higher?
posted by geoff. at 2:00 PM on September 14, 2005

Crystal Cove State Park, just north of Laguna Beach - $2254,915.
posted by caddis at 2:03 PM on September 14, 2005

oops
$224,915
posted by caddis at 2:04 PM on September 14, 2005

I spoke too soon, $252,000 near Montgomery Golf Club just north of Manhattan. And $353,000 near the Wee Burn Country Club! I'm surprised my neighborhood is $131,000. Modest houses with two or three car garages and low-end luxury cars. I thought that would start at $80,000.
posted by geoff. at 2:06 PM on September 14, 2005

The averages are being pulled up by the presence of the mega rich, what are the means?
posted by Phantast at 2:16 PM on September 14, 2005

I just clicked on Gates country, $219,901, I would have thought it would have been larger.
posted by mk1gti at 2:30 PM on September 14, 2005

Seattle, Bellevue
posted by mk1gti at 2:30 PM on September 14, 2005

Is HH income a census statistic, or is that info somehow pulled from the tax rolls? If it IS a self-reported stat, I wonder if the mega rich are as prone to lying to census takers (lying about the true amount of their wealth) as the mega poor are (lying about the number of illegal immigrants living in a building).
posted by jonson at 2:48 PM on September 14, 2005

If you go to a less populated area you can really isolate single people. How hard would it be to connect this person to a postal code and an address and a phone number and a name?

Variable 1 Miles 3 Miles 5 Miles
Total Population 0 1 1
Total Households 0 1 1
Median Income 0 $25,679 $25,679
Average Income 0 $35,766 $35,766
Median Age 0 44.80 44.80
Male Population 0 1 1
Female Population 0 0 0
White Population 0 1 1
Black Population 0 0 0
Am. Indian Population 0 0 0
Pac. Islander Population 0 0 0
Asian Population 0 0 0
Other Population 0 0 0
2+ Race Population 0 0 0
Hispanic Population 0 0 0
posted by Rumple at 3:30 PM on September 14, 2005

Thats from rural Idaho, BTW
posted by Rumple at 3:31 PM on September 14, 2005

How can the median and average incomes differ with only one data point?
posted by caddis at 6:16 PM on September 14, 2005

OMG caddis you just made my head asplode. How indeed.
posted by AspectRatio at 9:52 PM on September 14, 2005

If you go to a less populated area you can really isolate single people. How hard would it be to connect this person to a postal code and an address and a phone number and a name?

Not just difficult, impossible.

Now I think what you're suggesting is that if you found a unit of aggregation (say a zip code) with a small number of people you could then look at the various variables and figure there is only one person with X combination of variables and you could find the income (or whatever) for person with X combination of variables and then you would know a specific person's income.

That's something that people who do large scale data collection worry about. First, when reporting aggregated data they suppress any cells contain small numbers of people. So if you had Mean Income reported by Shoe Size and Hair Colour and there were only 15 Redheads with Size 8 shoes, then they would leave blank (suppress) the mean income for Size 8 Redheads. What's more, they would leave other cells so that you can just add/subtract etc. to find the values that should be in the missing cells.

If you got your hads on the data itself (non-aggregated) you still wouldn't be able to do what you suggest. Why? Because they deliberately switch people's values for different variables. So what they do is they would take two people in a census block and swap their income values (or values for another variable). Because they're swapped, the mean/median etc. income for the block is still accurate, but you don't know for any individual if that is their real income or one that was swapped with someone else. They do this here and there for each variable. Because the number of people sampled is so large, the effects on correlations and other statistics is negligible, but it protects the privacy of respondents.

This second method of protecting privacy (Data-swapping) seems the most relevant here. You have no idea if that income is correct for that person, or for that matter if the other variables you have are correct. They could be swapped.

Also, that's assuming this really does come from individual level data. If it's being somehow parsed out of aggregate data (note that that age is 44.8.... while a person can be 44.8 years old and this would be simple enough to calculate from birth date, I bet the census data sets record age in whole years for individuals as is the convention), then there's a whole other set of problems with accuracy.

Anyway, rest assured that there's no way in hell the census bureau is letting out data that could be used to identify individuals.

Here is the census bureau description of how suppression and data swapping are done.
posted by duck at 9:28 AM on September 15, 2005

I wish I could mark that as "best answer" duck.
posted by caddis at 9:41 AM on September 15, 2005

I hate being in the income bracket pulling the average down. I blame Bush. /flee
posted by cleverusername at 2:53 PM on September 15, 2005