Does anyone know if they looked at how well the model they developed predicted demographics in low car ownership areas like New York City, where less then
half of the households have cars? It did not seem like they used any data from the NY area. In built up, high density areas, many of the cars are in structured parking or people do not own cars. On-Street cars may not represent who is living in the area.
Same for high density areas in other cities.
From: ctpp-news [mailto:ctpp-news-bounces@chrispy.net]
On Behalf Of Krishnan Viswanathan
Sent: Friday, December 01, 2017 10:15 AM
To: Ed Christopher
Cc: ctpp-news@chrispy.net
Subject: Re: [CTPP] The Washington Post: Scientists can now figure out detailed, accurate neighborhood demographics using Google Street View photos
Totally agree with you Ed. I think the Washington Post headline is a dose of hyperbole. I have a question about spurious correlations and as far as I know the ACS does not ask what type of vehicle is there in the household so this paragraph
in their paper gave me pause:
Using ACS and presidential election voting data for regions in our training set, we train a logistic regression model to estimate race and education levels and
a ridge regression model to estimate income and voter preferences on the basis of the collection of vehicles seen in a region. This simple linear model is sufficient to identify positive and negative associations between the presence of specific vehicles (such
as Hondas) and particular demographics (i.e., the percentage of Asians) or voter preferences (i.e., Democrat).
On Fri, Dec 1, 2017 at 9:38 AM, Ed Christopher <edc@berwyned.com> wrote:
Interesting stuff Krishnan--
If their basic assumption were true, that vehicle ownership somehow translates into demographics (ie voter behavior), then why not just cut Google out and process vehicle registration files. It seems to me that would be a lot easier and cheaper. Then again
you have to buy the basic assumption. Also, if you look at precinct by precinct voter behavior two things will surprise you. First, precincts are not all one color (red or blue) in most places and the number of people who do vote are very small when considering
the total population. While I found this work interesting I would not be out their trying to oversell what its capabilities are without a whole lot more work and research. As I see it we have a very long way to go before we have something that is a kin to
the ACS and all its by-products.
On 11/30/2017 8:39 PM, Krishnan Viswanathan wrote: