Friday, October 3, 2008

Election for Nerds Special!


Click the picture to enlarge. Further explanation (kinda technical) added below.



I've been wanting to do this for a while...


From left to right is the distribution of Hardcore for Nerds readers from the United States, in terms of the percentage of the readership (via Google Analytics) from each state, relative to its percentage of the population overall. It is expressed in terms of % difference: for example, New York on the extreme left makes up 6.31% of the US population and 10.31% of visits to this blog from the US, hence a percentage point difference of almost exactly (+)4%.


The clear columns going up and down are the distribution of red (down) and blue (up) states, according to FiveThirtyEight.com as of today [Friday Oct. 3rd]. So while Florida (second from the extreme right) may switch to a red state in the end, Alaska on the left is not realistically going to switch to blue. You can adjust the exact data points according to whatever predictions you follow, but the overall trend is clear.

And also, I suppose, fairly obvious to begin with. However, this is the statistical proof* that Hardcore for Nerds is a blue state blog...


* as a politics student and someone who is generally very good at mathematics, I absolutely do not stand over the validity of this exercise in any way.


EXPLANATION: Basically this graph shows a strong correlation between whether a state has above or below average readership of the blog (as against its share of the national population) and whether it is a 'red' or 'blue' state in current electoral terms. (Blue for Democratic - Obama - and red for Republican - McCain, incidentally in complete reversion of the traditional British-Irish/European political scheme, which assigns blue for conservative and red for left-wing or socialist)

The switch from blue to red occurs almost exactly with the point that readership drops from exactly average (no difference between % of readership and % of population) to below average. With the exception of Alaska, all the above-average readership states are blue, leading to the statement above.

On the below-average side, it's a little bit more complicated. Some of them, like Delaware and (I guess) Washington, are firmly in the blue camp. Others are more easily identified as swing states: Ohio, Michigan, Florida and New Mexico.

Please note that the order of the states on the graph is in decreasing positive (and then increasing negative) magnitude of difference between % of readership and % of population. They are not adjusted for population. So a large magnitude of difference does not necessarily represent a huge disparity of readership; rather it represents a relative measure, a function of a) the population size and b) the actual disparity of readership per head.

Therefore the fact of whether a state is close or far away from the centre of the graph (low % difference) changes its significance according to where the state stands in terms of its population size. States that are large should be closer to the edges - if they are not, then they have a more average level of readership than those that are; and small states should be in the centre; if they are not, then they have a high disparity (for example Connecticut, which probably has a lot to do with this blog.)


For bonus marks, you can tell me why this means, statistically speaking, that the three swing states on the far right could be expected to be in that position, or - to phrase the question differently, and a slight hint - why the states in that position could be expected to be (important) swing ones?


A few more notes:

Population data from Wikipedia, of course.

US readership comprises 48% of Hardcore for Nerds readership during the period July 25th-October 2nd, with a total of 7,297 visits.

'Not stated' made up 1.45% of visits, while other US territories make up 1.44% of the total population.


12 comments:

Ape Mummy said...

This is the most cheerfully nerdy thing I've heard all week.

cretin said...

yes, that's the positive way of looking at it.

o_O

*#..(brad said...

eh, uhhh....

far left from nevada does stuff.(?)

gabbagabbahey said...

ape - thanks, yeah that was kind of the intent!

cretin - why wouldn't you want to be positive?

brad - I don't think I understand. there's a far left in Nevada?

lex dexter said...

this last was a very special episode of hardcore for nerds.

cretin said...

i always liked the very special episode of punky brewster where the older girls forced her to try some grass, some uppers, and some nose candy in order to get into their club. but she said no and then nancy reagan came out and gave her a lollipop. probably for the best on her part. i wouldn't want to get mixed up with a pre-teen drug cartel either.

luciferyellow said...

I have a nerdy suggestion to make: Throw out the bars that illustrate Red state/blue state and instead color the names of the states in the respective color. The red/blue state designation is binary information anyway; but at first I kept trying to compare the relative percentage difference to the "value" of the white bars (which is really only 1 and -1, but it took me a while to internalize it) looking for some additional information. This way you keep all the information, but unclutter the graphic (make the blue bars grey or something). Just a thought ;-).

gabbagabbahey said...

hey lucifer, good suggestion. (but)

the graph is quite cluttered, mostly because I'm a bit rusty on Excel and the new Windows interface. I was trying to have the bars blue and red, but that would just have cluttered it up even more!

you're right that it's binary information - I entered it in that way - but it's binary in that it's either positive or negative, 1 or -1, not 1 or 0. so the bars should be there to show them going up or down (a biased interpretation for 'blue' and 'red', of course) and furthermore, in the same direction as the % diff - but not, as you note, with any particular significance in the sizes.

luciferyellow said...

"so the bars should be there to show them going up or down (a biased interpretation for 'blue' and 'red', of course) and furthermore, in the same direction as the % diff -"

Ha,ha yes - downward is evil and so are the red states.
But seriously, you are making my point - or rather a point I wasn't even making before: Whether the bars go up or down is irrelevant, states are either "blue" or "red", it does not add any information with respect to the % difference. So if you make the names of the states red and blue, respectively, you convey the same information (maybe as I would argue even more intuitively), while getting rid of the perception bias of "up" and "down".
Ok, back to work, I should be making some bar graphs of my own ;-) Just poking the bear a bit here

gabbagabbahey said...

agree with you that the names in red and blue could be more intuitive. no idea how to go about it technically, though.

however, whether the red/blue bars go up or down, I think, is neither completely irrelevant nor completely arbitrary. scientifically, it may not be that valid but the visual connection between negative % diff and "negative" political placement works for me. (though not maybe for someone else viewing the graph for the first time)

someone, with an opp. bias, using the same stats to show my readership decreasing in states with *traditional* values, and reversing the polarity (yes, it can be done outside of BTF!) of the red/blue bars could be just as valid.

or, I could just have put up a really simple graph which I worked up a day ago - a pie chart - showing 80% (79.75) of readers coming from blue states, and 20% (20.25) from red! not actually that apparent from this graph, that level of difference! :)

ps. I know I've had you linked up for quite a while, but I should say I like the name and explanation for the name of your blog. quite nerdy! and the blog itself's way hardcore, of course.

luciferyellow said...

Ha, ha, I am proud to say that through my frequent visits here I probably single-handedly keep South Carolina from dropping into the abyss that the other red states occupy in your graph.
And yes, I am a (proud) nerd. The picture at the top of my blog is straight from my microscope at work.

gabbagabbahey said...

-lucifer; I just did some readjusting of the visitor stats to broadband penetration, and on that measure South Carolina moves into the above-average category - and only the second red state (after Alaska) to have above-average readership. well done, I guess!